1 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
4 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
5 paradoxical and zero sized accesses.
7 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
10 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
13 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
15 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
16 to the list of symbol prefixes that must be made linker-
19 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
21 * config.host: Add support for aarch64-*-darwin.
22 * config/aarch64/host-aarch64-darwin.c: New file.
23 * config/aarch64/x-darwin: New file.
25 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
27 PR tree-optimization/103093
28 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
30 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
32 PR tree-optimization/102943
33 * gimple-range-cache.cc (class update_list): New.
34 (update_list::add): Replace add_to_update.
35 (update_list::pop): New.
36 (ranger_cache::ranger_cache): Adjust.
37 (ranger_cache::~ranger_cache): Adjust.
38 (ranger_cache::add_to_update): Delete.
39 (ranger_cache::propagate_cache): Adjust to new class.
40 (ranger_cache::propagate_updated_value): Ditto.
41 (ranger_cache::fill_block_cache): Ditto.
42 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
44 2021-11-05 Richard Biener <rguenther@suse.de>
46 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
47 comment and expand on another one. Combine nested if.
49 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
52 * config/pa/pa-modes.def: Add OImode integer type.
53 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
55 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
56 to UNITS_PER_WORD if IN_LIBGCC2.
57 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
58 negvti2, ashlti3, shrpd_internal): New patterns.
59 Change some multi instruction types to multi.
61 2021-11-05 Jakub Jelinek <jakub@redhat.com>
65 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
68 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
71 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
73 * config/aarch64/constraints.md: Support const in Usw.
75 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
77 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
78 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
80 2021-11-05 Richard Biener <rguenther@suse.de>
82 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
83 (LOOP_VINFO_N_STMTS): Likewise.
84 (vec_info_for_bb): Remove unused function.
85 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
86 Initialize n_stmts member.
87 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
88 (vect_create_loop_vinfo): Do not set loop->aux.
89 (vect_analyze_loop_2): Do not get n_stmts as argument,
90 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
92 (vect_analyze_loop_1): Remove callback, get the mode iterator
93 and autodetected_vector_mode as argument, advancing the
94 iterator and initializing autodetected_vector_mode here.
95 (vect_analyze_loop): Split analysis loop into two, first
96 processing main loops only and then epilogues.
98 2021-11-05 Martin Jambor <mjambor@suse.cz>
100 * ipa-prop.c (compute_complex_assign_jump_func): Remove
101 unnecessary check for RECORD_TYPE.
103 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
105 * doc/install.texi2html: Do not generate old.html any longer.
107 2021-11-05 Martin Liska <mliska@suse.cz>
110 * opts.c (finish_options): Reset flag_gtoggle when it is used.
112 2021-11-05 Jakub Jelinek <jakub@redhat.com>
115 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
116 in CONST_WIDE_INT handling. Adjust recursive calls.
117 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
118 new add_const_value_attribute argument.
119 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
120 add_const_value_attribute argument.
122 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
124 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
127 2021-11-05 Richard Biener <rguenther@suse.de>
129 * tree-vectorizer.h (struct vect_loop_form_info): New.
130 (vect_analyze_loop_form): Adjust.
131 (vect_create_loop_vinfo): New.
132 * tree-parloops.c (gather_scalar_reductions): Adjust for
133 vect_analyze_loop_form API change.
134 * tree-vect-loop.c: Include <functional>.
135 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
136 take struct vect_loop_form_info as output parameter and adjust.
137 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
138 split out call to the original vect_analyze_loop_form_1.
139 (vect_reanalyze_as_main_loop): Rename to...
140 (vect_analyze_loop_1): ... this, factor out the call to
141 vect_analyze_loop_form and generalize to be able to use it twice ...
142 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
145 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
148 * config/rs6000/fusion.md: Regenerate.
149 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
151 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
153 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
156 2021-11-04 Tamar Christina <tamar.christina@arm.com>
158 * config/aarch64/aarch64-builtins.c
159 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
160 ashr_simd, lshr, lshr_simd.
161 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
162 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
163 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
165 2021-11-04 Tamar Christina <tamar.christina@arm.com>
167 * match.pd: New negate+shift pattern.
169 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
171 PR tree-optimization/103079
172 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
174 (gimple_range_calc_op2): Ditto.
176 2021-11-04 Martin Jambor <mjambor@suse.cz>
179 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
180 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
181 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
182 parameter to mark_dead_statements.
183 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
184 (ipa_param_body_adjustments::mark_dead_statements): New parameter
185 debugstack, push into it all SSA names used in debug statements,
186 produce m_dead_ssa_debug_equiv mapping for the removed param.
187 (replace_with_mapped_expr): New function.
188 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
189 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
190 (ipa_param_body_adjustments::common_initialization): Gather and
191 procecc SSA which will be removed but are in debug statements. Simplify.
192 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
194 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
195 when avoiding a copy of an unnecessary statement. Remap removed SSA
196 names in existing debug statements.
197 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
198 parameters if we have already done so.
200 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
203 * gimple.c (gimple_call_static_chain_flags): Handle case when
204 nested function does not bind locally.
206 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
208 * config/aarch64/aarch64.c (aarch64_function_value): Generate
209 a register rtx for Neon vector-tuple modes.
210 (aarch64_layout_arg): Likewise.
212 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
214 * lower-subreg.c (simple_move): Prevent decomposition if
215 modes are not tieable.
217 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
218 Richard Sandiford <richard.sandiford@arm.com>
220 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
221 (v2x4hi_UP): Likewise.
222 (v2x4hf_UP): Likewise.
223 (v2x4bf_UP): Likewise.
224 (v2x2si_UP): Likewise.
225 (v2x2sf_UP): Likewise.
226 (v2x1di_UP): Likewise.
227 (v2x1df_UP): Likewise.
228 (v2x16qi_UP): Likewise.
229 (v2x8hi_UP): Likewise.
230 (v2x8hf_UP): Likewise.
231 (v2x8bf_UP): Likewise.
232 (v2x4si_UP): Likewise.
233 (v2x4sf_UP): Likewise.
234 (v2x2di_UP): Likewise.
235 (v2x2df_UP): Likewise.
236 (v3x8qi_UP): Likewise.
237 (v3x4hi_UP): Likewise.
238 (v3x4hf_UP): Likewise.
239 (v3x4bf_UP): Likewise.
240 (v3x2si_UP): Likewise.
241 (v3x2sf_UP): Likewise.
242 (v3x1di_UP): Likewise.
243 (v3x1df_UP): Likewise.
244 (v3x16qi_UP): Likewise.
245 (v3x8hi_UP): Likewise.
246 (v3x8hf_UP): Likewise.
247 (v3x8bf_UP): Likewise.
248 (v3x4si_UP): Likewise.
249 (v3x4sf_UP): Likewise.
250 (v3x2di_UP): Likewise.
251 (v3x2df_UP): Likewise.
252 (v4x8qi_UP): Likewise.
253 (v4x4hi_UP): Likewise.
254 (v4x4hf_UP): Likewise.
255 (v4x4bf_UP): Likewise.
256 (v4x2si_UP): Likewise.
257 (v4x2sf_UP): Likewise.
258 (v4x1di_UP): Likewise.
259 (v4x1df_UP): Likewise.
260 (v4x16qi_UP): Likewise.
261 (v4x8hi_UP): Likewise.
262 (v4x8hf_UP): Likewise.
263 (v4x8bf_UP): Likewise.
264 (v4x4si_UP): Likewise.
265 (v4x4sf_UP): Likewise.
266 (v4x2di_UP): Likewise.
267 (v4x2df_UP): Likewise.
268 (TYPES_GETREGP): Delete.
269 (TYPES_SETREGP): Likewise.
270 (TYPES_LOADSTRUCT_U): Define.
271 (TYPES_LOADSTRUCT_P): Likewise.
272 (TYPES_LOADSTRUCT_LANE_U): Likewise.
273 (TYPES_LOADSTRUCT_LANE_P): Likewise.
274 (TYPES_STORE1P): Move for consistency.
275 (TYPES_STORESTRUCT_U): Define.
276 (TYPES_STORESTRUCT_P): Likewise.
277 (TYPES_STORESTRUCT_LANE_U): Likewise.
278 (TYPES_STORESTRUCT_LANE_P): Likewise.
279 (aarch64_simd_tuple_types): Define.
280 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
281 (aarch64_init_simd_builtin_functions): Update frontend lookup
282 for builtin functions after handling arm_neon.h pragma.
283 (register_tuple_type): Manually set modes of single-integer
284 tuple types. Record tuple types.
285 * config/aarch64/aarch64-modes.def
286 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
287 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
288 (SVE_MODES): Give single-vector modes priority over vector-
290 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
291 be after all single-vector modes.
292 * config/aarch64/aarch64-simd-builtins.def: Update builtin
293 generator macros to reflect modifications to the backend
295 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
296 Use vector-tuple mode iterator and rename to...
297 (aarch64_simd_ld2<vstruct_elt>): This.
298 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
300 (aarch64_simd_ld2r<vstruct_elt>): This.
301 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
302 iterator and rename to...
303 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
304 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
306 (vec_load_lanes<mode><vstruct_elt>): This.
307 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
309 (aarch64_simd_st2<vstruct_elt>): This.
310 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
311 iterator and rename to...
312 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
313 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
315 (vec_store_lanes<mode><vstruct_elt>): This.
316 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
318 (aarch64_simd_ld3<vstruct_elt>): This.
319 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
321 (aarch64_simd_ld3r<vstruct_elt>): This.
322 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
323 iterator and rename to...
324 (vec_load_lanesci<mode>): This.
325 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
327 (aarch64_simd_st3<vstruct_elt>): This.
328 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
329 iterator and rename to...
330 (vec_store_lanesci<mode>): This.
331 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
333 (aarch64_simd_ld4<vstruct_elt>): This.
334 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
336 (aarch64_simd_ld4r<vstruct_elt>): This.
337 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
338 iterator and rename to...
339 (vec_load_lanesxi<mode>): This.
340 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
342 (aarch64_simd_st4<vstruct_elt>): This.
343 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
344 iterator and rename to...
345 (vec_store_lanesxi<mode>): This.
346 (mov<mode>): Define for Neon vector-tuple modes.
347 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
349 (aarch64_ld1x3<vstruct_elt>): This.
350 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
352 (aarch64_ld1_x3_<vstruct_elt>): This.
353 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
355 (aarch64_ld1x4<vstruct_elt>): This.
356 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
358 (aarch64_ld1_x4_<vstruct_elt>): This.
359 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
361 (aarch64_st1x2<vstruct_elt>): This.
362 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
364 (aarch64_st1_x2_<vstruct_elt>): This.
365 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
367 (aarch64_st1x3<vstruct_elt>): This.
368 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
370 (aarch64_st1_x3_<vstruct_elt>): This.
371 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
373 (aarch64_st1x4<vstruct_elt>): This.
374 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
376 (aarch64_st1_x4_<vstruct_elt>): This.
377 (*aarch64_mov<mode>): Define for vector-tuple modes.
378 (*aarch64_be_mov<mode>): Likewise.
379 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
380 mode iterator and rename to...
381 (aarch64_ld<nregs>r<vstruct_elt>): This.
382 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
384 (aarch64_ld2<vstruct_elt>_dreg): This.
385 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
387 (aarch64_ld3<vstruct_elt>_dreg): This.
388 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
390 (aarch64_ld4<vstruct_elt>_dreg): This.
391 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
392 iterator and rename to...
393 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
394 iterator and rename to...
395 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
396 (aarch64_ld1x2<VQ:mode>): Delete.
397 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
399 (aarch64_ld1x2<vstruct_elt>): This.
400 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
401 tuple mode iterator and rename to...
402 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
403 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
404 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
405 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
407 (aarch64_st2<vstruct_elt>_dreg): This.
408 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
410 (aarch64_st3<vstruct_elt>_dreg): This.
411 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
413 (aarch64_st4<vstruct_elt>_dreg): This.
414 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
415 iterator and rename to...
416 (aarch64_st<nregs><vstruct_elt>): This.
417 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
418 iterator and rename to aarch64_st<nregs><vstruct_elt>.
419 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
420 tuple mode iterator and rename to...
421 (aarch64_st<nregs>_lane<vstruct_elt>): This.
422 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
423 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
425 (aarch64_simd_ld1<vstruct_elt>_x2): This.
426 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
427 Refactor to include new vector-tuple modes.
428 (aarch64_classify_vector_mode): Add cases for new vector-
430 (aarch64_advsimd_partial_struct_mode_p): Define.
431 (aarch64_advsimd_full_struct_mode_p): Likewise.
432 (aarch64_advsimd_vector_array_mode): Likewise.
433 (aarch64_sve_data_mode): Change location in file.
434 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
435 (aarch64_hard_regno_nregs): Handle case of partial Neon
437 (aarch64_classify_address): Refactor to include handling of
438 Neon vector-tuple modes.
439 (aarch64_print_operand): Print "d" for "%R" for a partial
440 Neon vector structure.
441 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
442 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
443 modes with scalar machines modes larger than 8 bytes.
444 (aarch64_can_change_mode_class): Don't allow changes between
445 partial and full Neon vector-structure modes.
446 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
447 builtin and remove boiler-plate code for opaque mode.
448 (vst2_lane_f32): Likewise.
449 (vst2_lane_f64): Likewise.
450 (vst2_lane_p8): Likewise.
451 (vst2_lane_p16): Likewise.
452 (vst2_lane_p64): Likewise.
453 (vst2_lane_s8): Likewise.
454 (vst2_lane_s16): Likewise.
455 (vst2_lane_s32): Likewise.
456 (vst2_lane_s64): Likewise.
457 (vst2_lane_u8): Likewise.
458 (vst2_lane_u16): Likewise.
459 (vst2_lane_u32): Likewise.
460 (vst2_lane_u64): Likewise.
461 (vst2q_lane_f16): Likewise.
462 (vst2q_lane_f32): Likewise.
463 (vst2q_lane_f64): Likewise.
464 (vst2q_lane_p8): Likewise.
465 (vst2q_lane_p16): Likewise.
466 (vst2q_lane_p64): Likewise.
467 (vst2q_lane_s8): Likewise.
468 (vst2q_lane_s16): Likewise.
469 (vst2q_lane_s32): Likewise.
470 (vst2q_lane_s64): Likewise.
471 (vst2q_lane_u8): Likewise.
472 (vst2q_lane_u16): Likewise.
473 (vst2q_lane_u32): Likewise.
474 (vst2q_lane_u64): Likewise.
475 (vst3_lane_f16): Likewise.
476 (vst3_lane_f32): Likewise.
477 (vst3_lane_f64): Likewise.
478 (vst3_lane_p8): Likewise.
479 (vst3_lane_p16): Likewise.
480 (vst3_lane_p64): Likewise.
481 (vst3_lane_s8): Likewise.
482 (vst3_lane_s16): Likewise.
483 (vst3_lane_s32): Likewise.
484 (vst3_lane_s64): Likewise.
485 (vst3_lane_u8): Likewise.
486 (vst3_lane_u16): Likewise.
487 (vst3_lane_u32): Likewise.
488 (vst3_lane_u64): Likewise.
489 (vst3q_lane_f16): Likewise.
490 (vst3q_lane_f32): Likewise.
491 (vst3q_lane_f64): Likewise.
492 (vst3q_lane_p8): Likewise.
493 (vst3q_lane_p16): Likewise.
494 (vst3q_lane_p64): Likewise.
495 (vst3q_lane_s8): Likewise.
496 (vst3q_lane_s16): Likewise.
497 (vst3q_lane_s32): Likewise.
498 (vst3q_lane_s64): Likewise.
499 (vst3q_lane_u8): Likewise.
500 (vst3q_lane_u16): Likewise.
501 (vst3q_lane_u32): Likewise.
502 (vst3q_lane_u64): Likewise.
503 (vst4_lane_f16): Likewise.
504 (vst4_lane_f32): Likewise.
505 (vst4_lane_f64): Likewise.
506 (vst4_lane_p8): Likewise.
507 (vst4_lane_p16): Likewise.
508 (vst4_lane_p64): Likewise.
509 (vst4_lane_s8): Likewise.
510 (vst4_lane_s16): Likewise.
511 (vst4_lane_s32): Likewise.
512 (vst4_lane_s64): Likewise.
513 (vst4_lane_u8): Likewise.
514 (vst4_lane_u16): Likewise.
515 (vst4_lane_u32): Likewise.
516 (vst4_lane_u64): Likewise.
517 (vst4q_lane_f16): Likewise.
518 (vst4q_lane_f32): Likewise.
519 (vst4q_lane_f64): Likewise.
520 (vst4q_lane_p8): Likewise.
521 (vst4q_lane_p16): Likewise.
522 (vst4q_lane_p64): Likewise.
523 (vst4q_lane_s8): Likewise.
524 (vst4q_lane_s16): Likewise.
525 (vst4q_lane_s32): Likewise.
526 (vst4q_lane_s64): Likewise.
527 (vst4q_lane_u8): Likewise.
528 (vst4q_lane_u16): Likewise.
529 (vst4q_lane_u32): Likewise.
530 (vst4q_lane_u64): Likewise.
531 (vtbl3_s8): Likewise.
532 (vtbl3_u8): Likewise.
533 (vtbl3_p8): Likewise.
534 (vtbl4_s8): Likewise.
535 (vtbl4_u8): Likewise.
536 (vtbl4_p8): Likewise.
537 (vld1_u8_x3): Likewise.
538 (vld1_s8_x3): Likewise.
539 (vld1_u16_x3): Likewise.
540 (vld1_s16_x3): Likewise.
541 (vld1_u32_x3): Likewise.
542 (vld1_s32_x3): Likewise.
543 (vld1_u64_x3): Likewise.
544 (vld1_s64_x3): Likewise.
545 (vld1_f16_x3): Likewise.
546 (vld1_f32_x3): Likewise.
547 (vld1_f64_x3): Likewise.
548 (vld1_p8_x3): Likewise.
549 (vld1_p16_x3): Likewise.
550 (vld1_p64_x3): Likewise.
551 (vld1q_u8_x3): Likewise.
552 (vld1q_s8_x3): Likewise.
553 (vld1q_u16_x3): Likewise.
554 (vld1q_s16_x3): Likewise.
555 (vld1q_u32_x3): Likewise.
556 (vld1q_s32_x3): Likewise.
557 (vld1q_u64_x3): Likewise.
558 (vld1q_s64_x3): Likewise.
559 (vld1q_f16_x3): Likewise.
560 (vld1q_f32_x3): Likewise.
561 (vld1q_f64_x3): Likewise.
562 (vld1q_p8_x3): Likewise.
563 (vld1q_p16_x3): Likewise.
564 (vld1q_p64_x3): Likewise.
565 (vld1_u8_x2): Likewise.
566 (vld1_s8_x2): Likewise.
567 (vld1_u16_x2): Likewise.
568 (vld1_s16_x2): Likewise.
569 (vld1_u32_x2): Likewise.
570 (vld1_s32_x2): Likewise.
571 (vld1_u64_x2): Likewise.
572 (vld1_s64_x2): Likewise.
573 (vld1_f16_x2): Likewise.
574 (vld1_f32_x2): Likewise.
575 (vld1_f64_x2): Likewise.
576 (vld1_p8_x2): Likewise.
577 (vld1_p16_x2): Likewise.
578 (vld1_p64_x2): Likewise.
579 (vld1q_u8_x2): Likewise.
580 (vld1q_s8_x2): Likewise.
581 (vld1q_u16_x2): Likewise.
582 (vld1q_s16_x2): Likewise.
583 (vld1q_u32_x2): Likewise.
584 (vld1q_s32_x2): Likewise.
585 (vld1q_u64_x2): Likewise.
586 (vld1q_s64_x2): Likewise.
587 (vld1q_f16_x2): Likewise.
588 (vld1q_f32_x2): Likewise.
589 (vld1q_f64_x2): Likewise.
590 (vld1q_p8_x2): Likewise.
591 (vld1q_p16_x2): Likewise.
592 (vld1q_p64_x2): Likewise.
593 (vld1_s8_x4): Likewise.
594 (vld1q_s8_x4): Likewise.
595 (vld1_s16_x4): Likewise.
596 (vld1q_s16_x4): Likewise.
597 (vld1_s32_x4): Likewise.
598 (vld1q_s32_x4): Likewise.
599 (vld1_u8_x4): Likewise.
600 (vld1q_u8_x4): Likewise.
601 (vld1_u16_x4): Likewise.
602 (vld1q_u16_x4): Likewise.
603 (vld1_u32_x4): Likewise.
604 (vld1q_u32_x4): Likewise.
605 (vld1_f16_x4): Likewise.
606 (vld1q_f16_x4): Likewise.
607 (vld1_f32_x4): Likewise.
608 (vld1q_f32_x4): Likewise.
609 (vld1_p8_x4): Likewise.
610 (vld1q_p8_x4): Likewise.
611 (vld1_p16_x4): Likewise.
612 (vld1q_p16_x4): Likewise.
613 (vld1_s64_x4): Likewise.
614 (vld1_u64_x4): Likewise.
615 (vld1_p64_x4): Likewise.
616 (vld1q_s64_x4): Likewise.
617 (vld1q_u64_x4): Likewise.
618 (vld1q_p64_x4): Likewise.
619 (vld1_f64_x4): Likewise.
620 (vld1q_f64_x4): Likewise.
621 (vld2_s64): Likewise.
622 (vld2_u64): Likewise.
623 (vld2_f64): Likewise.
626 (vld2_p64): Likewise.
627 (vld2_s16): Likewise.
628 (vld2_p16): Likewise.
629 (vld2_s32): Likewise.
631 (vld2_u16): Likewise.
632 (vld2_u32): Likewise.
633 (vld2_f16): Likewise.
634 (vld2_f32): Likewise.
635 (vld2q_s8): Likewise.
636 (vld2q_p8): Likewise.
637 (vld2q_s16): Likewise.
638 (vld2q_p16): Likewise.
639 (vld2q_p64): Likewise.
640 (vld2q_s32): Likewise.
641 (vld2q_s64): Likewise.
642 (vld2q_u8): Likewise.
643 (vld2q_u16): Likewise.
644 (vld2q_u32): Likewise.
645 (vld2q_u64): Likewise.
646 (vld2q_f16): Likewise.
647 (vld2q_f32): Likewise.
648 (vld2q_f64): Likewise.
649 (vld3_s64): Likewise.
650 (vld3_u64): Likewise.
651 (vld3_f64): Likewise.
654 (vld3_s16): Likewise.
655 (vld3_p16): Likewise.
656 (vld3_s32): Likewise.
658 (vld3_u16): Likewise.
659 (vld3_u32): Likewise.
660 (vld3_f16): Likewise.
661 (vld3_f32): Likewise.
662 (vld3_p64): Likewise.
663 (vld3q_s8): Likewise.
664 (vld3q_p8): Likewise.
665 (vld3q_s16): Likewise.
666 (vld3q_p16): Likewise.
667 (vld3q_s32): Likewise.
668 (vld3q_s64): Likewise.
669 (vld3q_u8): Likewise.
670 (vld3q_u16): Likewise.
671 (vld3q_u32): Likewise.
672 (vld3q_u64): Likewise.
673 (vld3q_f16): Likewise.
674 (vld3q_f32): Likewise.
675 (vld3q_f64): Likewise.
676 (vld3q_p64): Likewise.
677 (vld4_s64): Likewise.
678 (vld4_u64): Likewise.
679 (vld4_f64): Likewise.
682 (vld4_s16): Likewise.
683 (vld4_p16): Likewise.
684 (vld4_s32): Likewise.
686 (vld4_u16): Likewise.
687 (vld4_u32): Likewise.
688 (vld4_f16): Likewise.
689 (vld4_f32): Likewise.
690 (vld4_p64): Likewise.
691 (vld4q_s8): Likewise.
692 (vld4q_p8): Likewise.
693 (vld4q_s16): Likewise.
694 (vld4q_p16): Likewise.
695 (vld4q_s32): Likewise.
696 (vld4q_s64): Likewise.
697 (vld4q_u8): Likewise.
698 (vld4q_u16): Likewise.
699 (vld4q_u32): Likewise.
700 (vld4q_u64): Likewise.
701 (vld4q_f16): Likewise.
702 (vld4q_f32): Likewise.
703 (vld4q_f64): Likewise.
704 (vld4q_p64): Likewise.
705 (vld2_dup_s8): Likewise.
706 (vld2_dup_s16): Likewise.
707 (vld2_dup_s32): Likewise.
708 (vld2_dup_f16): Likewise.
709 (vld2_dup_f32): Likewise.
710 (vld2_dup_f64): Likewise.
711 (vld2_dup_u8): Likewise.
712 (vld2_dup_u16): Likewise.
713 (vld2_dup_u32): Likewise.
714 (vld2_dup_p8): Likewise.
715 (vld2_dup_p16): Likewise.
716 (vld2_dup_p64): Likewise.
717 (vld2_dup_s64): Likewise.
718 (vld2_dup_u64): Likewise.
719 (vld2q_dup_s8): Likewise.
720 (vld2q_dup_p8): Likewise.
721 (vld2q_dup_s16): Likewise.
722 (vld2q_dup_p16): Likewise.
723 (vld2q_dup_s32): Likewise.
724 (vld2q_dup_s64): Likewise.
725 (vld2q_dup_u8): Likewise.
726 (vld2q_dup_u16): Likewise.
727 (vld2q_dup_u32): Likewise.
728 (vld2q_dup_u64): Likewise.
729 (vld2q_dup_f16): Likewise.
730 (vld2q_dup_f32): Likewise.
731 (vld2q_dup_f64): Likewise.
732 (vld2q_dup_p64): Likewise.
733 (vld3_dup_s64): Likewise.
734 (vld3_dup_u64): Likewise.
735 (vld3_dup_f64): Likewise.
736 (vld3_dup_s8): Likewise.
737 (vld3_dup_p8): Likewise.
738 (vld3_dup_s16): Likewise.
739 (vld3_dup_p16): Likewise.
740 (vld3_dup_s32): Likewise.
741 (vld3_dup_u8): Likewise.
742 (vld3_dup_u16): Likewise.
743 (vld3_dup_u32): Likewise.
744 (vld3_dup_f16): Likewise.
745 (vld3_dup_f32): Likewise.
746 (vld3_dup_p64): Likewise.
747 (vld3q_dup_s8): Likewise.
748 (vld3q_dup_p8): Likewise.
749 (vld3q_dup_s16): Likewise.
750 (vld3q_dup_p16): Likewise.
751 (vld3q_dup_s32): Likewise.
752 (vld3q_dup_s64): Likewise.
753 (vld3q_dup_u8): Likewise.
754 (vld3q_dup_u16): Likewise.
755 (vld3q_dup_u32): Likewise.
756 (vld3q_dup_u64): Likewise.
757 (vld3q_dup_f16): Likewise.
758 (vld3q_dup_f32): Likewise.
759 (vld3q_dup_f64): Likewise.
760 (vld3q_dup_p64): Likewise.
761 (vld4_dup_s64): Likewise.
762 (vld4_dup_u64): Likewise.
763 (vld4_dup_f64): Likewise.
764 (vld4_dup_s8): Likewise.
765 (vld4_dup_p8): Likewise.
766 (vld4_dup_s16): Likewise.
767 (vld4_dup_p16): Likewise.
768 (vld4_dup_s32): Likewise.
769 (vld4_dup_u8): Likewise.
770 (vld4_dup_u16): Likewise.
771 (vld4_dup_u32): Likewise.
772 (vld4_dup_f16): Likewise.
773 (vld4_dup_f32): Likewise.
774 (vld4_dup_p64): Likewise.
775 (vld4q_dup_s8): Likewise.
776 (vld4q_dup_p8): Likewise.
777 (vld4q_dup_s16): Likewise.
778 (vld4q_dup_p16): Likewise.
779 (vld4q_dup_s32): Likewise.
780 (vld4q_dup_s64): Likewise.
781 (vld4q_dup_u8): Likewise.
782 (vld4q_dup_u16): Likewise.
783 (vld4q_dup_u32): Likewise.
784 (vld4q_dup_u64): Likewise.
785 (vld4q_dup_f16): Likewise.
786 (vld4q_dup_f32): Likewise.
787 (vld4q_dup_f64): Likewise.
788 (vld4q_dup_p64): Likewise.
789 (vld2_lane_u8): Likewise.
790 (vld2_lane_u16): Likewise.
791 (vld2_lane_u32): Likewise.
792 (vld2_lane_u64): Likewise.
793 (vld2_lane_s8): Likewise.
794 (vld2_lane_s16): Likewise.
795 (vld2_lane_s32): Likewise.
796 (vld2_lane_s64): Likewise.
797 (vld2_lane_f16): Likewise.
798 (vld2_lane_f32): Likewise.
799 (vld2_lane_f64): Likewise.
800 (vld2_lane_p8): Likewise.
801 (vld2_lane_p16): Likewise.
802 (vld2_lane_p64): Likewise.
803 (vld2q_lane_u8): Likewise.
804 (vld2q_lane_u16): Likewise.
805 (vld2q_lane_u32): Likewise.
806 (vld2q_lane_u64): Likewise.
807 (vld2q_lane_s8): Likewise.
808 (vld2q_lane_s16): Likewise.
809 (vld2q_lane_s32): Likewise.
810 (vld2q_lane_s64): Likewise.
811 (vld2q_lane_f16): Likewise.
812 (vld2q_lane_f32): Likewise.
813 (vld2q_lane_f64): Likewise.
814 (vld2q_lane_p8): Likewise.
815 (vld2q_lane_p16): Likewise.
816 (vld2q_lane_p64): Likewise.
817 (vld3_lane_u8): Likewise.
818 (vld3_lane_u16): Likewise.
819 (vld3_lane_u32): Likewise.
820 (vld3_lane_u64): Likewise.
821 (vld3_lane_s8): Likewise.
822 (vld3_lane_s16): Likewise.
823 (vld3_lane_s32): Likewise.
824 (vld3_lane_s64): Likewise.
825 (vld3_lane_f16): Likewise.
826 (vld3_lane_f32): Likewise.
827 (vld3_lane_f64): Likewise.
828 (vld3_lane_p8): Likewise.
829 (vld3_lane_p16): Likewise.
830 (vld3_lane_p64): Likewise.
831 (vld3q_lane_u8): Likewise.
832 (vld3q_lane_u16): Likewise.
833 (vld3q_lane_u32): Likewise.
834 (vld3q_lane_u64): Likewise.
835 (vld3q_lane_s8): Likewise.
836 (vld3q_lane_s16): Likewise.
837 (vld3q_lane_s32): Likewise.
838 (vld3q_lane_s64): Likewise.
839 (vld3q_lane_f16): Likewise.
840 (vld3q_lane_f32): Likewise.
841 (vld3q_lane_f64): Likewise.
842 (vld3q_lane_p8): Likewise.
843 (vld3q_lane_p16): Likewise.
844 (vld3q_lane_p64): Likewise.
845 (vld4_lane_u8): Likewise.
846 (vld4_lane_u16): Likewise.
847 (vld4_lane_u32): Likewise.
848 (vld4_lane_u64): Likewise.
849 (vld4_lane_s8): Likewise.
850 (vld4_lane_s16): Likewise.
851 (vld4_lane_s32): Likewise.
852 (vld4_lane_s64): Likewise.
853 (vld4_lane_f16): Likewise.
854 (vld4_lane_f32): Likewise.
855 (vld4_lane_f64): Likewise.
856 (vld4_lane_p8): Likewise.
857 (vld4_lane_p16): Likewise.
858 (vld4_lane_p64): Likewise.
859 (vld4q_lane_u8): Likewise.
860 (vld4q_lane_u16): Likewise.
861 (vld4q_lane_u32): Likewise.
862 (vld4q_lane_u64): Likewise.
863 (vld4q_lane_s8): Likewise.
864 (vld4q_lane_s16): Likewise.
865 (vld4q_lane_s32): Likewise.
866 (vld4q_lane_s64): Likewise.
867 (vld4q_lane_f16): Likewise.
868 (vld4q_lane_f32): Likewise.
869 (vld4q_lane_f64): Likewise.
870 (vld4q_lane_p8): Likewise.
871 (vld4q_lane_p16): Likewise.
872 (vld4q_lane_p64): Likewise.
873 (vqtbl2_s8): Likewise.
874 (vqtbl2_u8): Likewise.
875 (vqtbl2_p8): Likewise.
876 (vqtbl2q_s8): Likewise.
877 (vqtbl2q_u8): Likewise.
878 (vqtbl2q_p8): Likewise.
879 (vqtbl3_s8): Likewise.
880 (vqtbl3_u8): Likewise.
881 (vqtbl3_p8): Likewise.
882 (vqtbl3q_s8): Likewise.
883 (vqtbl3q_u8): Likewise.
884 (vqtbl3q_p8): Likewise.
885 (vqtbl4_s8): Likewise.
886 (vqtbl4_u8): Likewise.
887 (vqtbl4_p8): Likewise.
888 (vqtbl4q_s8): Likewise.
889 (vqtbl4q_u8): Likewise.
890 (vqtbl4q_p8): Likewise.
891 (vqtbx2_s8): Likewise.
892 (vqtbx2_u8): Likewise.
893 (vqtbx2_p8): Likewise.
894 (vqtbx2q_s8): Likewise.
895 (vqtbx2q_u8): Likewise.
896 (vqtbx2q_p8): Likewise.
897 (vqtbx3_s8): Likewise.
898 (vqtbx3_u8): Likewise.
899 (vqtbx3_p8): Likewise.
900 (vqtbx3q_s8): Likewise.
901 (vqtbx3q_u8): Likewise.
902 (vqtbx3q_p8): Likewise.
903 (vqtbx4_s8): Likewise.
904 (vqtbx4_u8): Likewise.
905 (vqtbx4_p8): Likewise.
906 (vqtbx4q_s8): Likewise.
907 (vqtbx4q_u8): Likewise.
908 (vqtbx4q_p8): Likewise.
909 (vst1_s64_x2): Likewise.
910 (vst1_u64_x2): Likewise.
911 (vst1_f64_x2): Likewise.
912 (vst1_s8_x2): Likewise.
913 (vst1_p8_x2): Likewise.
914 (vst1_s16_x2): Likewise.
915 (vst1_p16_x2): Likewise.
916 (vst1_s32_x2): Likewise.
917 (vst1_u8_x2): Likewise.
918 (vst1_u16_x2): Likewise.
919 (vst1_u32_x2): Likewise.
920 (vst1_f16_x2): Likewise.
921 (vst1_f32_x2): Likewise.
922 (vst1_p64_x2): Likewise.
923 (vst1q_s8_x2): Likewise.
924 (vst1q_p8_x2): Likewise.
925 (vst1q_s16_x2): Likewise.
926 (vst1q_p16_x2): Likewise.
927 (vst1q_s32_x2): Likewise.
928 (vst1q_s64_x2): Likewise.
929 (vst1q_u8_x2): Likewise.
930 (vst1q_u16_x2): Likewise.
931 (vst1q_u32_x2): Likewise.
932 (vst1q_u64_x2): Likewise.
933 (vst1q_f16_x2): Likewise.
934 (vst1q_f32_x2): Likewise.
935 (vst1q_f64_x2): Likewise.
936 (vst1q_p64_x2): Likewise.
937 (vst1_s64_x3): Likewise.
938 (vst1_u64_x3): Likewise.
939 (vst1_f64_x3): Likewise.
940 (vst1_s8_x3): Likewise.
941 (vst1_p8_x3): Likewise.
942 (vst1_s16_x3): Likewise.
943 (vst1_p16_x3): Likewise.
944 (vst1_s32_x3): Likewise.
945 (vst1_u8_x3): Likewise.
946 (vst1_u16_x3): Likewise.
947 (vst1_u32_x3): Likewise.
948 (vst1_f16_x3): Likewise.
949 (vst1_f32_x3): Likewise.
950 (vst1_p64_x3): Likewise.
951 (vst1q_s8_x3): Likewise.
952 (vst1q_p8_x3): Likewise.
953 (vst1q_s16_x3): Likewise.
954 (vst1q_p16_x3): Likewise.
955 (vst1q_s32_x3): Likewise.
956 (vst1q_s64_x3): Likewise.
957 (vst1q_u8_x3): Likewise.
958 (vst1q_u16_x3): Likewise.
959 (vst1q_u32_x3): Likewise.
960 (vst1q_u64_x3): Likewise.
961 (vst1q_f16_x3): Likewise.
962 (vst1q_f32_x3): Likewise.
963 (vst1q_f64_x3): Likewise.
964 (vst1q_p64_x3): Likewise.
965 (vst1_s8_x4): Likewise.
966 (vst1q_s8_x4): Likewise.
967 (vst1_s16_x4): Likewise.
968 (vst1q_s16_x4): Likewise.
969 (vst1_s32_x4): Likewise.
970 (vst1q_s32_x4): Likewise.
971 (vst1_u8_x4): Likewise.
972 (vst1q_u8_x4): Likewise.
973 (vst1_u16_x4): Likewise.
974 (vst1q_u16_x4): Likewise.
975 (vst1_u32_x4): Likewise.
976 (vst1q_u32_x4): Likewise.
977 (vst1_f16_x4): Likewise.
978 (vst1q_f16_x4): Likewise.
979 (vst1_f32_x4): Likewise.
980 (vst1q_f32_x4): Likewise.
981 (vst1_p8_x4): Likewise.
982 (vst1q_p8_x4): Likewise.
983 (vst1_p16_x4): Likewise.
984 (vst1q_p16_x4): Likewise.
985 (vst1_s64_x4): Likewise.
986 (vst1_u64_x4): Likewise.
987 (vst1_p64_x4): Likewise.
988 (vst1q_s64_x4): Likewise.
989 (vst1q_u64_x4): Likewise.
990 (vst1q_p64_x4): Likewise.
991 (vst1_f64_x4): Likewise.
992 (vst1q_f64_x4): Likewise.
993 (vst2_s64): Likewise.
994 (vst2_u64): Likewise.
995 (vst2_f64): Likewise.
998 (vst2_s16): Likewise.
999 (vst2_p16): Likewise.
1000 (vst2_s32): Likewise.
1001 (vst2_u8): Likewise.
1002 (vst2_u16): Likewise.
1003 (vst2_u32): Likewise.
1004 (vst2_f16): Likewise.
1005 (vst2_f32): Likewise.
1006 (vst2_p64): Likewise.
1007 (vst2q_s8): Likewise.
1008 (vst2q_p8): Likewise.
1009 (vst2q_s16): Likewise.
1010 (vst2q_p16): Likewise.
1011 (vst2q_s32): Likewise.
1012 (vst2q_s64): Likewise.
1013 (vst2q_u8): Likewise.
1014 (vst2q_u16): Likewise.
1015 (vst2q_u32): Likewise.
1016 (vst2q_u64): Likewise.
1017 (vst2q_f16): Likewise.
1018 (vst2q_f32): Likewise.
1019 (vst2q_f64): Likewise.
1020 (vst2q_p64): Likewise.
1021 (vst3_s64): Likewise.
1022 (vst3_u64): Likewise.
1023 (vst3_f64): Likewise.
1024 (vst3_s8): Likewise.
1025 (vst3_p8): Likewise.
1026 (vst3_s16): Likewise.
1027 (vst3_p16): Likewise.
1028 (vst3_s32): Likewise.
1029 (vst3_u8): Likewise.
1030 (vst3_u16): Likewise.
1031 (vst3_u32): Likewise.
1032 (vst3_f16): Likewise.
1033 (vst3_f32): Likewise.
1034 (vst3_p64): Likewise.
1035 (vst3q_s8): Likewise.
1036 (vst3q_p8): Likewise.
1037 (vst3q_s16): Likewise.
1038 (vst3q_p16): Likewise.
1039 (vst3q_s32): Likewise.
1040 (vst3q_s64): Likewise.
1041 (vst3q_u8): Likewise.
1042 (vst3q_u16): Likewise.
1043 (vst3q_u32): Likewise.
1044 (vst3q_u64): Likewise.
1045 (vst3q_f16): Likewise.
1046 (vst3q_f32): Likewise.
1047 (vst3q_f64): Likewise.
1048 (vst3q_p64): Likewise.
1049 (vst4_s64): Likewise.
1050 (vst4_u64): Likewise.
1051 (vst4_f64): Likewise.
1052 (vst4_s8): Likewise.
1053 (vst4_p8): Likewise.
1054 (vst4_s16): Likewise.
1055 (vst4_p16): Likewise.
1056 (vst4_s32): Likewise.
1057 (vst4_u8): Likewise.
1058 (vst4_u16): Likewise.
1059 (vst4_u32): Likewise.
1060 (vst4_f16): Likewise.
1061 (vst4_f32): Likewise.
1062 (vst4_p64): Likewise.
1063 (vst4q_s8): Likewise.
1064 (vst4q_p8): Likewise.
1065 (vst4q_s16): Likewise.
1066 (vst4q_p16): Likewise.
1067 (vst4q_s32): Likewise.
1068 (vst4q_s64): Likewise.
1069 (vst4q_u8): Likewise.
1070 (vst4q_u16): Likewise.
1071 (vst4q_u32): Likewise.
1072 (vst4q_u64): Likewise.
1073 (vst4q_f16): Likewise.
1074 (vst4q_f32): Likewise.
1075 (vst4q_f64): Likewise.
1076 (vst4q_p64): Likewise.
1077 (vtbx4_s8): Likewise.
1078 (vtbx4_u8): Likewise.
1079 (vtbx4_p8): Likewise.
1080 (vld1_bf16_x2): Likewise.
1081 (vld1q_bf16_x2): Likewise.
1082 (vld1_bf16_x3): Likewise.
1083 (vld1q_bf16_x3): Likewise.
1084 (vld1_bf16_x4): Likewise.
1085 (vld1q_bf16_x4): Likewise.
1086 (vld2_bf16): Likewise.
1087 (vld2q_bf16): Likewise.
1088 (vld2_dup_bf16): Likewise.
1089 (vld2q_dup_bf16): Likewise.
1090 (vld3_bf16): Likewise.
1091 (vld3q_bf16): Likewise.
1092 (vld3_dup_bf16): Likewise.
1093 (vld3q_dup_bf16): Likewise.
1094 (vld4_bf16): Likewise.
1095 (vld4q_bf16): Likewise.
1096 (vld4_dup_bf16): Likewise.
1097 (vld4q_dup_bf16): Likewise.
1098 (vst1_bf16_x2): Likewise.
1099 (vst1q_bf16_x2): Likewise.
1100 (vst1_bf16_x3): Likewise.
1101 (vst1q_bf16_x3): Likewise.
1102 (vst1_bf16_x4): Likewise.
1103 (vst1q_bf16_x4): Likewise.
1104 (vst2_bf16): Likewise.
1105 (vst2q_bf16): Likewise.
1106 (vst3_bf16): Likewise.
1107 (vst3q_bf16): Likewise.
1108 (vst4_bf16): Likewise.
1109 (vst4q_bf16): Likewise.
1110 (vld2_lane_bf16): Likewise.
1111 (vld2q_lane_bf16): Likewise.
1112 (vld3_lane_bf16): Likewise.
1113 (vld3q_lane_bf16): Likewise.
1114 (vld4_lane_bf16): Likewise.
1115 (vld4q_lane_bf16): Likewise.
1116 (vst2_lane_bf16): Likewise.
1117 (vst2q_lane_bf16): Likewise.
1118 (vst3_lane_bf16): Likewise.
1119 (vst3q_lane_bf16): Likewise.
1120 (vst4_lane_bf16): Likewise.
1121 (vst4q_lane_bf16): Likewise.
1122 * config/aarch64/geniterators.sh: Modify iterator regex to
1123 match new vector-tuple modes.
1124 * config/aarch64/iterators.md (insn_count): Extend mode
1125 attribute with vector-tuple type information.
1127 (Vendreg): Likewise.
1130 (VSTRUCT_2D): New mode iterator.
1131 (VSTRUCT_2DNX): Likewise.
1132 (VSTRUCT_2DX): Likewise.
1133 (VSTRUCT_2Q): Likewise.
1134 (VSTRUCT_2QD): Likewise.
1135 (VSTRUCT_3D): Likewise.
1136 (VSTRUCT_3DNX): Likewise.
1137 (VSTRUCT_3DX): Likewise.
1138 (VSTRUCT_3Q): Likewise.
1139 (VSTRUCT_3QD): Likewise.
1140 (VSTRUCT_4D): Likewise.
1141 (VSTRUCT_4DNX): Likewise.
1142 (VSTRUCT_4DX): Likewise.
1143 (VSTRUCT_4Q): Likewise.
1144 (VSTRUCT_4QD): Likewise.
1145 (VSTRUCT_D): Likewise.
1146 (VSTRUCT_Q): Likewise.
1147 (VSTRUCT_QD): Likewise.
1148 (VSTRUCT_ELT): New mode attribute.
1149 (vstruct_elt): Likewise.
1150 * genmodes.c (VECTOR_MODE): Add default prefix and order
1152 (VECTOR_MODE_WITH_PREFIX): Define.
1153 (make_vector_mode): Add mode prefix and order parameters.
1155 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1157 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
1159 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1161 * expr.c (emit_group_load_1): Remove historic workaround.
1163 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1165 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
1166 Factor out main loop to...
1167 (aarch64_init_simd_builtin_functions): This new function.
1168 (register_tuple_type): Define.
1169 (aarch64_scalar_builtin_type_p): Define.
1170 (handle_arm_neon_h): Define.
1171 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
1172 pragma for arm_neon.h.
1173 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
1175 (handle_arm_neon_h): Likewise.
1176 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
1177 Remove static modifier.
1178 * config/aarch64/arm_neon.h (target): Remove Neon vector
1179 structure type definitions.
1181 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1183 PR tree-optimization/102943
1184 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1185 Prefer range_of_expr unless there are no statements in the BB.
1187 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1189 PR tree-optimization/102943
1190 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
1191 Avoid duplicate calculation of paths.
1193 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1195 PR tree-optimization/102943
1196 * gimple-range-path.cc (path_range_query::compute_phi_relations):
1197 Only compute relations for SSA names in the import list.
1198 (path_range_query::compute_outgoing_relations): Same.
1199 * gimple-range-path.h (path_range_query::import_p): New.
1201 2021-11-04 Richard Biener <rguenther@suse.de>
1203 PR rtl-optimization/103075
1204 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
1205 false for a VOIDmode operand.
1207 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1209 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
1210 variables private and add "m_" to their names. Remove is_loop.
1211 (aarch64_record_potential_advsimd_unrolling): Replace with...
1212 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
1213 (aarch64_analyze_loop_vinfo): Replace with...
1214 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
1215 Move initialization of (m_)vec_flags to add_stmt_cost.
1216 (aarch64_analyze_bb_vinfo): Delete.
1217 (aarch64_count_ops): Replace with...
1218 (aarch64_vector_costs::count_ops): ...this.
1219 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
1220 using m_costing_for_scalar to test whether we're costing
1221 scalar or vector code.
1222 (aarch64_adjust_body_cost_sve): Replace with...
1223 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
1224 (aarch64_adjust_body_cost): Replace with...
1225 (aarch64_vector_costs::adjust_body_cost): ...this.
1226 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
1228 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1230 * target.def (targetm.vectorize.init_cost): Replace with...
1231 (targetm.vectorize.create_costs): ...this.
1232 (targetm.vectorize.add_stmt_cost): Delete.
1233 (targetm.vectorize.finish_cost): Likewise.
1234 (targetm.vectorize.destroy_cost_data): Likewise.
1235 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
1236 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1237 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1238 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1239 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1240 * doc/tm.texi: Regenerate.
1241 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
1243 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
1244 (vector_costs): New class.
1245 (init_cost): Take a vec_info and return a vector_costs.
1246 (dump_stmt_cost): Remove data parameter.
1247 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
1248 (add_stmt_costs): Likewise.
1249 (finish_cost): Replace data parameter with a vector_costs.
1250 (destroy_cost_data): Delete.
1251 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
1253 (vec_info::vec_info): Remove the target_cost_data parameter and
1254 initialize the member variable to null instead.
1255 (vec_info::~vec_info): Delete target_cost_data instead of calling
1257 (vector_costs::add_stmt_cost): New function.
1258 (vector_costs::finish_cost): Likewise.
1259 (vector_costs::record_stmt_cost): Likewise.
1260 (vector_costs::adjust_cost_for_freq): Likewise.
1261 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
1262 call to vec_info::vec_info.
1263 (vect_compute_single_scalar_iteration_cost): Update after above
1264 changes to costing interface.
1265 (vect_analyze_loop_operations): Likewise.
1266 (vect_estimate_min_profitable_iters): Likewise.
1267 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
1268 at the start_over point, where it needs to be recreated after
1269 trying without slp. Update retry code accordingly.
1270 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
1271 to vec_info::vec_info.
1272 (vect_slp_analyze_operation): Update after above changes to costing
1274 (vect_bb_vectorization_profitable_p): Likewise.
1275 * targhooks.h (default_init_cost): Replace with...
1276 (default_vectorize_create_costs): ...this.
1277 (default_add_stmt_cost): Delete.
1278 (default_finish_cost, default_destroy_cost_data): Likewise.
1279 * targhooks.c (default_init_cost): Replace with...
1280 (default_vectorize_create_costs): ...this.
1281 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
1282 (default_finish_cost, default_destroy_cost_data): Delete.
1283 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
1284 vector_costs. Add a constructor.
1285 (aarch64_init_cost): Replace with...
1286 (aarch64_vectorize_create_costs): ...this.
1287 (aarch64_add_stmt_cost): Replace with...
1288 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
1289 to adjust the cost for inner loops.
1290 (aarch64_finish_cost): Replace with...
1291 (aarch64_vector_costs::finish_cost): ...this.
1292 (aarch64_destroy_cost_data): Delete.
1293 (TARGET_VECTORIZE_INIT_COST): Replace with...
1294 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1295 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1296 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1297 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1298 * config/i386/i386.c (ix86_vector_costs): New structure.
1299 (ix86_init_cost): Replace with...
1300 (ix86_vectorize_create_costs): ...this.
1301 (ix86_add_stmt_cost): Replace with...
1302 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
1303 to adjust the cost for inner loops.
1304 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
1305 (TARGET_VECTORIZE_INIT_COST): Replace with...
1306 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1307 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1308 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1309 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1310 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
1311 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1312 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1313 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1314 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1315 (rs6000_cost_data): Inherit from vector_costs.
1316 Add a constructor. Drop loop_info, cost and costing_for_scalar
1317 in favor of the corresponding vector_costs member variables.
1318 Add "m_" to the names of the remaining member variables and
1320 (rs6000_density_test): Replace with...
1321 (rs6000_cost_data::density_test): ...this.
1322 (rs6000_init_cost): Replace with...
1323 (rs6000_vectorize_create_costs): ...this.
1324 (rs6000_update_target_cost_per_stmt): Replace with...
1325 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
1326 (rs6000_add_stmt_cost): Replace with...
1327 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
1328 to adjust the cost for inner loops.
1329 (rs6000_adjust_vect_cost_per_loop): Replace with...
1330 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
1331 (rs6000_finish_cost): Replace with...
1332 (rs6000_cost_data::finish_cost): ...this. Group loop code
1333 into a single if statement and pass the loop_vinfo down to
1335 (rs6000_destroy_cost_data): Delete.
1337 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1339 PR tree-optimization/103062
1340 PR tree-optimization/103062
1341 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
1342 Increase size of allocation by 1.
1343 (ssa_equiv_stack::push_replacement): Grow as needed.
1344 (ssa_equiv_stack::get_replacement): Same.
1345 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
1346 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
1347 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
1348 (pointer_equiv_analyzer::get_equiv): Same.
1349 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
1350 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
1351 const markers. Use auto_vec instead of tree *.
1353 2021-11-04 Richard Biener <rguenther@suse.de>
1355 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
1356 true parameter and inline valueization.
1357 (vn_nary_op_lookup_1): Inline valueization from ...
1358 (vn_nary_op_compute_hash): ... here and remove it here.
1359 * tree-ssa-pre.c (phi_translate_1): Do not valueize
1360 before vn_nary_lookup_pieces.
1361 (get_representative_for): Mark created SSA representatives
1364 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1366 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
1367 that the operand has a vector mode. Use subreg_lowpart_offset
1368 to test whether an index corresponds to the low part.
1370 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1372 * read-rtl.c: Remove dead !GENERATOR_FILE block.
1373 * read-rtl-function.c (function_reader::consolidate_singletons):
1374 Generate canonical CONST_VECTORs.
1376 2021-11-04 liuhongt <hongtao.liu@intel.com>
1379 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
1380 (regmem_or_bitnot_regmem_operand): .. and extend to handle
1382 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
1383 the operands which are required to be register_operand.
1384 (*<avx512>_vpternlog<mode>_2): Ditto.
1385 (*<avx512>_vpternlog<mode>_3): Ditto.
1386 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
1387 vector HFmodes since it's not a real AVX512FP16 instruction.
1389 2021-11-04 liuhongt <hongtao.liu@intel.com>
1392 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
1393 .COPYSIGN (a,b) when a and b are same type as the truncation
1394 type and has less precision than extend type.
1396 2021-11-04 Richard Biener <rguenther@suse.de>
1398 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
1401 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1403 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
1404 (VI1_AVX512VNNI): Likewise.
1405 (SDOT_VPDP_SUF): New mode_attr.
1408 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
1409 vpdpwssd when VNNI targets available.
1410 (usdot_prod<mode>): New expander for vector QImode.
1412 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1414 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
1415 parentheses to base and stride.
1416 (_tile_stream_loadd_internal): Likewise.
1417 (_tile_stored_internal): Likewise.
1419 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
1421 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
1422 arguments to `reg_class_subset_p'.
1424 2021-11-03 Joseph Myers <joseph@codesourcery.com>
1427 * fold-const.c (fold_init): New function.
1428 * fold-const.h (fold_init): New prototype.
1430 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1432 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
1434 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1436 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
1437 a range from edge if arg == phidef.
1439 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1441 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
1442 for constant before any other processing.
1444 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1446 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
1448 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1450 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
1451 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
1452 no explcit context and current_bb is set.
1453 (gimple_ranger::fold_stmt): New.
1454 * gimple-range.h (current_bb, fold_stmt): New.
1455 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
1457 2021-11-03 Richard Biener <rguenther@suse.de>
1459 PR tree-optimization/102970
1460 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
1461 when translating a MEM_REF over a backedge.
1463 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
1465 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
1466 * config/aarch64/aarch64-tune.md: Regenerate.
1467 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
1468 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
1469 * doc/invoke.texi: Add documentation for Ampere-1 core.
1471 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
1473 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
1475 (ldr_got_small_<mode>): Remove pattern.
1476 (ldr_got_small_sidi): Likewise.
1477 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
1478 GOT accesses as moves.
1479 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
1480 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
1481 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
1483 2021-11-03 Martin Liska <mliska@suse.cz>
1485 * gcov.c (read_line): Remove dead variable.
1487 2021-11-03 Martin Liska <mliska@suse.cz>
1490 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
1491 (ipa_fn_summary::account_size_time): Likewise.
1492 (edge_set_predicate): Likewise.
1493 (set_hint_predicate): Likewise.
1494 (add_freqcounting_predicate): Likewise.
1495 (evaluate_conditions_for_known_args): Likewise.
1496 (evaluate_properties_for_edge): Likewise.
1497 (remap_freqcounting_preds_after_dup): Likewise.
1498 (ipa_fn_summary_t::duplicate): Likewise.
1499 (set_cond_stmt_execution_predicate): Likewise.
1500 (set_switch_stmt_execution_predicate): Likewise.
1501 (compute_bb_predicates): Likewise.
1502 (will_be_nonconstant_expr_predicate): Likewise.
1503 (will_be_nonconstant_predicate): Likewise.
1504 (phi_result_unknown_predicate): Likewise.
1505 (predicate_for_phi_result): Likewise.
1506 (analyze_function_body): Likewise.
1507 (compute_fn_summary): Likewise.
1508 (summarize_calls_size_and_time): Likewise.
1509 (estimate_calls_size_and_time): Likewise.
1510 (ipa_call_context::estimate_size_and_time): Likewise.
1511 (remap_edge_summaries): Likewise.
1512 (remap_freqcounting_predicate): Likewise.
1513 (ipa_merge_fn_summary_after_inlining): Likewise.
1514 (ipa_update_overall_fn_summary): Likewise.
1515 (read_ipa_call_summary): Likewise.
1516 (inline_read_section): Likewise.
1517 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
1518 * ipa-predicate.c (predicate::add_clause): Likewise.
1519 (ipa_predicate::add_clause): Likewise.
1520 (predicate::or_with): Likewise.
1521 (ipa_predicate::or_with): Likewise.
1522 (predicate::evaluate): Likewise.
1523 (ipa_predicate::evaluate): Likewise.
1524 (predicate::probability): Likewise.
1525 (ipa_predicate::probability): Likewise.
1526 (dump_condition): Likewise.
1527 (dump_clause): Likewise.
1528 (predicate::dump): Likewise.
1529 (ipa_predicate::dump): Likewise.
1530 (predicate::debug): Likewise.
1531 (ipa_predicate::debug): Likewise.
1532 (predicate::remap_after_duplication): Likewise.
1533 (ipa_predicate::remap_after_duplication): Likewise.
1534 (predicate::remap_after_inlining): Likewise.
1535 (ipa_predicate::remap_after_inlining): Likewise.
1536 (predicate::stream_in): Likewise.
1537 (ipa_predicate::stream_in): Likewise.
1538 (predicate::stream_out): Likewise.
1539 (ipa_predicate::stream_out): Likewise.
1540 (add_condition): Likewise.
1541 * ipa-predicate.h (class predicate): Likewise.
1542 (class ipa_predicate): Likewise.
1543 (add_condition): Likewise.
1545 2021-11-03 Richard Biener <rguenther@suse.de>
1547 * bitmap.h (bitmap_bit_p): Change the return type to bool.
1548 * bitmap.c (bitmap_bit_p): Likewise.
1549 * sbitmap.h (bitmap_bit_p): Likewise.
1550 (bitmap_set_bit): Return whether the bit changed.
1551 (bitmap_clear_bit): Likewise.
1552 * tree-ssa.c (verify_vssa): Make use of the changed state
1553 from bitmap_set_bit.
1555 2021-11-03 Richard Biener <rguenther@suse.de>
1557 PR middle-end/103033
1558 * internal-fn.c (expand_DEFERRED_INIT): Elide the
1559 native_interpret_expr path in favor of folding the
1560 VIEW_CONVERT_EXPR generated when punning the RHS.
1562 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1564 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
1567 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
1570 * ipa-modref.c (callee_to_caller_flags): New function.
1571 (modref_eaf_analysis::analyze_ssa_name): Use it.
1572 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
1574 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1576 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
1579 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1580 Uroš Bizjak <ubizjak@gmail.com>
1582 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
1583 rotations by non-constant amounts.
1585 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1587 * ipa-modref.c: Fix anonymous namespace placement.
1588 (class modref_eaf_analysis): New class.
1589 (analyze_ssa_name_flags): Turn to ...
1590 (modref_eaf_analysis::analyze_ssa_name): ... this one.
1591 (merge_call_lhs_flags): Turn to ...
1592 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
1593 (modref_eaf_analysis::merge_with_ssa_name): New member function.
1594 (record_escape_points): Turn to ...
1595 (modref_eaf_analysis::record_escape_points): ... this one.
1596 (analyze_parms): Updat
1597 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
1599 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1601 * gimple.c (gimple_call_static_chain_flags): New function.
1602 * gimple.h (gimple_call_static_chain_flags): Declare
1603 * ipa-modref.c (modref_summary::modref_summary): Initialize
1605 (modref_summary_lto::modref_summary_lto): Likewise.
1606 (modref_summary::useful_p): Test static_chain_flags.
1607 (modref_summary_lto::useful_p): Likewise.
1608 (struct modref_summary_lto): Add static_chain_flags.
1609 (modref_summary::dump): Dump static_chain_flags.
1610 (modref_summary_lto::dump): Likewise.
1611 (struct escape_point): Add static_cahin_arg.
1612 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
1613 (analyze_parms): Handle static chains.
1614 (modref_summaries::duplicate): Duplicate static_chain_flags.
1615 (modref_summaries_lto::duplicate): Likewise.
1616 (modref_write): Stream static_chain_flags.
1617 (read_section): Likewise.
1618 (modref_merge_call_site_flags): Handle static_chain_flags.
1619 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
1620 * tree-ssa-structalias.c (handle_rhs_call): Use
1621 gimple_static_chain_flags.
1623 2021-11-02 Richard Biener <rguenther@suse.de>
1625 PR tree-optimization/103029
1626 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
1627 the PHI nodes in the loop maintain their original operand
1630 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1632 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
1633 (EAF_NOREAD): Renumber.
1634 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
1635 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
1636 (deref_flags): Likewise.
1637 (modref_lattice::init): Likewise.
1638 (modref_lattice::merge): Likewise.
1639 (merge_call_lhs_flags): Likewise.
1640 (analyze_ssa_name_flags): Likewise.
1641 (modref_merge_call_site_flags): Likewise.
1642 * tree-ssa-structalias.c (handle_call_arg): Likewise.
1644 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
1646 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
1647 check and remove an unused local variable with shNadd/shNadd.uw
1650 2021-11-02 Tamar Christina <tamar.christina@arm.com>
1652 PR tree-optimization/103007
1653 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
1656 2021-11-02 Richard Biener <rguenther@suse.de>
1658 PR middle-end/103038
1659 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
1660 (can_native_interpret_type_p): Likewise.
1661 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
1662 RHS if the LHS is an SSA name.
1664 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
1666 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
1667 * langhooks-def.h (lhd_simulate_record_decl): Declare.
1668 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
1669 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
1670 * langhooks.c (lhd_simulate_record_decl): New function.
1672 2021-11-02 Jakub Jelinek <jakub@redhat.com>
1675 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
1676 modes with TImode inner mode if 32-bit.
1678 2021-11-02 liuhongt <hongtao.liu@intel.com>
1680 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
1681 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
1682 vect_slp_v4si_store_unalign): Document efficient target.
1683 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
1684 vect_slp_v16qi_store_unalign_1): Ditto.
1685 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
1686 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
1687 (struct_4char_block_move, struct_8char_block_move,
1688 struct_16char_block_move): Ditto.
1690 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1691 Jakub Jelinek <jakub@redhat.com>
1694 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
1695 ix86_expand_ti_to_v1ti): New helper functions.
1696 (ix86_expand_v1ti_shift): Check if the amount operand is an
1697 integer constant, and expand as a TImode shift if it isn't.
1698 (ix86_expand_v1ti_rotate): Check if the amount operand is an
1699 integer constant, and expand as a TImode rotate if it isn't.
1700 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
1701 right shifts of V1TImode quantities.
1702 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
1703 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
1704 to QImode general_operand, and let the helper functions lower
1705 shifts by non-constant operands, as TImode shifts. Make
1706 conditional on TARGET_64BIT.
1707 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
1708 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
1709 Make conditional on TARGET_64BIT.
1711 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1713 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
1714 "*tabort_1_plus"): Remove operands non-null check.
1716 2021-11-01 Martin Liska <mliska@suse.cz>
1718 * opt-functions.awk: Add new sanity checking.
1719 * optc-gen.awk: Add new argument to integer_range_info.
1720 * params.opt: Update 2 params which have negative IntegerRange.
1722 2021-11-01 qing zhao <qing.zhao@oracle.com>
1724 * gimplify.c (gimplify_decl_expr): Do not add call to
1725 __builtin_clear_padding when a variable is a gimple register
1726 or it might not have padding.
1727 (gimplify_init_constructor): Likewise.
1729 2021-11-01 Tamar Christina <tamar.christina@arm.com>
1731 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
1732 movi, dup and extract costing fields.
1733 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
1734 thunderx_extra_costs, thunderx2t99_extra_costs,
1735 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
1737 * config/arm/aarch-cost-tables.h (generic_extra_costs,
1738 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
1739 exynosm1_extra_costs, xgene1_extra_costs): Likewise
1740 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
1741 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
1743 2021-11-01 Tamar Christina <tamar.christina@arm.com>
1745 * cse.c (add_to_set): New.
1746 (find_sets_in_insn): Register constants in sets.
1747 (canonicalize_insn): Use auto_vec instead.
1748 (cse_insn): Try materializing using vec_dup.
1749 * rtl.h (simplify_context::simplify_gen_vec_select,
1750 simplify_gen_vec_select): New.
1751 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
1753 2021-11-01 David Malcolm <dmalcolm@redhat.com>
1755 * common.opt (fdiagnostics-escape-format=): New.
1756 (diagnostics_escape_format): New enum.
1757 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
1758 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
1759 * diagnostic-format-json.cc (json_end_diagnostic): Add
1760 "escape-source" attribute.
1761 * diagnostic-show-locus.c
1762 (exploc_with_display_col::exploc_with_display_col): Replace
1763 "tabstop" param with a cpp_char_column_policy and add an "aspect"
1764 param. Use these to compute m_display_col accordingly.
1765 (struct char_display_policy): New struct.
1766 (layout::m_policy): New field.
1767 (layout::m_escape_on_output): New field.
1768 (def_policy): New function.
1769 (make_range): Update for changes to exploc_with_display_col ctor.
1770 (default_print_decoded_ch): New.
1771 (width_per_escaped_byte): New.
1772 (escape_as_bytes_width): New.
1773 (escape_as_bytes_print): New.
1774 (escape_as_unicode_width): New.
1775 (escape_as_unicode_print): New.
1777 (layout::layout): Initialize new fields. Update m_exploc ctor
1778 call for above change to ctor.
1779 (layout::maybe_add_location_range): Update for changes to
1780 exploc_with_display_col ctor.
1781 (layout::calculate_x_offset_display): Update for change to
1783 (layout::print_source_line): Pass policy
1784 to cpp_display_width_computation. Capture cpp_decoded_char when
1785 calling process_next_codepoint. Move printing of source code to
1786 m_policy.m_print_cb.
1787 (line_label::line_label): Pass in policy rather than context.
1788 (layout::print_any_labels): Update for change to line_label ctor.
1789 (get_affected_range): Pass in policy rather than context, updating
1790 calls to location_compute_display_column accordingly.
1791 (get_printed_columns): Likewise, also for cpp_display_width.
1792 (correction::correction): Pass in policy rather than tabstop.
1793 (correction::compute_display_cols): Pass m_policy rather than
1794 m_tabstop to cpp_display_width.
1795 (correction::m_tabstop): Replace with...
1796 (correction::m_policy): ...this.
1797 (line_corrections::line_corrections): Pass in policy rather than
1799 (line_corrections::m_context): Replace with...
1800 (line_corrections::m_policy): ...this.
1801 (line_corrections::add_hint): Update to use m_policy rather than
1803 (line_corrections::add_hint): Likewise.
1804 (layout::print_trailing_fixits): Likewise.
1805 (selftest::test_display_widths): New.
1806 (selftest::test_layout_x_offset_display_utf8): Update to use
1807 policy rather than tabstop.
1808 (selftest::test_one_liner_labels_utf8): Add test of escaping
1810 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
1811 use policy rather than tabstop.
1812 (selftest::test_overlapped_fixit_printing): Likewise.
1813 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
1814 (selftest::test_overlapped_fixit_printing_2): Likewise.
1815 (selftest::test_tab_expansion): Likewise.
1816 (selftest::test_escaping_bytes_1): New.
1817 (selftest::test_escaping_bytes_2): New.
1818 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
1819 * diagnostic.c (diagnostic_initialize): Initialize
1820 context->escape_format.
1821 (convert_column_unit): Update to use default character width policy.
1822 (selftest::test_diagnostic_get_location_text): Likewise.
1823 * diagnostic.h (enum diagnostics_escape_format): New enum.
1824 (diagnostic_context::escape_format): New field.
1825 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
1826 (-fdiagnostics-format=): Add "escape-source" attribute to examples
1827 of JSON output, and document it.
1828 * input.c (location_compute_display_column): Pass in "policy"
1829 rather than "tabstop", passing to
1830 cpp_byte_column_to_display_column.
1831 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
1832 * input.h (class cpp_char_column_policy): New forward decl.
1833 (location_compute_display_column): Pass in "policy" rather than
1835 * opts.c (common_handle_option): Handle
1836 OPT_fdiagnostics_escape_format_.
1837 * selftest.c (temp_source_file::temp_source_file): New ctor
1838 overload taking a size_t.
1839 * selftest.h (temp_source_file::temp_source_file): Likewise.
1841 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
1843 * dbgcnt.def: Add debug counter for back_thread[12] and
1844 back_threadfull[12].
1845 * passes.def: Pass "first" argument to each back threading pass.
1846 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
1848 (back_threader::debug_counter): New.
1849 (back_threader::maybe_register_path): Call debug_counter.
1851 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
1853 * tree-ssa-threadbackward.c (BT_NONE): New.
1856 (back_threader::back_threader): Add flags.
1857 Move loop initialization here.
1858 (back_threader::~back_threader): New.
1859 (back_threader::find_taken_edge_switch): Change solver and ranger
1861 (back_threader::find_taken_edge_cond): Same.
1862 (back_threader::find_paths_to_names): Same.
1863 (back_threader::find_paths): Same.
1864 (back_threader::dump): Same.
1865 (try_thread_blocks): Merge into thread_blocks.
1866 (back_threader::thread_blocks): New.
1867 (do_early_thread_jumps): Merge into thread_blocks.
1868 (do_thread_jumps): Merge into thread_blocks.
1869 (back_threader::thread_through_all_blocks): Remove.
1871 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
1873 PR tree-optimization/103003
1874 * value-relation.cc (dom_oracle::register_relation): If the 2
1875 ssa names are the same, don't register any relation.
1877 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
1879 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
1880 redundant check for calls_eh_return.
1881 * config/aarch64/aarch64.md (*do_return): Likewise.
1883 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
1885 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
1886 duplicate_loop_to_header_edge to
1887 duplicate_loop_body_to_header_edge.
1888 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
1889 * cfghooks.h (struct cfg_hooks): Likewise.
1890 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
1891 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
1892 (clone_loop_to_header_edge): Likewise.
1893 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
1894 * cfgrtl.c (struct cfg_hooks): Likewise.
1895 * doc/loop.texi: Likewise.
1896 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
1897 (unroll_loop_runtime_iterations): Likewise.
1898 (unroll_loop_stupid): Likewise.
1899 (apply_opt_in_copies): Likewise.
1900 * tree-cfg.c (struct cfg_hooks): Likewise.
1901 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
1902 (try_peel_loop): Likewise.
1903 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
1904 (gimple_duplicate_loop_body_to_header_edge): Likewise.
1905 (tree_transform_and_unroll_loop): Likewise.
1906 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
1909 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
1911 * cfgloopmanip.c (loop_version): Refactor loopify to
1912 loop_version. Move condition generation after loopify.
1914 * cfgloopmanip.h (loopify): Delete.
1916 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
1918 * ipa-fnsummary.c: Include tree-dfa.h.
1919 (points_to_local_or_readonly_memory_p): Return true on return
1921 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
1924 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
1926 PR middle-end/102972
1927 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
1928 internal Fortran name; new permit_num_teams arg to permit
1929 omp_get_num_teams and omp_get_team_num.
1930 (scan_omp_1_stmt): Update call to it, add missing call for
1931 reverse offload, and check for strictly nested API calls in teams.
1933 2021-10-30 Jakub Jelinek <jakub@redhat.com>
1935 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
1937 2021-10-29 David Malcolm <dmalcolm@redhat.com>
1939 * selftest.c (assert_streq): Add newlines when emitting non-equal
1942 2021-10-29 David Malcolm <dmalcolm@redhat.com>
1944 * Makefile.in: Fix syntax for reference to LIBDEPS in
1947 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
1949 * doc/install.texi: Bump required minimum DejaGnu version.
1951 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
1952 Andrew MacLeod <amacleod@redhat.com>
1954 * value-relation.cc (path_oracle::killing_def): Add a
1955 self-equivalence so we don't look to the root oracle.
1957 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
1959 * passes.def: Replace the pass_thread_jumps before VRP* with
1960 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
1961 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
1962 Remove hyphen from "thread-full" name.
1964 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
1966 PR tree-optimization/102983
1967 * gimple-range-cache.h (propagate_updated_value): Make public.
1968 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
1969 when processing gcond stmts.
1971 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
1973 * gimple.c (gimple_call_retslot_flags): New function.
1974 * gimple.h (gimple_call_retslot_flags): Declare.
1975 * ipa-modref.c: Include tree-cfg.h.
1976 (struct escape_entry): Turn parm_index to signed.
1977 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
1978 (modref_summary::modref_summary): Initialize retslot_flags.
1979 (struct modref_summary_lto): Likewise.
1980 (modref_summary::useful_p): Check retslot_flags.
1981 (modref_summary_lto::useful_p): Likewise.
1982 (modref_summary::dump): Dump retslot_flags.
1983 (modref_summary_lto::dump): Likewise.
1984 (struct escape_point): Add hidden_args enum.
1985 (analyze_ssa_name_flags): Ignore return slot return;
1986 use gimple_call_retslot_flags.
1987 (record_escape_points): Break out from ...
1988 (analyze_parms): ... here; handle retslot_flags.
1989 (modref_summaries::duplicate): Duplicate retslot_flags.
1990 (modref_summaries_lto::duplicate): Likewise.
1991 (modref_write_escape_summary): Stream parm_index as signed.
1992 (modref_read_escape_summary): Likewise.
1993 (modref_write): Stream retslot_flags.
1994 (read_section): Likewise.
1995 (struct escape_map): Fix typo in comment.
1996 (update_escape_summary_1): Fix whitespace.
1997 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
1998 (modref_merge_call_site_flags): Merge retslot_flags.
1999 * ipa-modref.h (struct modref_summary): Add retslot_flags.
2000 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
2002 2021-10-29 Tamar Christina <tamar.christina@arm.com>
2004 PR tree-optimization/102977
2005 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
2006 (vect_detect_pair_op): Add crosslane check.
2007 (vect_match_call_complex_mla): Remove.
2008 (class complex_mul_pattern): Update comment.
2009 (complex_mul_pattern::matches): Update detection.
2010 (class complex_fma_pattern): Remove.
2011 (complex_fma_pattern::matches): Remove.
2012 (complex_fma_pattern::recognize): Remove.
2013 (complex_fma_pattern::build): Remove.
2014 (class complex_fms_pattern): Update comment.
2015 (complex_fms_pattern::matches): Remove.
2016 (complex_operations_pattern::recognize): Remove complex_fma_pattern
2018 2021-10-29 Jakub Jelinek <jakub@redhat.com>
2020 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
2023 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
2026 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
2027 patterns match and emit for VSX xxpermdi.
2029 2021-10-29 liuhongt <hongtao.liu@intel.com>
2032 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
2034 (V16HF_FTYPE_V16HF): Ditto.
2035 (V32HF_FTYPE_V32HF): Ditto.
2036 (V8HF_FTYPE_V8HF_ROUND): Ditto.
2037 (V16HF_FTYPE_V16HF_ROUND): Ditto.
2038 (V32HF_FTYPE_V32HF_ROUND): Ditto.
2039 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
2040 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
2041 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
2042 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
2043 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
2044 * config/i386/i386-builtins.c
2045 (ix86_builtin_vectorized_function): Enable vectorization for
2046 HFmode FLOOR/CEIL/TRUNC operation.
2047 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
2049 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
2052 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2053 Andrew MacLeod <amacleod@redhat.com>
2055 * value-relation.cc (path_oracle::killing_def): Walk the
2056 equivalency list and remove SSA from any equivalencies.
2058 2021-10-28 Stafford Horne <shorne@gmail.com>
2060 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
2063 2021-10-28 Jakub Jelinek <jakub@redhat.com>
2065 PR tree-optimization/102951
2066 * fold-const.h (address_compare): Declare.
2067 * fold-const.c (address_compare): New function.
2068 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
2069 address_compare helper.
2070 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
2073 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
2075 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
2077 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
2079 PR tree-optimization/102940
2080 * tree-vrp.c (execute_ranger_vrp): Reset scev.
2082 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
2084 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
2086 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
2088 * configure: Regenerate.
2089 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
2091 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
2093 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
2095 2021-10-28 Richard Biener <rguenther@suse.de>
2098 * fold-const.c (fold_convert_const): Avoid int to float
2099 constant folding with -frounding-math and inexact result.
2100 * simplify-rtx.c (simplify_const_unary_operation): Likewise
2101 for both float and unsigned_float.
2103 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2105 * tree-ssa-threadbackward.c
2106 (back_threader::find_taken_edge_switch): Use find_case_label_range
2107 instead of find_taken_edge.
2109 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2111 * tree-ssa-threadbackward.c (class back_threader_registry):
2112 Inherit from back_jt_path_registry.
2113 (back_threader_registry::thread_through_all_blocks): Remove.
2114 (back_threader_registry::register_path): Remove
2115 m_lowlevel_registry prefix.
2117 2021-10-28 Richard Biener <rguenther@suse.de>
2120 * fold-const.c (fold_convert_const_real_from_real): Honor
2121 -frounding-math if the conversion is not exact.
2122 * simplify-rtx.c (simplify_const_unary_operation): Do not
2123 simplify FLOAT_TRUNCATE with sign dependent rounding.
2125 2021-10-28 Richard Biener <rguenther@suse.de>
2127 PR tree-optimization/102949
2128 * tree-vect-stmts.c (ensure_base_align): Look at the
2129 dr_info of a group leader and assert we are looking at
2130 one with analyzed alignment.
2132 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
2135 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
2136 V1T1 mode for unaligned load and store.
2138 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
2140 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
2143 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
2145 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
2148 2021-10-28 Alexandre Oliva <oliva@adacore.com>
2150 * common.opt (fharden-compares): New.
2151 (fharden-conditional-branches): New.
2152 * doc/invoke.texi: Document new options.
2153 * gimple-harden-conditionals.cc: New.
2154 * Makefile.in (OBJS): Build it.
2155 * passes.def: Add new passes.
2156 * tree-pass.h (make_pass_harden_compares): Declare.
2157 (make_pass_harden_conditional_branches): Declare.
2159 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2162 * config/rs6000/altivec.md: Add vsx register constraints.
2163 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
2164 (vsx_xxsel<mode>2): Likewise.
2165 (vsx_xxsel<mode>3): Likewise.
2166 (vsx_xxsel<mode>4): Likewise.
2168 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2171 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
2172 (altivec_vsel<mode>): ... this and update define.
2173 (*altivec_vsel<mode>_uns): Delete.
2174 (altivec_vsel<mode>2): New define_insn.
2175 (altivec_vsel<mode>3): Likewise.
2176 (altivec_vsel<mode>4): Likewise.
2177 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
2178 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
2180 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
2181 selection instead of per element.
2182 * config/rs6000/vector.md:
2183 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
2184 (vsx_xxsel<mode>): ... this and update define.
2185 (*vsx_xxsel<mode>_uns): Delete.
2186 (vsx_xxsel<mode>2): New define_insn.
2187 (vsx_xxsel<mode>3): Likewise.
2188 (vsx_xxsel<mode>4): Likewise.
2190 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
2192 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
2193 HFmode rsqrt without TARGET_SSE_MATH.
2194 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
2195 btrunc condition to be restricted by -ftrapping-math, adjust
2196 use_rsqrt_p function call.
2197 * config/i386/i386.md (rcphf2): New define_insn.
2198 (rsqrthf2): Likewise.
2199 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
2200 (div<mode>3): New expander for HF mode.
2201 (rsqrt<mode>2): Likewise.
2202 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
2203 (*avx512fp16_vmrsqrtv8hf2): Likewise.
2205 2021-10-27 Saagar Jha <saagar@saagarjha.com>
2207 * config.gcc: Adjust for Darwin21.
2208 * config/darwin-c.c (macosx_version_as_macro): Likewise.
2209 * config/darwin-driver.c (validate_macosx_version_min):
2211 (darwin_find_version_from_kernel): Likewise.
2213 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2215 * gimple-range-path.cc
2216 (path_range_query::range_defined_in_block): Call killing_def.
2218 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2220 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
2221 operands as well as relation.
2222 * gimple-range-path.cc
2223 (path_range_query::compute_ranges_in_block): Compute PHI relations
2224 first. Compute outgoing relations at the end.
2225 (path_range_query::compute_ranges): Remove call to compute_relations.
2226 (path_range_query::compute_relations): Remove.
2227 (path_range_query::maybe_register_phi_relation): New.
2228 (path_range_query::compute_phi_relations): Abstract out
2229 registering one PHI relation to...
2230 (path_range_query::compute_outgoing_relations): ...here.
2231 * gimple-range-path.h (class path_range_query): Remove
2233 Add maybe_register_phi_relation.
2235 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2236 Andrew MacLeod <amacleod@redhat.com>
2238 * value-relation.cc (path_oracle::killing_def): Kill second
2241 2021-10-27 Martin Jambor <mjambor@suse.cz>
2243 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
2244 profile feedback depending on their local availability.
2246 2021-10-27 Martin Jambor <mjambor@suse.cz>
2248 * params.opt (param_ipa_cp_profile_count_base): New parameter.
2249 * doc/invoke.texi (Optimize Options): Add entry for
2250 ipa-cp-profile-count-base.
2251 * ipa-cp.c (max_count): Replace with base_count, replace all
2252 occurrences too, unless otherwise stated.
2253 (ipcp_cloning_candidate_p): identify mostly-directly called
2254 functions based on their counts, not max_count.
2255 (compare_edge_profile_counts): New function.
2256 (ipcp_propagate_stage): Instead of setting max_count, find the
2257 appropriate edge count in a sorted vector of counts of eligible
2258 edges and make it the base_count.
2260 2021-10-27 Martin Jambor <mjambor@suse.cz>
2262 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
2263 n_nonrec_calls and itself, document all fields.
2264 (init_caller_stats): Initialize the above new fields.
2265 (gather_caller_stats): Gather self-recursive counts and calls number.
2266 (get_info_about_necessary_edges): Gather counts of self-recursive and
2267 other edges bringing in the requested value separately.
2268 (dump_profile_updates): Rework to dump info about a single node only.
2269 (lenient_count_portion_handling): New function.
2270 (struct gather_other_count_struct): New type.
2271 (gather_count_of_non_rec_edges): New function.
2272 (struct desc_incoming_count_struct): New type.
2273 (analyze_clone_icoming_counts): New function.
2274 (adjust_clone_incoming_counts): Likewise.
2275 (update_counts_for_self_gen_clones): Likewise.
2276 (update_profiling_info): Rewritten.
2277 (update_specialized_profile): Adjust call to dump_profile_updates.
2278 (create_specialized_node): Do not update profiling info.
2279 (decide_about_value): New parameter self_gen_clones, either push new
2280 clones into it or updat their profile counts. For self-recursively
2281 generated values, use a portion of the node count instead of count
2282 from self-recursive edges to estimate goodness.
2283 (decide_whether_version_node): Gather clones for self-generated values
2284 in a new vector, update their profiles at once at the end.
2286 2021-10-27 Richard Biener <rguenther@suse.de>
2288 * tree-vectorizer.c (vect_transform_loops): New function,
2290 (try_vectorize_loop_1): ... here. Simplify as epilogues
2291 are now fully handled in the split part.
2293 2021-10-27 Jakub Jelinek <jakub@redhat.com>
2295 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
2296 iterators with pointer types.
2297 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
2299 2021-10-26 Martin Sebor <msebor@redhat.com>
2301 PR tree-optimization/102238
2302 PR tree-optimization/102919
2303 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
2304 (array_elt_at_offset): Move to pointer-query.
2305 (set_aggregate_size_and_offset): New function.
2306 (field_at_offset): Move to pointer-query.
2307 (get_origin_and_offset): Rename...
2308 (get_origin_and_offset_r): this. Add an argument. Make aggregate
2309 handling more robust.
2310 (get_origin_and_offset): New.
2311 (alias_offset): Add an argument.
2312 (format_string): Use subobject size determined by get_origin_and_offset.
2313 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
2314 Improve/correct handling of aggregates.
2315 (array_elt_at_offset): Same.
2316 * pointer-query.h (field_at_offset): Declare.
2317 (array_elt_at_offset): Declare.
2319 2021-10-26 Martin Sebor <msebor@redhat.com>
2321 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
2323 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
2324 (expand_builtin_fork_or_exec): Same.
2325 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
2326 compute_objsize additional arguments.
2327 (inbounds_memaccess_p): Same.
2328 (array_bounds_checker::check_array_bounds): Add an assert. Stash
2329 statement in a member.
2330 (check_array_bounds_dom_walker::before_dom_children): Same.
2331 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
2332 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
2333 (handle_printf_call): Pass a new argument.
2334 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
2335 (check_access): Add an argument and pass it along to callees.
2336 (check_read_access): Make a member function.
2337 (pass_waccess::check_strcat): Pass access_data ctor additional
2339 (pass_waccess::check_strncat): Same.
2340 (pass_waccess::check_stxcpy): Same.
2341 (pass_waccess::check_stxncpy): Same.
2342 (pass_waccess::check_strncmp): Same.
2343 (pass_waccess::check_read_access): Same.
2344 (pass_waccess::check_builtin): Same.
2345 (pass_waccess::maybe_check_access_sizes): Same.
2346 (pass_waccess::maybe_check_dealloc_call): Same.
2347 * gimple-ssa-warn-access.h (check_read_access): Declare a new
2349 * pointer-query.cc (compute_objsize_r): Add an argument.
2350 (gimple_call_return_array): Same.
2351 (gimple_call_alloc_size): Same.
2352 (access_ref::access_ref): Same.
2353 (access_ref::get_ref): Same.
2354 (pointer_query::get_ref): Same.
2355 (handle_min_max_size): Pass an arguments to callees.
2356 (handle_array_ref): Add an argument.
2357 (handle_mem_ref): Same.
2358 (compute_objsize): Same.
2359 * pointer-query.h (struct access_ref): Adjust signatures.
2360 (struct access_data): Same.
2361 (gimple_call_alloc_size): Add an argument.
2362 (gimple_parm_array_size): Same.
2363 (compute_objsize): Same.
2364 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
2365 argument to compute_objsize.
2366 (strlen_pass::maybe_warn_overflow): Same.
2367 (maybe_diag_stxncpy_trunc): Same.
2369 2021-10-26 Martin Sebor <msebor@redhat.com>
2371 PR middle-end/102453
2372 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
2373 (pass_waccess::check_atomic_builtin): Call it.
2375 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
2377 PR rtl-optimization/102842
2378 * lra-constraints.c (match_reload): Ignore out in checking values
2380 (curr_insn_transform): Collect outputs before doing reloads of operands.
2382 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
2385 * config/rs6000/x86intrin.h: Move some included headers to new
2386 headers. Include new immintrin.h instead of those headers.
2387 * config/rs6000/immintrin.h: New.
2388 * config/rs6000/x86gprintrin.h: New.
2389 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
2391 2021-10-26 Richard Biener <rguenther@suse.de>
2393 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
2394 Remove byte_offset parameter.
2395 (vect_create_data_ref_ptr): Likewise.
2396 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2398 (vect_create_data_ref_ptr): Likewise.
2399 * tree-vect-stmts.c (vectorizable_store): Adjust.
2400 (vectorizable_load): Likewise.
2402 2021-10-26 Richard Biener <rguenther@suse.de>
2404 PR tree-optimization/96109
2405 * tree-vectorizer.h (dr_misalignment): Add optional offset
2407 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
2408 offset applied for negative stride accesses.
2409 (vect_enhance_data_refs_alignment): Compute negative stride
2410 access offset and pass it to dr_misalignment.
2411 * tree-vect-stmts.c (get_negative_load_store_type): Pass
2412 negative offset to dr_misalignment.
2413 (get_group_load_store_type): Likewise.
2414 (get_load_store_type): Likewise.
2415 (vectorizable_store): Remove asserts about alignment.
2416 (vectorizable_load): Likewise.
2418 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2420 PR tree-optimization/102897
2421 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
2423 2021-10-26 Richard Biener <rguenther@suse.de>
2425 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2426 Take offset in bytes.
2427 (vect_create_data_ref_ptr): Likewise.
2428 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
2429 offset by element size.
2430 (vect_create_cond_for_align_checks): Likewise.
2431 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
2432 (vectorizable_load): Remove duplicate leftover from merge
2435 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
2437 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
2438 function to expand V1TI mode logical shifts by integer constants.
2439 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
2440 mode rotations by integer constants.
2441 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
2442 ix86_expand_v1ti_rotate): Prototype new functions here.
2443 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
2444 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
2446 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2448 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2449 Avoid threading circular paths.
2451 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2453 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
2454 Attempt to resolve all incoming paths to a PHI.
2455 (back_threader::resolve_def): Always return true for PHIs.
2457 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2459 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
2460 Always try to resolve path without looking back.
2461 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
2462 edge is a back edge.
2464 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2466 PR tree-optimization/102789
2467 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2468 update inits of simd_lane_access.
2470 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2472 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
2473 move to vr-values.c.
2474 (simplify_casted_conds): Move to vrp_folder class.
2475 (execute_vrp): Call via vrp_folder now.
2476 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
2477 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
2478 * vr-values.h (simplify_casted_cond): Add prototype.
2480 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2482 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
2485 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
2487 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
2489 (_mm_testc_si128): Likewise.
2490 (_mm_testnzc_si128): Likewise.
2491 (_mm_blend_ps): Likewise.
2492 (_mm_blendv_ps): Likewise.
2493 (_mm_blend_pd): Likewise.
2494 (_mm_blendv_pd): Likewise.
2495 (_mm_ceil_pd): Likewise.
2496 (_mm_ceil_sd): Likewise.
2497 (_mm_ceil_ps): Likewise.
2498 (_mm_ceil_ss): Likewise.
2499 (_mm_floor_pd): Likewise.
2500 (_mm_floor_sd): Likewise.
2501 (_mm_floor_ps): Likewise.
2502 (_mm_floor_ss): Likewise.
2503 (_mm_minpos_epu16): Likewise.
2504 (_mm_mul_epi32): Likewise.
2505 (_mm_cvtepi8_epi16): Likewise.
2506 (_mm_packus_epi32): Likewise.
2507 (_mm_cmpgt_epi64): Likewise.
2509 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
2511 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
2512 shifts of the mode's smin_value and smax_value when the bit count
2513 operand doesn't have side-effects.
2514 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
2515 when the bit count operand doesn't have side-effects.
2516 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
2517 compile-time evaluation of saturating left shifts with constant
2520 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2522 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
2525 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2527 * flag-types.h (enum ranger_debug): Adjust values.
2528 * params.opt (ranger_debug): Ditto.
2530 2021-10-25 Martin Jambor <mjambor@suse.cz>
2532 PR tree-optimization/102886
2533 * tree-sra.c (totally_scalarize_subtree): Fix the out of
2536 2021-10-25 Andrew Pinski <apinski@marvell.com>
2538 * tree-ssa-dce.c (simple_dce_from_worklist):
2539 Check stmt_unremovable_because_of_non_call_eh_p also
2540 before removing the statement.
2542 2021-10-25 Richard Biener <rguenther@suse.de>
2544 PR tree-optimization/102905
2545 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2546 Use vect_supportable_dr_alignment again to determine whether
2547 an access is supported when not aligned.
2549 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2551 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
2554 2021-10-25 Jim Wilson <jimw@sifive.com>
2555 Kito Cheng <kito.cheng@sifive.com>
2556 Jia-Wei Chen <jiawei@iscas.ac.cn>
2557 Shi-Hua Liao <shihua@iscas.ac.cn>
2559 * config/riscv/bitmanip.md (shiftm1): New.
2560 (*bset<mode>): Ditto.
2561 (*bset<mode>_mask): Ditto.
2562 (*bset<mode>_1): Ditto.
2563 (*bset<mode>_1_mask): Ditto.
2564 (*bseti<mode>): Ditto.
2565 (*bclr<mode>): Ditto.
2566 (*bclri<mode>): Ditto.
2567 (*binv<mode>): Ditto.
2568 (*binvi<mode>): Ditto.
2569 (*bext<mode>): Ditto.
2571 * config/riscv/predicates.md (splittable_const_int_operand):
2573 (single_bit_mask_operand): New.
2574 (not_single_bit_mask_operand): Ditto.
2575 (const31_operand): Ditto.
2576 (const63_operand): Ditto.
2577 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
2578 (riscv_output_move): Ditto.
2579 (riscv_print_operand): Handle new operand type: T and S.
2580 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
2582 2021-10-25 Jim Wilson <jimw@sifive.com>
2584 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
2587 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2589 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
2591 (riscv_rtx_costs): Ditto.
2593 2021-10-25 Jim Wilson <jimw@sifive.com>
2594 Kito Cheng <kito.cheng@sifive.com>
2595 Jia-Wei Chen <jiawei@iscas.ac.cn>
2597 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
2598 (bitmanip_minmax): New.
2599 (clz_ctz_pcnt): New.
2600 (bitmanip_optab): New.
2601 (bitmanip_insn): New.
2602 (*<optab>_not<mode>): New.
2603 (*xor_not<mode>): New.
2604 (<bitmanip_optab>si2): New.
2605 (*<bitmanip_optab>disi2): New.
2606 (<bitmanip_optab>di2): New.
2607 (*zero_extendhi<GPR:mode>2_bitmanip): New.
2608 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
2609 (*zero_extendhi<GPR:mode>2_zbb): New.
2612 (rotrsi3_sext): New.
2615 (rotlsi3_sext): New.
2616 (bswap<mode>2): New.
2617 (<bitmanip_optab><mode>3): New.
2618 * config/riscv/riscv.md (type): Add rotate.
2619 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
2620 (*zero_extendhi<GPR:mode>2): New.
2621 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
2622 (*extend<SHORT:mode><SUPERQI:mode>2): New.
2624 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2626 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
2628 (riscv_rtx_costs): Ditto.
2630 2021-10-25 Jim Wilson <jimw@sifive.com>
2631 Kito Cheng <kito.cheng@sifive.com>
2632 Jia-Wei Chen <jiawei@iscas.ac.cn>
2634 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
2639 (riscv_rtx_costs): Ditto.
2640 * config/riscv/riscv.md: Include bitmanip.md
2641 (type): Add bitmanip bype.
2642 (zero_extendsidi2): Change to define_expand pattern.
2643 (*zero_extendsidi2_internal): New.
2644 (zero_extendsidi2_shifted): Disable for ZBA.
2646 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2648 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
2649 Add zba, zbb, zbc and zbs.
2650 (riscv_ext_flag_table): Ditto.
2651 * config/riscv/riscv-opts.h (MASK_ZBA): New.
2655 (TARGET_ZBA): Ditto.
2656 (TARGET_ZBB): Ditto.
2657 (TARGET_ZBC): Ditto.
2658 (TARGET_ZBS): Ditto.
2659 * config/riscv/riscv.opt (riscv_zb_subext): New.
2661 2021-10-25 liuhongt <hongtao.liu@intel.com>
2664 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
2665 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
2667 2021-10-25 Richard Biener <rguenther@suse.de>
2669 PR tree-optimization/102920
2670 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
2671 controlling VN_TOP matching behavior.
2672 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
2673 (vn_phi_eq): Do not optimistically match VN_TOP.
2675 2021-10-25 konglin1 <lingling.kong@intel.com>
2677 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
2678 define_insn_and_split.
2679 (fma_<mode>_fadd_fcmul):Likewise
2680 (fma_<complexopname>_<mode>_fma_zero):Likewise
2682 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
2684 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
2685 TARGET_DISABLE_FPREGS.
2686 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
2687 MASK_DISABLE_FPREGS.
2688 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
2689 cost of hardware integer multiplication.
2690 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
2691 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
2692 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
2693 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
2694 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
2695 SOFT_FLOAT. Revise comment.
2696 (msoft-float): New option.
2698 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
2700 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
2702 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
2704 * doc/install.texi (Specific): Remove obsolete details
2705 around GNU/Linux on Itanium.
2706 (Specific): Remove reference to Windows for Itanium.
2708 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
2710 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
2713 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
2715 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
2717 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
2719 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
2720 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
2721 (one_cmplv1ti2): New define expand.
2723 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
2726 * Makefile.in: Handle dvidir and install-dvi target.
2727 * configure: Regenerate.
2728 * configure.ac: Add install-dvi to target_list.
2730 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
2732 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
2735 2021-10-22 Richard Biener <rguenther@suse.de>
2737 PR tree-optimization/102893
2738 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
2739 test for an exit edge.
2741 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
2742 Andrew MacLeod <amacleod@redhat.com>
2744 * gimple-range-path.cc (path_range_query::compute_phi_relations):
2745 Kill any global relations we may know before registering a new
2747 * value-relation.cc (path_oracle::killing_def): New.
2748 * value-relation.h (path_oracle::killing_def): New.
2750 2021-10-22 Richard Biener <rguenther@suse.de>
2753 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
2755 (vn_phi_lookup): Likewise.
2757 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
2760 * doc/invoke.texi: Document -fcf-protection requires i686 or
2763 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
2765 PR middle-end/102764
2766 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
2768 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
2770 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
2771 macro and all invocations.
2773 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2775 * doc/invoke.texi (ranger-debug): Document.
2776 * flag-types.h (enum ranger_debug): New.
2777 (enum evrp_mode): Remove debug values.
2778 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
2779 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
2780 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
2781 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
2782 (execute_early_vrp): Use evrp-mode directly.
2783 * params.opt (enum evrp_mode): Remove debug values.
2784 (ranger-debug): New.
2785 (ranger-logical-depth): Relocate to be in alphabetical order.
2787 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2789 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
2790 * flag-types.h: (enum vrp_mode): New.
2791 * params.opt: (vrp1-mode, vrp2-mode): New.
2792 * tree-vrp.c (vrp_pass_num): New.
2793 (pass_vrp::pass_vrp): Set pass number.
2794 (pass_vrp::execute): Choose which VRP mode to execute.
2796 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2798 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
2799 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
2800 * tree-vrp.c (class rvrp_folder): Relocate here.
2801 (execute_ranger_vrp): New.
2802 * tree-vrp.h (execute_ranger_vrp): Export.
2804 2021-10-21 Martin Liska <mliska@suse.cz>
2808 * opts.c (finish_options): Process flag_var_tracking* options
2809 here as they can be adjusted by optimize attribute.
2810 Process also flag_syntax_only and flag_gtoggle.
2811 * toplev.c (process_options): Remove it here.
2812 * common.opt: Make debug_nonbind_markers_p as PerFunction
2813 attribute as it depends on optimization level.
2815 2021-10-21 Martin Jambor <mjambor@suse.cz>
2817 PR tree-optimization/102505
2818 * tree-sra.c (totally_scalarize_subtree): Check that the
2819 encountered field fits within the acces we would like to put it
2822 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
2824 * tree-ssa-threadbackward.c
2825 (back_threader::maybe_register_path): Remove circular paths check.
2827 2021-10-21 Richard Biener <rguenther@suse.de>
2829 * toplev.c (process_options): Move the initial debug_hooks
2831 (toplev::main): ... before the call of the post_options
2834 2021-10-21 Richard Biener <rguenther@suse.de>
2836 PR tree-optimization/102847
2837 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
2838 load cost in the prologue for VMAT_INVARIANT.
2840 2021-10-21 Richard Biener <rguenther@suse.de>
2842 PR tree-optimization/102847
2843 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
2844 handle VMAT_INVARIANT as a splat in the prologue.
2846 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
2849 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
2850 move to use the same logic as HImode.
2852 2021-10-21 Richard Biener <rguenther@suse.de>
2854 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
2855 superfluous gimple_call_nothrow_p check.
2857 2021-10-21 Andrew Pinski <apinski@marvell.com>
2859 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
2860 Mark the ssa-name of the rhs as one to be removed.
2861 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
2862 Call simple_dce_from_worklist at the end to a simple dce.
2864 2021-10-21 Andrew Pinski <apinski@marvell.com>
2866 * tree-cfg.c (maybe_remove_writeonly_store): New function
2867 factored out from ...
2868 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
2870 2021-10-21 Andrew Pinski <apinski@marvell.com>
2872 * tree-cfg.c (execute_fixup_cfg): Remove comment
2873 about standalone pass.
2875 2021-10-21 Andrew Pinski <apinski@marvell.com>
2877 * tree-cfg.c (execute_fixup_cfg): Output when the statement
2878 is removed when it is a write only var.
2880 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
2882 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2883 Avoid threading circular paths.
2885 2021-10-20 Alex Coplan <alex.coplan@arm.com>
2887 * calls.c (initialize_argument_information): Remove some dead
2888 code, remove handling for function_arg returning const_int.
2889 * doc/tm.texi: Delete documentation for unused target hooks.
2890 * doc/tm.texi.in: Likewise.
2891 * target.def (load_bounds_for_arg): Delete.
2892 (store_bounds_for_arg): Delete.
2893 (load_returned_bounds): Delete.
2894 (store_returned_bounds): Delete.
2895 * targhooks.c (default_load_bounds_for_arg): Delete.
2896 (default_store_bounds_for_arg): Delete.
2897 (default_load_returned_bounds): Delete.
2898 (default_store_returned_bounds): Delete.
2899 * targhooks.h (default_load_bounds_for_arg): Delete.
2900 (default_store_bounds_for_arg): Delete.
2901 (default_load_returned_bounds): Delete.
2902 (default_store_returned_bounds): Delete.
2904 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
2906 * doc/extend.texi (Basic Asm): Clarify that asm is not an
2908 * doc/invoke.texi (-fno-asm): Fix description for C++.
2910 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
2912 * doc/install.texi: Remove link to old.html
2914 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2916 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
2918 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2920 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
2922 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2924 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
2926 * config/aarch64/constraints.md (D1): New.
2928 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2930 * config/aarch64/aarch64-simd.md
2931 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
2932 (*aarch64_topbits_shuffle<mode>_le): New.
2933 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
2934 (*aarch64_topbits_shuffle<mode>_be): New.
2935 * config/aarch64/predicates.md
2936 (aarch64_simd_shift_imm_vec_exact_top): New.
2938 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2940 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
2941 *aarch64_<srn_op>shrn<mode>2_vect_le,
2942 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
2943 * config/aarch64/iterators.md (srn_op): New.
2945 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
2947 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
2949 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
2950 ctx->outer instead of ctx.
2952 2021-10-20 Martin Liska <mliska@suse.cz>
2955 2021-10-19 Martin Liska <mliska@suse.cz>
2958 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
2959 * system.h (strip_whilespaces): New function.
2961 2021-10-20 Martin Liska <mliska@suse.cz>
2964 2021-10-19 Martin Liska <mliska@suse.cz>
2967 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
2970 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
2972 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
2973 Change to 8 with -Os, 11 otherwise.
2975 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
2977 * config/aarch64/aarch64.c (neoversev1_tunings):
2978 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
2979 (neoversen2_tunings): Likewise.
2981 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
2984 * config/aarch64/aarch64.md (frint_pattern): Update comment.
2985 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
2986 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
2987 * config/aarch64/arm_neon.h: Likewise.
2988 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
2990 2021-10-20 Martin Liska <mliska@suse.cz>
2992 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
2993 already declared global variable.
2994 (arm_unwind_emit_set): Use out_file as function argument.
2995 (arm_unwind_emit): Likewise.
2996 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
2997 (machopic_output_stub_indirection): Likewise.
2998 (machopic_output_indirection): Likewise.
2999 (machopic_finish): Likewise.
3000 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
3001 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
3002 * config/ia64/ia64.c (process_epilogue): Likewise.
3003 (process_cfa_adjust_cfa): Likewise.
3004 (process_cfa_register): Likewise.
3005 (process_cfa_offset): Likewise.
3006 (ia64_asm_unwind_emit): Likewise.
3007 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
3009 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
3011 * config/aarch64/aarch64-builtins.c
3012 (aarch64_general_gimple_fold_builtin):
3013 lower vld1 and vst1 variants of the neon builtins
3014 * config/aarch64/aarch64-protos.h:
3015 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
3016 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
3019 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
3021 * match.pd: Generate IFN_TRUNC.
3023 2021-10-20 Richard Biener <rguenther@suse.de>
3025 PR tree-optimization/102853
3026 * tree-data-ref.c (split_constant_offset_1): Bail out
3027 immediately if the expression traps on overflow.
3029 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3031 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
3033 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3035 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
3036 Remove superflous debugging message.
3037 (back_jt_path_registry::duplicate_thread_path): Same.
3039 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3041 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
3043 (back_threader_registry::register_path): Remove m_threaded_paths.
3045 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3047 PR tree-optimization/102814
3048 * doc/invoke.texi: Document --param=max-fsm-thread-length.
3049 * params.opt: Add --param=max-fsm-thread-length.
3050 * tree-ssa-threadbackward.c
3051 (back_threader_profitability::profitable_path_p): Fail on paths
3052 longer than max-fsm-thread-length.
3054 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
3056 PR middle-end/102764
3057 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
3058 statement to reset the current location for the outgoing edges.
3060 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3061 Richard Biener <rguenther@suse.de>
3063 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
3064 on the same line as the threading cancellation.
3065 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
3066 Avoid threading through loop headers where the path remains in the
3069 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
3071 * tree-object-size.c (unknown): Make into a function. Adjust
3073 (unknown_object_size): Simplify implementation.
3075 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
3076 Kewen Lin <linkw@linux.ibm.com>
3078 * doc/sourcebuild.texi (Effective-Target Keywords): Document
3079 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
3080 vect_slp_v16qi_store, vect_slp_v2hi_store,
3081 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
3083 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
3085 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
3087 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
3091 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
3092 * config/rs6000/pmmintrin.h: Same.
3093 * config/rs6000/smmintrin.h: Same.
3094 * config/rs6000/tmmintrin.h: Same.
3096 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
3098 * config.gcc (extra_headers): Add nmmintrin.h.
3100 2021-10-19 Richard Biener <rguenther@suse.de>
3102 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
3103 misalignment parameter.
3104 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
3105 Do not change DR_MISALIGNMENT in place, instead pass the
3106 adjusted misalignment to vect_supportable_dr_alignment.
3107 (vect_peeling_supportable): Likewise.
3108 (vect_peeling_hash_get_lowest_cost): Adjust.
3109 (vect_enhance_data_refs_alignment): Likewise.
3110 (vect_vfa_access_size): Likewise.
3111 (vect_supportable_dr_alignment): Add misalignment
3112 parameter and simplify.
3113 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
3114 (get_group_load_store_type): Likewise.
3115 (get_load_store_type): Likewise.
3117 2021-10-19 Clément Chigot <clement.chigot@atos.net>
3119 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
3120 __tls_get_addr reference to .text csect.
3122 2021-10-19 Martin Liska <mliska@suse.cz>
3125 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
3128 2021-10-19 Richard Biener <rguenther@suse.de>
3130 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
3131 (vect_get_load_cost): Likewise.
3132 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3133 alignment support scheme and misalignment as arguments
3135 (vect_get_peeling_costs_all_drs): Compute that info here
3136 and note that we shouldn't need to.
3137 * tree-vect-stmts.c (vect_model_store_cost): Get
3138 alignment support scheme and misalignment as arguments.
3139 (vect_get_store_cost): Likewise.
3140 (vect_model_load_cost): Likewise.
3141 (vect_get_load_cost): Likewise.
3142 (vectorizable_store): Pass down alignment support scheme
3143 and misalignment to costing.
3144 (vectorizable_load): Likewise.
3146 2021-10-19 Richard Biener <rguenther@suse.de>
3148 * tree-vect-stmts.c (get_negative_load_store_type): Add
3149 offset output parameter and initialize it.
3150 (get_group_load_store_type): Likewise.
3151 (get_load_store_type): Likewise.
3152 (vectorizable_store): Use offset as computed by
3153 get_load_store_type.
3154 (vectorizable_load): Likewise.
3156 2021-10-19 Richard Biener <rguenther@suse.de>
3158 PR tree-optimization/102827
3159 * tree-if-conv.c (predicate_statements): Add pe parameter
3160 and use that edge to insert invariant stmts on.
3161 (combine_blocks): Pass through pe.
3162 (tree_if_conversion): Compute the edge to insert invariant
3163 stmts on and pass it along.
3165 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
3168 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
3169 sssubaddv2hi3): Swap the order of operators in vec_concat.
3171 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
3173 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
3174 (altivec_vmrghb_direct): New.
3175 (*altivec_vmrghh_internal): Delete.
3176 (altivec_vmrghh_direct): New.
3177 (*altivec_vmrghw_internal): Delete.
3178 (altivec_vmrghw_direct_<mode>): New.
3179 (altivec_vmrghw_direct): Delete.
3180 (*altivec_vmrglb_internal): Delete.
3181 (altivec_vmrglb_direct): New.
3182 (*altivec_vmrglh_internal): Delete.
3183 (altivec_vmrglh_direct): New.
3184 (*altivec_vmrglw_internal): Delete.
3185 (altivec_vmrglw_direct_<mode>): New.
3186 (altivec_vmrglw_direct): Delete.
3187 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3188 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
3190 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
3191 (vsx_xxmrglw_<mode>): Adjust.
3193 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
3195 * passes.def: Change threading comment before pass_ccp pass.
3197 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
3199 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
3200 Modify the expansion for sign extension. All extensions are done
3201 within VSX registers.
3203 2021-10-19 Richard Biener <rguenther@suse.de>
3205 * tree-vect-stmts.c (get_group_load_store_type): Add
3206 misalignment output parameter and initialize it.
3207 (get_group_load_store_type): Likewise.
3208 (vectorizable_store): Remove now redundant queries.
3209 (vectorizable_load): Likewise.
3211 2021-10-19 Richard Biener <rguenther@suse.de>
3213 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
3214 check_aligned argument.
3215 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
3217 (vect_peeling_hash_insert): Add supportable_if_not_aligned
3218 argument and do not call vect_supportable_dr_alignment here.
3219 (vect_peeling_supportable): Adjust.
3220 (vect_enhance_data_refs_alignment): Compute whether the
3221 access is supported with different alignment here and
3222 pass that down to vect_peeling_hash_insert.
3223 (vect_vfa_access_size): Adjust.
3224 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
3225 (vect_get_load_cost): Likewise.
3226 (get_negative_load_store_type): Likewise.
3227 (get_group_load_store_type): Likewise.
3228 (get_load_store_type): Likewise.
3230 2021-10-19 Martin Liska <mliska@suse.cz>
3233 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
3234 * system.h (strip_whilespaces): New function.
3236 2021-10-19 dianhong xu <dianhong.xu@intel.com>
3238 * config/i386/avx512fp16intrin.h:
3239 (_mm512_set1_pch): New intrinsic.
3240 * config/i386/avx512fp16vlintrin.h:
3241 (_mm256_set1_pch): New intrinsic.
3242 (_mm_set1_pch): Ditto.
3244 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
3246 PR tree-optimization/102796
3247 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
3248 normally. Return get_tree_range for non gimple_range_ssa_p names.
3249 (gimple_ranger::range_of_stmt): Use get_tree_range for non
3250 gimple_range_ssa_p names.
3252 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
3255 * config/i386/i386.c (ix86_print_operand_address):
3256 Error out for non-address_operand asm operands.
3258 2021-10-18 Richard Biener <rguenther@suse.de>
3260 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
3261 not auto-convert dr_alignment_support to bool.
3262 (vect_peeling_supportable): Likewise.
3263 (vect_enhance_data_refs_alignment): Likewise.
3264 (vect_supportable_dr_alignment): Commonize read/write case.
3265 * tree-vect-stmts.c (vect_get_store_cost): Use
3266 dr_alignment_support, not int, for the vect_supportable_dr_alignment
3268 (vect_get_load_cost): Likewise.
3270 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
3272 * tree-object-size.c (object_sizes_execute): Consolidate LHS
3273 null check and do it early.
3275 2021-10-18 Richard Biener <rguenther@suse.de>
3277 * tree-vect-stmts.c (vectorizable_store): Use the
3278 computed alignment scheme instead of querying
3281 2021-10-18 Richard Biener <rguenther@suse.de>
3283 * tree-vect-stmts.c (vectorizable_store): Do not recompute
3284 alignment scheme already determined by get_load_store_type.
3286 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3288 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
3289 Clone corresponding pass.
3291 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3293 * combine.c (recog_for_combine): For an unrecognized move/set of
3294 a constant, try force_const_mem to place it in the constant pool.
3295 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
3296 Delete global variables (that are no longer assigned a cost value).
3297 (cse_insn): Simplify logic for deciding whether to place a folded
3298 constant in the constant pool using force_const_mem.
3299 (cse_main): Remove zero initialization of constant_pool_entries_cost
3300 and constant_pool_entries_regcost.
3301 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
3302 fractionally more expensive, when optimizing for speed.
3304 2021-10-18 Martin Liska <mliska@suse.cz>
3306 PR gcov-profile/102746
3307 PR gcov-profile/102747
3308 * gcov.c (main): Return return_code.
3309 (output_gcov_file): Mark return_code when error happens.
3310 (generate_results): Likewise.
3311 (read_graph_file): Likewise.
3312 (read_count_file): Likewise.
3314 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3316 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
3317 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
3318 a popcount, allowing compile-time evaluation/simplification.
3319 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
3321 2021-10-18 Richard Biener <rguenther@suse.de>
3323 PR tree-optimization/102788
3324 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
3326 (vect_pattern_recog_1): Likewise.
3327 (vect_recog_bool_pattern): Continue matching the pattern
3328 even if we do not have a vector type for a conversion
3331 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3333 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
3334 Evalute SS_NEG and SS_ABS of a constant argument.
3336 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
3339 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
3340 * internal-fn.def: Add entry for COND_NEG internal function.
3341 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
3342 NEGATE_EXPR, COND_NEG mapping.
3343 * optabs.def: Add entry for cond_neg_optab.
3344 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
3345 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
3346 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
3348 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3350 * tree-ssa-strlen.c (class strlen_pass): Rename from
3352 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
3353 (handle_assign): Same.
3354 (adjust_last_stmt): Same.
3355 (maybe_warn_overflow): Same.
3356 (handle_builtin_strlen): Same.
3357 (handle_builtin_strchr): Same.
3358 (handle_builtin_strcpy): Same.
3359 (handle_builtin_strncat): Same.
3360 (handle_builtin_stxncpy_strncat): Same.
3361 (handle_builtin_memcpy): Same.
3362 (handle_builtin_strcat): Same.
3363 (handle_alloc_call): Same.
3364 (handle_builtin_memset): Same.
3365 (handle_builtin_memcmp): Same.
3366 (get_len_or_size): Same.
3367 (strxcmp_eqz_result): Same.
3368 (handle_builtin_string_cmp): Same.
3369 (handle_pointer_plus): Same.
3370 (count_nonzero_bytes_addr): Same.
3371 (count_nonzero_bytes): Same.
3372 (handle_store): Same.
3373 (strlen_check_and_optimize_call): Same.
3374 (handle_integral_assign): Same.
3375 (check_and_optimize_stmt): Same.
3376 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
3378 2021-10-18 Richard Biener <rguenther@suse.de>
3380 PR tree-optimization/102798
3381 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
3382 Only copy points-to info to newly generated SSA names.
3384 2021-10-18 Martin Liska <mliska@suse.cz>
3386 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
3387 * gcov.c (get_cycles_count): Likewise.
3388 * lto-compress.c (lto_compression_zlib): Likewise.
3389 (lto_uncompression_zlib): Likewise.
3390 * targhooks.c (default_pch_valid_p): Likewise.
3392 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
3394 * tree-pass.h (make_pass_thread_jumps_full): New.
3395 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
3396 (try_thread_blocks): Add resolve and speed arguments.
3397 (pass_thread_jumps::execute): Inline.
3398 (do_early_thread_jumps): New.
3399 (do_thread_jumps): New.
3400 (make_pass_thread_jumps): Move.
3401 (pass_early_thread_jumps::gate): Inline.
3402 (pass_early_thread_jumps::execute): Inline.
3403 (class pass_thread_jumps_full): New.
3405 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
3407 * configure.ac: Treat powerpc64*-*-freebsd* the same as
3408 powerpc64-*-freebsd*.
3409 * configure: Regenerate.
3411 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
3413 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
3414 set_ptr_nonull to set_ptr_nonnull in comments.
3416 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
3418 PR tree-optimization/102720
3419 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
3420 of call used and clobbered sets.
3422 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3424 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
3425 skip gomp and itm when r or nodefaultlibs is given.
3426 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
3427 Update ordering of exclusions, remove duplicate 'v' addition
3428 (collect2 will add this from the main command line).
3430 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3432 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
3433 filelist and framework options in specs instead of code.
3434 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
3435 specs that are really driver ones.
3436 (DARWIN_CC1_SPEC): Likewise.
3437 (CPP_SPEC): Likewise.
3438 (SYSROOT_SPEC): Append space.
3439 (LINK_SYSROOT_SPEC): Remove most driver link specs.
3440 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
3441 (STARTFILE_SPEC): Likewise.
3442 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
3443 (ASM_SPEC): Update driver-related specs.
3444 (ASM_FINAL_SPEC): Likewise.
3445 * config/darwin.opt: Remove now unused option aliases.
3446 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
3447 (ASM_SPEC): Update driver-related specs.
3449 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
3451 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
3452 allow standard_sse_constant_p immediate constants.
3454 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3456 * config.gcc: Add tm-dwarf2.h to tm_d-file.
3458 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
3460 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
3461 that occur in abnormal phis.
3462 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
3463 abnormal and EH edges.
3464 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
3465 (rvrp_folder::value_on_edge): Ditto.
3466 (rvrp_folder::value_of_stmt): Ditto.
3467 (hybrid_folder::value_of_expr): Ditto for ranger queries.
3468 (hybrid_folder::value_on_edge): Ditto.
3469 (hybrid_folder::value_of_stmt): Ditto.
3470 * value-query.cc (gimple_range_global): Always return a range if
3471 the type is supported.
3473 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
3475 * config/pa/pa.md: Consistently use "rG" constraint for copy
3476 instruction in move patterns.
3478 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
3480 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
3483 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
3484 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
3485 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
3486 (ELFABIVERSION_AMDGPU_HSA_V4): New.
3487 (SET_SRAM_ECC_UNSUPPORTED): New.
3488 (copy_early_debug_info): Create elf flags to match the other objects.
3489 (main): Just let the attribute flags pass through.
3491 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3493 * tree-loop-distribution.c (reduction_var_overflows_first):
3494 Pass the type of reduction_var as first argument as it is also
3495 done for the load type.
3496 (loop_distribution::transform_reduction_loop): Add missing
3497 TREE_TYPE while determining precission of reduction_var.
3499 2021-10-15 Richard Biener <rguenther@suse.de>
3501 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
3503 * toplev.c (process_options): Warn when STABS debugging is
3504 enabled but not the preferred format.
3505 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
3507 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
3509 2021-10-15 Richard Biener <rguenther@suse.de>
3512 * tree-inline.c (copy_bb): Avoid underflowing nargs.
3514 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3516 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
3517 HFmode input operand to HImode.
3518 (ix86_vectorize_vec_perm_const): Likewise.
3519 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
3521 (*avx512f_permvar_truncv8siv8hi_1_hf):
3524 2021-10-15 Richard Biener <rguenther@suse.de>
3526 PR middle-end/102682
3527 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
3528 not create a paradoxical subreg.
3530 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3532 * config/i386/i386-expand.c (ix86_expand_vector_init):
3533 For half_vector concat for HFmode, handle them like HImode.
3535 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3537 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
3538 (back_threader::back_threader): Same.
3539 (back_threader::resolve_phi): Try to solve without looking back if
3541 (back_threader::find_paths_to_names): Same.
3542 (try_thread_blocks): Pass resolve argument to back threader.
3543 (pass_early_thread_jumps::execute): Same.
3545 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3547 * doc/invoke.texi: Remove max-fsm-thread-length,
3548 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
3550 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
3552 (back_threader_registry::back_threader_registry): Same.
3553 (back_threader_profitability::profitable_path_p): Remove
3554 param_max_fsm_thread-length.
3555 (back_threader_registry::register_path): Remove
3556 m_max_allowable_paths.
3558 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3560 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
3562 (back_threader::~back_threader): Do not release m_path.
3564 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
3566 PR tree-optimization/102738
3567 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
3569 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
3571 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
3572 (omp_mark_declare_variant): Move from c-omp.c.
3573 (omp_context_name_list_prop): Update for Fortran strings.
3574 * omp-general.h (omp_check_context_selector): New prototype.
3575 (omp_mark_declare_variant): New prototype.
3577 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3579 * config/rs6000/rs6000.c (rs6000_density_test): Move early
3580 exit test further up the function.
3582 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3584 * config/arm/arm.c (arm_add_stmt_cost): Delete.
3585 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3587 2021-10-14 Martin Jambor <mjambor@suse.cz>
3589 * doc/invoke.texi (Optimize Options): Add entry for
3590 ipa-cp-recursive-freq-factor.
3592 2021-10-14 Tamar Christina <tamar.christina@arm.com>
3594 * match.pd: New rule.
3596 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
3599 * ipa-modref-tree.h (modref_access_node::update2):
3600 Also check that parm_offset is unchanged.
3601 (modref_ref_node::insert_access): Fix updating of
3604 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3606 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
3609 2021-10-14 Richard Biener <rguenther@suse.de>
3611 PR tree-optimization/102659
3612 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
3613 rewrite pointer typed undefined overflow operations.
3614 (predicate_statements): Likewise. Make sure to emit invariant
3615 conversions in the preheader.
3616 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
3617 non-empty preheaders.
3618 * tree-data-ref.c (dr_analyze_indices): Strip useless
3619 conversions to the MEM_REF base type.
3621 2021-10-14 Martin Liska <mliska@suse.cz>
3623 * common.opt: Stop using AUTODETECT_VALUE
3624 and use EnabledBy where possible.
3625 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
3626 * toplev.c (AUTODETECT_VALUE): Remove macro.
3627 (process_options): Simplify by using EnabledBy and
3628 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
3631 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
3633 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
3635 (vld1_s32_x3): Likewise.
3637 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3639 PR tree-optimization/102736
3640 PR tree-optimization/102736
3641 * gimple-range-path.cc (path_range_query::range_on_path_entry):
3642 Assert that the requested range is defined outside the path.
3643 (path_range_query::ssa_range_in_phi): Do not call
3644 range_on_path_entry for SSA names that are defined within the
3647 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
3649 * config/darwin-driver.c (darwin_find_version_from_kernel):
3650 Quote internal identifiers and avoid contractions in
3652 (darwin_default_min_version): Likewise.
3653 (darwin_driver_init): Likewise.
3655 2021-10-14 Martin Jambor <mjambor@suse.cz>
3657 * params.opt (ipa-cp-recursive-freq-factor): New.
3658 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
3659 scc_no, self_recursion_generated_level, same_scc and
3660 self_recursion_generated_p.
3661 (ipcp_lattice::add_value): Replaced parameter unlimited with
3662 same_lat_gen_level, usit it determine limit of values and store it to
3664 (ipcp_lattice<valtype>::print): Dump the new fileds.
3665 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
3666 parameter and store it to the new value.
3667 (self_recursively_generated_p): Removed.
3668 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
3669 instead of self_recursively_generated_p, store self generation level
3671 (value_topo_info<valtype>::add_val): Set scc_no.
3672 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
3673 recursively feeding values and self generated values by appropriate
3676 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3678 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
3679 redundant test for flag_vect_cost_model.
3681 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3683 * bitmap.c (debug): New overloaded function for auto_bitmaps.
3684 * bitmap.h (debug): Same.
3686 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
3688 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
3689 Adjust builtin call.
3690 (_mm512_mask3_fcmadd_pch): Likewise.
3691 (_mm512_mask_fmadd_pch): Likewise
3692 (_mm512_mask3_fmadd_pch): Likewise
3693 (_mm512_mask_fcmadd_round_pch): Likewise
3694 (_mm512_mask3_fcmadd_round_pch): Likewise
3695 (_mm512_mask_fmadd_round_pch): Likewise
3696 (_mm512_mask3_fmadd_round_pch): Likewise
3697 (_mm_mask_fcmadd_sch): Likewise
3698 (_mm_mask3_fcmadd_sch): Likewise
3699 (_mm_mask_fmadd_sch): Likewise
3700 (_mm_mask3_fmadd_sch): Likewise
3701 (_mm_mask_fcmadd_round_sch): Likewise
3702 (_mm_mask3_fcmadd_round_sch): Likewise
3703 (_mm_mask_fmadd_round_sch): Likewise
3704 (_mm_mask3_fmadd_round_sch): Likewise
3705 (_mm_fcmadd_round_sch): Likewise
3706 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
3707 Adjust builtin call.
3708 (_mm_mask3_fmadd_pch): Likewise
3709 (_mm256_mask_fmadd_pch): Likewise
3710 (_mm256_mask3_fmadd_pch): Likewise
3711 (_mm_mask_fcmadd_pch): Likewise
3712 (_mm_mask3_fcmadd_pch): Likewise
3713 (_mm256_mask_fcmadd_pch): Likewise
3714 (_mm256_mask3_fcmadd_pch): Likewise
3715 * config/i386/i386-builtin.def: Add mask3 builtin for complex
3716 fma, and adjust mask_builtin to corresponding expander.
3717 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3718 Skip eraseing embedded rounding for expanders that emits
3720 * config/i386/sse.md (complexmove): New mode_attr.
3721 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
3722 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
3723 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
3724 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
3725 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
3726 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
3727 * config/i386/subst.md (round_embedded_complex): New subst.
3729 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
3731 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
3732 (cbranchdf4): Likewise.
3733 Add missing move patterns for TARGET_SOFT_FLOAT.
3735 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
3737 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
3738 pseudo intermediate when moving a SUBREG into a hard register,
3739 by checking ix86_hardreg_mov_ok.
3740 (ix86_expand_vector_extract): Store zero-extended SImode
3741 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
3743 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
3744 complex (SUBREG) sets of (vector) hard registers before reload, by
3745 checking ix86_hardreg_mov_ok.
3747 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
3749 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
3751 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
3753 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
3755 2021-10-13 Martin Sebor <msebor@redhat.com>
3757 PR middle-end/102630
3758 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
3760 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
3762 * collect2.c (is_lto_object_file): Release simple-object
3763 resources, close files.
3765 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
3767 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
3768 when checking mve addressing modes.
3769 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
3771 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
3772 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
3773 there is a single Ux constraint.
3775 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
3777 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
3778 multiplication on 32-bit PA 1.1 and 2.0 targets.
3780 2021-10-13 Martin Liska <mliska@suse.cz>
3782 PR gcov-profile/90364
3783 * coverage.c (build_info): Emit checksum to the global variable.
3784 (build_info_type): Add new field for checksum.
3785 (coverage_obj_finish): Pass object_checksum.
3786 (coverage_init): Use 0 as checksum for .gcno files.
3787 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
3788 * gcov.c (read_graph_file): Read also checksum.
3789 * doc/invoke.texi: Document the behaviour change.
3791 2021-10-13 Richard Biener <rguenther@suse.de>
3793 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
3794 start at integer value 2.
3795 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
3796 the iterator for GSI_LAST_NEW_STMT.
3797 (gsi_insert_seq_nodes_after): Likewise.
3798 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
3799 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
3800 arguments to gsi_replace.
3802 2021-10-13 Martin Liska <mliska@suse.cz>
3805 * common.opt: Use EnabledBy instead of detection in
3806 finish_options and process_options.
3807 * opts.c (finish_options): Remove handling of
3808 x_flag_unroll_all_loops.
3809 * toplev.c (process_options): Likewise for flag_web and
3810 flag_rename_registers.
3812 2021-10-13 Richard Biener <rguenther@suse.de>
3814 PR tree-optimization/102659
3815 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
3816 (if_convertible_gimple_assign_stmt_p): Mark the loop for
3817 rewrite when stmts with undefined behavior on integer
3819 (combine_blocks): Predicate also when we need to rewrite stmts.
3820 (predicate_statements): Rewrite affected stmts to something
3821 with well-defined behavior on overflow.
3822 (tree_if_conversion): Initialize need_to_rewrite_undefined.
3824 2021-10-13 Richard Biener <rguenther@suse.de>
3827 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
3830 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
3832 * dwarf2ctf.c: Fix typo in comment.
3834 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
3837 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
3838 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
3839 avoid generating an automatic #define of vec_cpsgn. Use the
3840 correct built-in for V4SFmode that doesn't depend on VSX.
3842 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
3846 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
3847 define_insn_and_split pattern. Add alternative 1 and split it
3848 post reload to insert operand 1 into the low part of operand 0.
3849 (*sub<mode>_1_slp): Ditto.
3850 (*and<mode>_1_slp): Ditto.
3851 (*<any_or:code><mode>_1_slp): Ditto.
3852 (*ashl<mode>3_1_slp): Ditto.
3853 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
3854 (*<any_rotate:insn><mode>3_1_slp): Ditto.
3855 (*neg<mode>_1_slp): New insn_and_split pattern.
3856 (*one_cmpl<mode>_1_slp): Ditto.
3858 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
3860 * doc/install.texi: Update MinGW and mingw-64 Binaries
3863 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
3865 * doc/extend.texi (Common Variable Attributes): Fix typos in
3866 alloc_size documentation.
3868 2021-10-12 Richard Biener <rguenther@suse.de>
3870 PR tree-optimization/102696
3871 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
3872 the tree fatally failed when we reject a BIT_FIELD_REF.
3874 2021-10-12 Richard Biener <rguenther@suse.de>
3876 PR tree-optimization/102572
3877 * tree-vect-stmts.c (vect_build_gather_load_calls): When
3878 gathering the vectorized defs for the mask pass in the
3879 desired mask vector type so invariants will be handled
3882 2021-10-12 Tamar Christina <tamar.christina@arm.com>
3884 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
3885 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
3886 *fcmuo<mode>_nor_combine): New.
3888 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
3891 * config/sparc/sparc-modes.def (OI): New integer mode.
3893 2021-10-12 Jakub Jelinek <jakub@redhat.com>
3895 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
3896 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
3899 2021-10-12 Jakub Jelinek <jakub@redhat.com>
3901 * tree-vectorizer.h (loop_cost_model): New function.
3902 (unlimited_cost_model): Use it.
3903 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
3904 call instead of flag_vect_cost_model.
3905 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
3906 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
3907 of flag_simd_cost_model.
3909 2021-10-12 liuhongt <hongtao.liu@intel.com>
3912 * config/i386/i386-expand.c (emit_reduc_half): Handle
3914 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
3915 (reduc_plus_scal_v4qi): Ditto.
3917 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3919 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
3920 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
3921 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
3923 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3925 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
3926 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
3927 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
3928 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
3929 _mm_cvtepu32_epi64): New.
3931 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3933 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
3934 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
3936 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3938 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
3939 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
3940 _mm_max_epi32, _mm_max_epu32): New.
3942 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
3944 * ipa-modref-tree.h (struct modref_access_node): Revert
3946 (struct modref_ref_node): Likewise.
3948 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
3950 * ipa-modref-tree.h (modref_tree::global_access_p): New member
3953 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
3954 ignore_stores_eaf_flags): Move to ipa-modref.h
3955 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
3956 (modref_summary::global_memory_read_p): New member function.
3957 (modref_summary::global_memory_written_p): New member function.
3958 * ipa-modref.h (modref_summary::global_memory_read_p,
3959 modref_summary::global_memory_written_p): Declare.
3960 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
3961 ignore_stores_eaf_flags): move here.
3962 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
3964 (handle_rhs_call): Rewrite.
3965 (handle_call_arg): New function.
3966 (determine_global_memory_access): New function.
3967 (handle_const_call): Remove
3968 (handle_pure_call): Remove
3969 (find_func_aliases_for_call): Update use of handle_rhs_call.
3970 (compute_points_to_sets): Handle global memory acccesses
3973 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
3975 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
3976 documentation, mention UBSAN_OPTIONS, similar to what is done
3977 for AddressSanitizer.
3979 2021-10-11 Richard Biener <rguenther@suse.de>
3981 PR middle-end/102683
3982 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
3983 availability before building an integer type for storage
3986 2021-10-11 Richard Biener <rguenther@suse.de>
3988 PR middle-end/101480
3989 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
3992 2021-10-11 Martin Liska <mliska@suse.cz>
3994 * common.opt: Remove Init(2) for some options.
3995 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
3996 use rather OPTION_SET_P.
3998 2021-10-11 Martin Liska <mliska@suse.cz>
4000 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
4001 * flag-types.h (enum ira_region): Likewise.
4002 * toplev.c (process_options): Use OPTION_SET_P instead of
4003 IRA_REGION_AUTODETECT.
4005 2021-10-11 Jakub Jelinek <jakub@redhat.com>
4007 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
4008 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
4010 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4012 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
4013 * config/s390/s390.c (s390_rawmemchr): New function.
4014 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
4015 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
4016 the pattern vfees<mode> from vx-builtins.md.
4017 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
4019 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4021 * builtins.c (get_memory_rtx): Change to external linkage.
4022 * builtins.h (get_memory_rtx): Add function prototype.
4023 * doc/md.texi (rawmemchr<mode>): Document.
4024 * internal-fn.c (expand_RAWMEMCHR): Define.
4025 * internal-fn.def (RAWMEMCHR): Add.
4026 * optabs.def (rawmemchr_optab): Add.
4027 * tree-loop-distribution.c (find_single_drs): Change return code
4028 behaviour by also returning true if no single store was found
4030 (loop_distribution::classify_partition): Respect the new return
4031 code behaviour of function find_single_drs.
4032 (loop_distribution::execute): Call new function
4033 transform_reduction_loop in order to replace rawmemchr or strlen
4034 like loops by calls into builtins.
4035 (generate_reduction_builtin_1): New function.
4036 (generate_rawmemchr_builtin): New function.
4037 (generate_strlen_builtin_1): New function.
4038 (generate_strlen_builtin): New function.
4039 (generate_strlen_builtin_using_rawmemchr): New function.
4040 (reduction_var_overflows_first): New function.
4041 (determine_reduction_stmt_1): New function.
4042 (determine_reduction_stmt): New function.
4043 (loop_distribution::transform_reduction_loop): New function.
4045 2021-10-11 Martin Liska <mliska@suse.cz>
4047 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
4048 and remove legacy hashing code.
4050 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
4053 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
4054 CONST_INT for BEGIN and END, and use gcc_assert rather than
4057 2021-10-10 Jakub Jelinek <jakub@redhat.com>
4060 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
4061 use MO_VAL_SET if loc is not sp.
4063 2021-10-10 Andrew Pinski <apinski@marvell.com>
4065 PR tree-optimization/102622
4066 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
4067 Swap the order of a?0:pow2cst and a?0:-1 transformations.
4069 2021-10-09 liuhongt <hongtao.liu@intel.com>
4072 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
4074 (ix86_use_mask_cmp_p): Ditto.
4075 (ix86_expand_sse_movcc): Ditto.
4076 * config/i386/i386.md (setcc_hf_mask): New define_insn.
4077 (movhf_mask): Ditto.
4078 (UNSPEC_MOVCC_MASK): New unspec.
4079 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
4081 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
4083 PR rtl-optimization/102627
4084 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
4086 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
4088 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
4091 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
4093 * value-range.cc (irange::debug): New.
4094 * value-range.h (irange::debug): New.
4096 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
4098 PR tree-optimization/102385
4099 * predict.h (change_edge_frequency): Declare.
4100 * predict.c (change_edge_frequency): New function.
4101 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
4103 (tree_unroll_loop): Likewise.
4104 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
4105 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
4107 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
4108 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
4109 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
4110 to retrieve the exit edges. Make all the old profile update code
4111 conditional on !single_loop_p -- the case it was written for --
4112 and use a different approach for the single-loop case.
4114 2021-10-08 Martin Liska <mliska@suse.cz>
4116 * config/alpha/alpha.c (alpha_option_override): Use new macro
4118 * config/arc/arc.c (arc_override_options): Likewise.
4119 * config/arm/arm.c (arm_option_override): Likewise.
4120 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
4121 * config/c6x/c6x.c (c6x_option_override): Likewise.
4122 * config/csky/csky.c: Likewise.
4123 * config/darwin.c (darwin_override_options): Likewise.
4124 * config/frv/frv.c (frv_option_override): Likewise.
4125 * config/i386/djgpp.h: Likewise.
4126 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
4127 (ix86_max_noce_ifcvt_seq_cost): Likewise.
4128 * config/ia64/ia64.c (ia64_option_override): Likewise.
4129 (ia64_override_options_after_change): Likewise.
4130 * config/m32c/m32c.c (m32c_option_override): Likewise.
4131 * config/m32r/m32r.c (m32r_init): Likewise.
4132 * config/m68k/m68k.c (m68k_option_override): Likewise.
4133 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
4134 * config/mips/mips.c (mips_option_override): Likewise.
4135 * config/nios2/nios2.c (nios2_option_override): Likewise.
4136 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
4137 * config/pa/pa.c (pa_option_override): Likewise.
4138 * config/riscv/riscv.c (riscv_option_override): Likewise.
4139 * config/rs6000/aix71.h: Likewise.
4140 * config/rs6000/aix72.h: Likewise.
4141 * config/rs6000/aix73.h: Likewise.
4142 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
4143 (rs6000_override_options_after_change): Likewise.
4144 (rs6000_linux64_override_options): Likewise.
4145 (glibc_supports_ieee_128bit): Likewise.
4146 (rs6000_option_override_internal): Likewise.
4147 (rs6000_file_start): Likewise.
4148 (rs6000_darwin_file_start): Likewise.
4149 * config/rs6000/rtems.h: Likewise.
4150 * config/rs6000/sysv4.h: Likewise.
4151 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
4152 * config/s390/s390.c (s390_option_override): Likewise.
4153 * config/sh/linux.h: Likewise.
4154 * config/sh/netbsd-elf.h (while): Likewise.
4155 * config/sh/sh.c (sh_option_override): Likewise.
4156 * config/sol2.c (solaris_override_options): Likewise.
4157 * config/sparc/sparc.c (sparc_option_override): Likewise.
4158 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
4159 * config/visium/visium.c (visium_option_override): Likewise.
4160 * config/vxworks.c (vxworks_override_options): Likewise.
4161 * lto-opts.c (lto_write_options): Likewise.
4162 * omp-expand.c (expand_omp_simd): Likewise.
4163 * omp-general.c (omp_max_vf): Likewise.
4164 * omp-offload.c (oacc_xform_loop): Likewise.
4165 * opts.h (OPTION_SET_P): Likewise.
4166 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
4167 * toplev.c (process_options): Likewise.
4168 * tree-predcom.c: Likewise.
4169 * tree-sra.c (analyze_all_variable_accesses): Likewise.
4171 2021-10-08 liuhongt <hongtao.liu@intel.com>
4174 * config/i386/i386.c (ix86_optab_supported_p):
4175 Return true for HFmode.
4176 * match.pd: Simplify (_Float16) ceil ((double) x) to
4177 __builtin_ceilf16 (a) when a is _Float16 type and
4178 direct_internal_fn_supported_p.
4180 2021-10-08 liuhongt <hongtao.liu@intel.com>
4183 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
4184 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
4185 (reduc_<code>_scal_v4hi): New.
4187 2021-10-08 liuhongt <hongtao.liu@intel.com>
4189 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
4190 * doc/invoke.texi (Options That Control Optimization): Update
4192 * opts.c (default_options_table): Enable auto-vectorization at
4193 O2 with very-cheap cost model.
4194 (finish_options): Use cheap cost model for
4195 explicit -ftree{,-loop}-vectorize.
4197 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4199 * ctfc.c (ctfc_delete_container): Free hash table contents.
4201 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4203 * toplev.c (process_options): Do not warn for GNU GIMPLE.
4205 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
4207 * tree-object-size.c (addr_object_size,
4208 compute_builtin_object_size): Drop PDECL and POFF arguments.
4209 (addr_object_size): Adjust calls.
4210 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
4213 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
4215 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
4216 representing signed and unsigned high-part multiplication resp.
4217 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
4218 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
4219 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
4220 SS_DIV, US_DIV]: Similar simplifications for saturating
4222 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
4223 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
4224 Implement compile-time evaluation for constant operands.
4225 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
4227 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
4228 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
4229 Mention the new smul_highpart and umul_highpart RTX codes.
4230 * doc/invoke.texi: Silence @xref "compilation" warnings.
4232 2021-10-07 Martin Jambor <mjambor@suse.cz>
4235 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
4236 case when the source reference description corresponds to a
4237 referance taken in a function src->caller is inlined to.
4239 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
4242 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
4244 (modref_access_node::try_merge_with): Add sanity check that there
4245 are no redundant entries in the list.
4247 2021-10-07 Richard Biener <rguenther@suse.de>
4249 PR tree-optimization/102608
4250 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
4253 2021-10-07 Martin Liska <mliska@suse.cz>
4255 * toplev.c (toplev::main): Make
4256 save_opt_decoded_options a pointer type
4257 * toplev.h: Likewise.
4259 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4261 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
4262 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
4263 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
4265 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4267 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
4268 Adjust for new -msram-ecc=any behaviour.
4269 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
4270 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
4271 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
4272 (EF_AMDGPU_XNACK_V3): ... this.
4273 (EF_AMDGPU_SRAM_ECC): Rename to ...
4274 (EF_AMDGPU_SRAM_ECC_V3): ... this.
4275 (EF_AMDGPU_FEATURE_XNACK_V4): New.
4276 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
4277 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
4278 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
4279 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
4280 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
4281 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
4282 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
4283 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
4284 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
4285 (SET_XNACK_ON): New.
4286 (SET_XNACK_OFF): New.
4288 (SET_SRAM_ECC_ON): New.
4289 (SET_SRAM_ECC_ANY): New.
4290 (SET_SRAM_ECC_OFF): New.
4291 (TEST_SRAM_ECC_ANY): New.
4292 (TEST_SRAM_ECC_ON): New.
4293 (main): Implement HSACOv4 and -msram-ecc=any.
4295 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4297 * config.in: Regenerate.
4298 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
4302 (A_FIJI): Rename to ...
4304 (A_900): Rename to ...
4306 (A_906): Rename to ...
4308 (A_908): Rename to ...
4310 (SRAMOPT): New macro.
4311 (ASM_SPEC): Adjust xnack option usage.
4312 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
4313 * configure: Regenerate.
4314 * configure.ac: Detect LLVM assembler dialect.
4316 2021-10-07 Richard Biener <rguenther@suse.de>
4318 * tree-pretty-print.c (dump_generic_node): Do not elide
4319 printing '&' when dumping with -gimple.
4321 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4323 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
4325 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
4326 (adjust_imagpart_expr): Ditto.
4327 * value-range.cc (irange::irange_intersect): Call new routine if
4328 RHS is a single pair.
4329 (irange::intersect): New wide_int version.
4330 * value-range.h (class irange): New prototype.
4332 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4334 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
4335 Add parameter to limit size when recognizing switches.
4336 (gimple_outgoing_range::edge_range_p): Check size limit.
4337 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
4338 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
4339 that exceed the size limit.
4340 (gori_compute::gori_compute): Add initializer.
4341 * params.opt (evrp-switch-limit): New.
4342 * doc/invoke.texi: Update docs.
4344 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4346 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
4347 TYPE_MAX_VALUE instead of creating new trees when possible.
4349 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4351 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
4352 zero and non-zero more efficently.
4354 2021-10-06 Richard Biener <rguenther@suse.de>
4357 * dumpfile.h (TDF_GIMPLE_VAL): New.
4358 (dump_flag): Re-order and adjust TDF_* flags. Make
4359 the enum uint32_t. Use std::underlying_type in the
4361 (optgroup_flag): Likewise for the operator overloads.
4362 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
4363 in _Literal if TDF_GIMPLE_VAL.
4364 * gimple-pretty-print.c (dump_gimple_assign): Add
4365 TDF_GIMPLE_VAL to flags when dumping operands where only
4366 is_gimple_val are allowed.
4367 (dump_gimple_cond): Likewise.
4369 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
4371 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
4374 2021-10-05 qing zhao <qing.zhao@oracle.com>
4376 PR middle-end/102359
4377 * gimplify.c (gimplify_decl_expr): Not add initialization for an
4378 auto variable when it has been initialized by frontend.
4380 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
4382 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
4385 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4387 * common/config/avr/avr-common.c (avr_handle_option): Mark
4388 argument as ATTRIBUTE_UNUSED.
4390 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4392 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
4393 Undefine before redefinition.
4395 2021-10-05 Richard Biener <rguenther@suse.de>
4397 * toplev.c (no_backend): Remove global var.
4398 (process_options): Pass in no_backend, move post_options
4399 langhook call to toplev::main.
4400 (do_compile): Pass in no_backend, move process_options call
4402 (toplev::run_self_tests): Check no_backend at the caller.
4403 (toplev::main): Call post_options and process_options
4404 split out from do_compile, do self-tests only if
4405 no_backend is initialized.
4407 2021-10-05 Richard Biener <rguenther@suse.de>
4409 * tree-cfg.c (dump_function_to_file): Dump the UID of the
4410 function as part of the name when requested.
4411 * tree-pretty-print.c (dump_function_name): Dump the UID when
4412 requested and the langhook produced the actual name.
4414 2021-10-05 Richard Biener <rguenther@suse.de>
4416 PR middle-end/102587
4417 PR middle-end/102285
4418 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
4419 zero-initialization as last resort, use the constant
4420 size as given by the DEFERRED_INIT argument to build
4423 2021-10-04 Marek Polacek <polacek@redhat.com>
4426 * doc/invoke.texi: Document -Warray-compare.
4428 2021-10-04 Richard Biener <rguenther@suse.de>
4430 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
4431 variables are not to be initialized.
4433 2021-10-04 Richard Biener <rguenther@suse.de>
4435 * expr.h (non_mem_decl_p): Declare.
4436 (mem_ref_refers_to_non_mem_p): Likewise.
4437 * expr.c (non_mem_decl_p): Export.
4438 (mem_ref_refers_to_non_mem_p): Likewise.
4439 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
4440 but check the base with mem_ref_refers_to_non_mem_p
4443 2021-10-04 Richard Biener <rguenther@suse.de>
4445 PR tree-optimization/102570
4446 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
4447 we are using clique for the internal function code.
4448 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
4449 internal function code.
4450 (print_vn_reference_ops): Print the internal function code.
4451 (vn_reference_op_compute_hash): Hash it.
4452 (copy_reference_ops_from_call): Record it.
4453 (visit_stmt): Remove the restriction around internal function
4455 (fully_constant_vn_reference_p): Use fold_const_call and handle
4457 (vn_reference_eq): Compare call return types.
4458 * tree-ssa-pre.c (create_expression_by_pieces): Handle
4459 generating calls to internal functions.
4460 (compute_avail): Remove the restriction around internal function
4463 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
4465 PR tree-optimization/102560
4466 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
4467 marker for invalid_range.
4469 2021-10-04 Richard Biener <rguenther@suse.de>
4471 PR middle-end/102587
4472 * internal-fn.c (expand_DEFERRED_INIT): Guard register
4473 initialization path an avoid initializing VLA registers
4476 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
4478 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
4480 2021-10-03 Martin Liska <mliska@suse.cz>
4482 * toplev.c (toplev::main): Check opt_index if it is a part
4485 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
4487 PR tree-optimization/102563
4488 * range-op.cc (operator_lshift::op1_range): Do not clobber
4491 2021-10-02 Martin Liska <mliska@suse.cz>
4493 * toplev.c (toplev::main): save_decoded_options[0] is program
4494 name and so it should be skipped.
4496 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4498 PR tree-optimization/102546
4499 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
4500 X << Y is non-zero implies X is also non-zero.
4502 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4504 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4506 * config/aarch64/aarch64-tune.md: Regenerate.
4507 * doc/invoke.texi: Update docs.
4509 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4511 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4513 * config/aarch64/aarch64-tune.md: Regenerate.
4514 * doc/invoke.texi: Update docs.
4516 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4518 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4520 * config/aarch64/aarch64-tune.md: Regenerate.
4521 * doc/invoke.texi: Update docs.
4523 2021-10-01 Martin Sebor <msebor@redhat.com>
4526 * doc/invoke.texi (-Waddress): Update.
4527 * gengtype.c (write_types): Avoid -Waddress.
4528 * poly-int.h (POLY_SET_COEFF): Avoid using null.
4530 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
4533 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
4536 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4538 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
4540 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4542 * gimple-range-path.cc (path_range_query::compute_ranges): Use
4544 * gimple-range-path.h (class path_range_query): Remove shadowed
4546 (path_range_query::get_path_oracle): New.
4548 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4549 Richard Biener <rguenther@suse.de>
4552 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
4553 INT_MIN / -1 division detection from here ...
4554 (-fsanitize=signed-integer-overflow): ... and add it here.
4556 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4558 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
4560 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
4561 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
4562 (AARCH64_ISA_V9): New ISA flag.
4563 * doc/invoke.texi: Update docs.
4565 2021-10-01 Martin Liska <mliska@suse.cz>
4567 * toplev.c (toplev::main): Save decoded optimization options.
4568 * toplev.h (save_opt_decoded_options): New.
4569 * doc/extend.texi: Be more clear about optimize and target
4572 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4574 * explow.c: Include langhooks.h.
4575 (set_stack_check_libfunc): Build a proper function type.
4577 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4580 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
4581 not return true for external weak function symbols in medium model.
4583 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4585 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
4586 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
4587 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
4588 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
4589 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
4590 OMP_CLAUSE_SCHEDULE_STATIC.
4592 2021-10-01 Richard Biener <rguenther@suse.de>
4594 PR middle-end/102518
4595 * tree-inline.c (setup_one_parameter): Avoid substituting
4596 an invariant into contexts where a GIMPLE register is not valid.
4598 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4600 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
4601 * config/arm/arm-tables.opt: Regenerate.
4602 * config/arm/arm-tune.md: Regenerate.
4603 * doc/invoke.texi: Update docs.
4605 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
4608 * config/i386/i386.md
4609 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
4612 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
4614 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
4615 omp_{c,re}alloc, fix omp_alloc/omp_free.
4617 2021-09-30 Martin Liska <mliska@suse.cz>
4619 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
4620 asm_out_file and stream directly to MYFILE.
4622 2021-09-30 Richard Biener <rguenther@suse.de>
4624 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4625 Restore and fix condition under which we apply npeel to
4626 the DRs misalignment value.
4628 2021-09-30 Richard Biener <rguenther@suse.de>
4630 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4631 Fix npeel check for variable amount of peeling.
4633 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4635 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
4637 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4639 * gimple-range.cc (gimple_ranger::debug): New.
4640 * gimple-range.h (class gimple_ranger): Add debug.
4642 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4644 PR middle-end/102519
4645 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
4647 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
4650 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
4651 (btf_finalize): Empty the hash_map btf_var_ids.
4653 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
4655 * tree-vrp.c (thread_through_all_blocks): Return bool.
4656 (execute_vrp_threader): Return TODO_* flags.
4657 (pass_data_vrp_threader): Set todo_flags_finish to 0.
4659 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
4661 * timevar.def (TV_TREE_VRP_THREADER): New.
4662 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
4664 2021-09-29 David Faust <david.faust@oracle.com>
4666 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
4668 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
4670 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
4671 new type qualifier enum.
4672 (TYPES_TERNOP_SSSU): Likewise.
4673 (TYPES_TERNOP_PPPU): Likewise.
4674 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
4675 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
4677 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
4678 builtin and remove casts.
4679 (vqtbl1_s8): Likewise.
4680 (vqtbl1q_p8): Likewise.
4681 (vqtbl1q_s8): Likewise.
4682 (vqtbx1_s8): Likewise.
4683 (vqtbx1_p8): Likewise.
4684 (vqtbx1q_s8): Likewise.
4685 (vqtbx1q_p8): Likewise.
4686 (vtbl1_p8): Likewise.
4687 (vtbl2_p8): Likewise.
4688 (vtbx2_p8): Likewise.
4690 2021-09-29 Richard Biener <rguenther@suse.de>
4692 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
4694 (vect_update_misalignment_for_peel): Use it to update
4695 misaligned to the value necessary for an aligned access.
4696 (vect_get_peeling_costs_all_drs): Likewise.
4697 (vect_enhance_data_refs_alignment): Likewise.
4699 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4701 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
4702 emitted operations and adjust heuristic for code size.
4704 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4706 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
4707 emitted operations and adjust heuristic for code size.
4709 2021-09-29 Jakub Jelinek <jakub@redhat.com>
4711 PR middle-end/102504
4712 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
4713 in OMP_SCOPE clauses, not just on worksharing construct clauses.
4715 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
4717 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
4718 (<u>mulditi3): Call <su>muldi3_highpart.
4719 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
4720 (<u>mulsidi3): Call <su>mulsi3_highpart.
4721 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
4723 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
4725 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
4727 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
4729 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
4730 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
4733 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4735 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
4736 (AARCH64_FL_V8_7): Likewise.
4737 (AARCH64_FL_FOR_ARCH8_7): Likewise.
4738 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
4739 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
4740 * doc/invoke.texi: Document the above.
4742 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4744 * dbgcnt.c (dbg_cnt_counter): New.
4745 * dbgcnt.h (dbg_cnt_counter): New.
4746 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
4747 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
4748 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
4749 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
4752 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4754 * cfgcleanup.c (pass_jump::execute): Check
4755 flag_expensive_optimizations.
4756 (pass_jump_after_combine::gate): Same.
4757 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
4758 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
4759 * tree-ssa-threadupdate.c
4760 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
4763 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
4765 * tree-ssa-reassoc.c (biased_names): New global.
4766 (propagate_bias_p): New function.
4767 (loop_carried_phi): Remove.
4768 (propagate_rank): Propagate bias along single uses.
4769 (get_rank): Update biased_names when needed.
4771 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
4773 * passes.def (pass_reassoc): Rename parameter to early_p.
4774 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
4776 (phi_rank): Don't bias loop-carried phi ranks
4777 before vectorization pass.
4778 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
4779 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
4781 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
4783 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
4785 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
4787 2021-09-28 Jakub Jelinek <jakub@redhat.com>
4790 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
4791 special 80387 instruction XFmode constants if flag_rounding_math.
4793 2021-09-28 Richard Biener <rguenther@suse.de>
4795 PR tree-optimization/100112
4796 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
4797 referece into the hashtable twice in case last_vuse is
4798 different from the original vuse on the stmt.
4800 2021-09-28 Jakub Jelinek <jakub@redhat.com>
4802 PR middle-end/102492
4803 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
4804 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
4805 clauses on SIMD constructs.
4807 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4809 PR tree-optimization/102511
4810 * gimple-range-path.cc (path_range_query::range_on_path_entry):
4811 Return VARYING when nothing found.
4813 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
4816 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
4818 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
4819 (VALID_MMX_REG_MODE): Add V4HFmode.
4820 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
4821 vector mode condition.
4822 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
4824 (function_arg_32): Add V4HFmode.
4825 (function_arg_advance_32): Likewise.
4826 * config/i386/i386.md (mode): Add V4HF/V2HF.
4827 (MODE_SIZE): Likewise.
4828 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
4829 (V_32): Add V2HF mode.
4830 (VHF_32_64): New mode iterator.
4831 (*mov<mode>_internal): Adjust sse alternatives to support
4833 (*mov<mode>_internal): Adjust sse alternatives to support
4835 (<insn><mode>3): New define_insn for add/sub/mul/div.
4837 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4839 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
4841 (pass_early_thread_jumps::gate): Same.
4842 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
4843 Return if !flag_thread_jumps.
4844 * tree-ssa-threadupdate.c
4845 (jt_path_registry::register_jump_thread): Assert that
4846 flag_thread_jumps is true.
4848 2021-09-28 liuhongt <hongtao.liu@intel.com>
4851 (simplify_context::simplify_binary_operation_1): Relax
4852 condition of simplifying (vec_concat:M (vec_select op0
4853 index0)(vec_select op1 index1)) to allow different modes
4854 between op0 and M, but have same inner mode.
4856 2021-09-28 liuhongt <hongtao.liu@intel.com>
4858 * config/i386/i386-expand.c (emit_reduc_half): Handle
4859 V8HF/V16HF/V32HFmode.
4860 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
4861 (REDUC_SSE_SMINMAX_MODE): Ditto.
4862 (REDUC_PLUS_MODE): Add V16HF and V32HF.
4863 (REDUC_SMINMAX_MODE): Ditto.
4865 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4867 * gimple-range-path.cc
4868 (path_range_query::precompute_ranges_in_block): Rename to...
4869 (path_range_query::compute_ranges_in_block): ...this.
4870 (path_range_query::precompute_ranges): Rename to...
4871 (path_range_query::compute_ranges): ...this.
4872 (path_range_query::precompute_relations): Rename to...
4873 (path_range_query::compute_relations): ...this.
4874 (path_range_query::precompute_phi_relations): Rename to...
4875 (path_range_query::compute_phi_relations): ...this.
4876 * gimple-range-path.h: Rename precompute* to compute*.
4877 * tree-ssa-threadbackward.c
4878 (back_threader::find_taken_edge_switch): Same.
4879 (back_threader::find_taken_edge_cond): Same.
4880 * tree-ssa-threadedge.c
4881 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
4882 (hybrid_jt_state::register_equivs_stmt): Inline...
4883 * tree-ssa-threadedge.h: ...here.
4885 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4887 * tree-vrp.c (lhs_of_dominating_assert): Remove.
4888 (class vrp_jt_state): Remove.
4889 (class vrp_jt_simplifier): Remove.
4890 (vrp_jt_simplifier::simplify): Remove.
4891 (class vrp_jump_threader): Remove.
4892 (vrp_jump_threader::vrp_jump_threader): Remove.
4893 (vrp_jump_threader::~vrp_jump_threader): Remove.
4894 (vrp_jump_threader::before_dom_children): Remove.
4895 (vrp_jump_threader::after_dom_children): Remove.
4897 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4899 * passes.def (pass_vrp_threader): New.
4900 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
4901 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
4902 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
4903 (hybrid_jt_simplifier::simplify): New.
4904 (hybrid_jt_simplifier::compute_ranges_from_state): New.
4905 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
4906 (class hybrid_jt_simplifier): New.
4907 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
4909 (class hybrid_threader): New.
4910 (hybrid_threader::hybrid_threader): New.
4911 (hybrid_threader::~hybrid_threader): New.
4912 (hybrid_threader::before_dom_children): New.
4913 (hybrid_threader::after_dom_children): New.
4914 (execute_vrp_threader): New.
4915 (class pass_vrp_threader): New.
4916 (make_pass_vrp_threader): New.
4918 2021-09-27 Martin Liska <mliska@suse.cz>
4920 * output.h (enum section_flag): New.
4921 (SECTION_FORGET): Remove.
4922 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
4923 (SECTION_STYLE_MASK): Define it based on other enum
4925 * varasm.c (switch_to_section): Remove unused handling of
4928 2021-09-27 Martin Liska <mliska@suse.cz>
4930 * common.opt: Add new variable flag_default_complex_method.
4931 * opts.c (finish_options): Handle flags related to
4932 x_flag_complex_method.
4933 * toplev.c (process_options): Remove option handling related
4934 to flag_complex_method.
4936 2021-09-27 Richard Biener <rguenther@suse.de>
4938 PR middle-end/102450
4939 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
4940 type_for_size, instead use int_mode_for_size.
4942 2021-09-27 Andrew Pinski <apinski@marvell.com>
4945 * gimplify.c (gimplify_save_expr): Return early
4946 if the type of val is error_mark_node.
4948 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4950 * tree-ssanames.c (ssa_name_has_boolean_range): Use
4953 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4955 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
4956 vrp_visit_cond_stmt.
4957 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
4958 (cprop_into_stmt): Same.
4959 (dom_opt_dom_walker::optimize_stmt): Same.
4961 2021-09-27 Richard Biener <rguenther@suse.de>
4963 PR tree-optimization/97351
4964 PR tree-optimization/97352
4965 PR tree-optimization/82426
4966 * tree-vectorizer.h (dr_misalignment): Add vector type
4968 (aligned_access_p): Likewise.
4969 (known_alignment_for_access_p): Likewise.
4970 (vect_supportable_dr_alignment): Likewise.
4971 (vect_known_alignment_in_bytes): Likewise. Refactor.
4972 (DR_MISALIGNMENT): Remove.
4973 (vect_update_shared_vectype): Likewise.
4974 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
4975 a vector type with larger alignment requirement and apply
4976 the negative step adjustment here.
4977 (vect_calculate_target_alignment): Remove.
4978 (vect_compute_data_ref_alignment): Get explicit vector type
4979 argument, do not apply a negative step alignment adjustment
4981 (vect_slp_analyze_node_alignment): Re-analyze alignment
4982 when we re-visit the DR with a bigger desired alignment but
4983 keep more precise results from smaller alignments.
4984 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
4985 (vect_slp_analyze_node_operations_1): Do not update the
4986 shared vector type on stmts.
4987 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
4988 vector type of an SLP node to the representative stmt-info.
4989 (vect_transform_stmt): Likewise.
4991 2021-09-27 liuhongt <hongtao.liu@intel.com>
4994 2021-09-09 liuhongt <hongtao.liu@intel.com>
4997 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
4998 (reduc_plus_scal_v4sf): .. this, New define_expand.
4999 (reduc_plus_scal_v2df): .. and this, New define_expand.
5001 2021-09-26 liuhongt <hongtao.liu@intel.com>
5003 * doc/extend.texi (Half-Precision): Remove storage only
5004 description for _Float16 w/o avx512fp16.
5006 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
5008 * config/pru/constraints.md (Rrio): New constraint.
5009 * config/pru/predicates.md (regio_operand): New predicate.
5010 * config/pru/pru-pragma.c (pru_register_pragmas): Register
5011 the __regio_symbol address space.
5012 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
5013 * config/pru/pru.c (pru_symref2ioregno): New helper function.
5014 (pru_legitimate_address_p): Remove.
5015 (pru_addr_space_legitimate_address_p): Use the address space
5017 (pru_nongeneric_pointer_addrspace): New helper function.
5018 (pru_insert_attributes): New function to validate __regio_symbol
5020 (TARGET_INSERT_ATTRIBUTES): New macro.
5021 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
5022 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
5023 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
5024 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
5026 (*regio_nozext_writesi): New pattern to write to I/O registers.
5027 (*regio_zext_write_r30<EQS0:mode>): Ditto.
5028 * doc/extend.texi: Document the new PRU Named Address Space.
5030 2021-09-24 Patrick Palka <ppalka@redhat.com>
5034 * real.c (encode_ieee_double): Avoid unwanted sign extension.
5035 (encode_ieee_quad): Likewise.
5037 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
5039 PR rtl-optimization/102147
5040 * ira-build.c (ira_conflict_vector_profitable_p): Make
5041 profitability calculation independent of host compiler pointer and
5044 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
5046 * gimple-range-path.cc (path_range_query::path_range_query):
5047 Move debugging header...
5048 (path_range_query::precompute_ranges): ...here.
5049 (path_range_query::internal_range_of_expr): Do not call
5050 range_on_path_entry if NAME is defined in the current block.
5052 2021-09-24 Richard Biener <rguenther@suse.de>
5054 * cfghooks.c (verify_flow_info): Verify unallocated BB and
5055 edge flags are not set.
5057 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
5059 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
5061 (jt_path_registry::register_jump_thread): Call
5062 cancel_invalid_paths.
5063 * tree-ssa-threadupdate.h (class jt_path_registry): Add
5064 cancel_invalid_paths.
5066 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
5068 PR tree-optimization/102400
5069 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
5070 result_vdef to zero value.
5072 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
5074 PR tree-optimization/102451
5075 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
5078 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
5080 * config/i386/sse.md (cond_<insn><mode>): Extend to support
5082 (cond_mul<mode>): Likewise.
5083 (cond_div<mode>): Likewise.
5084 (cond_<code><mode>): Likewise.
5085 (cond_fma<mode>): Likewise.
5086 (cond_fms<mode>): Likewise.
5087 (cond_fnma<mode>): Likewise.
5088 (cond_fnms<mode>): Likewise.
5090 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
5092 PR tree-optimization/102463
5093 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
5094 there is no range-ops handler, don't look for a relation.
5096 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
5098 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
5099 non-executable_edge flag as parameter.
5100 * gimple-range-cache.h (ranger_cache): Adjust prototype.
5101 * gimple-range-gori.cc (gori_compute::gori_compute): Take
5102 non-executable_edge flag as parameter.
5103 (gori_compute::outgoing_edge_range_p): Check new flag.
5104 * gimple-range-gori.h (gori_compute): Adjust prototype.
5105 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
5106 (gimple_ranger::range_on_edge): Check new flag.
5107 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
5108 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
5109 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
5110 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
5111 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
5112 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
5113 (simplify_using_ranges::simplify_switch_using_ranges): Clear
5114 EDGE_EXECUTABLE like it originally did.
5115 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
5116 NON_EXECUTABLE flags.
5117 (simplify_using_ranges::simplify_using_ranges): Adjust.
5118 * vr-values.h (class simplify_using_ranges): Adjust.
5119 (simplify_using_ranges::set_range_query): Add non-executable flag param.
5121 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
5124 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
5125 zero-width bit fields and return indicator.
5126 (rs6000_discover_homogeneous_aggregate): Diagnose when the
5127 presence of a zero-width bit field changes parameter passing in
5130 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5132 * gimple-range-fold.cc (fold_using_range::range_of_phi):
5133 Remove dominator check.
5135 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5137 * gimple-range-path.cc (path_range_query::precompute_relations):
5138 Hoist edge calculations before using EDGE_SUCC.
5140 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
5142 * configure.ac: Fix --with-multilib-list description.
5143 * configure: Regenerate.
5145 2021-09-23 Richard Biener <rguenther@suse.de>
5147 PR tree-optimization/102448
5148 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
5149 Clear alignment info copied from DR_PTR_INFO.
5151 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5153 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
5155 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
5156 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
5157 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
5158 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
5159 (vcond<mode><mode>): New expander.
5160 (vcond<mode><sseintvecmodelower>): Likewise.
5161 (vcond<sseintvecmodelower><mode>): Likewise.
5162 (vcondu<mode><sseintvecmodelower>): Likewise.
5164 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5166 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
5168 (extendv4hf<mode>2): Likewise.
5169 (extendv2hfv2df2): Likewise.
5170 (trunc<mode><ssePHmodelower>2): Likewise.
5171 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
5172 (trunc<mode>v4hf2): ... this, and drop constraints.
5173 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
5174 (truncv2dfv2hf2): ... this, and likewise.
5176 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5178 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
5180 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
5182 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
5183 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
5184 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
5186 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5188 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
5189 (fixuns_trunchfhi2): Likewise.
5190 (*fixuns_trunchfsi2zext): New define_insn.
5191 * config/i386/sse.md (ssePHmodelower): New mode_attr.
5192 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5193 New expander for same element vector fix_truncate.
5194 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5195 Likewise for V4HF to V4SI/V4DI fix_truncate.
5196 (fix<fixunssuffix>_truncv2hfv2di2):
5197 Likeise for V2HF to V2DI fix_truncate.
5199 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5201 * config/i386/i386.md (<code>hf3): New expander.
5203 2021-09-23 liuhongt <hongtao.liu@intel.com>
5205 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
5206 (VFH_SF_AVX512VL): Extend to handle HFmode.
5207 (VF_SF_AVX512VL): Deleted.
5209 2021-09-23 liuhongt <hongtao.liu@intel.com>
5211 * config/i386/i386.md (rinthf2): New expander.
5212 (nearbyinthf2): New expander.
5214 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5216 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
5217 (class dom_jt_state): ...this and provide virtual overrides.
5218 (dom_jt_state::register_equiv): New.
5219 (class dom_jt_simplifier): Rename from
5220 dom_jump_threader_simplifier.
5221 (dom_jump_threader_simplifier::simplify): Rename...
5222 (dom_jt_simplifier::simplify): ...to this.
5223 (pass_dominator::execute): Use dom_jt_simplifier and
5225 * tree-ssa-threadedge.c (jump_threader::jump_threader):
5227 (jt_state::register_equivs_stmt): Abstract out...
5228 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
5230 (jump_threader::thread_around_empty_blocks): Update state.
5231 (jump_threader::thread_through_normal_block): Same.
5232 (jt_state::jt_state): Remove.
5233 (jt_state::push): Remove pass specific bits. Keep block vector
5235 (jt_state::append_path): New.
5236 (jt_state::pop): Remove pass specific bits.
5237 (jt_state::register_equiv): Same.
5238 (jt_state::record_ranges_from_stmt): Same.
5239 (jt_state::register_equivs_on_edge): Same. Rename...
5240 (jt_state::register_equivs_edge): ...to this.
5241 (jt_state::dump): New.
5242 (jt_state::debug): New.
5243 (jump_threader_simplifier::simplify): Remove.
5244 (jt_state::get_path): New.
5245 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
5246 class. Expose common functionality as virtual methods.
5247 (class jump_threader_simplifier): Same. Rename...
5248 (class jt_simplifier): ...to this.
5249 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
5250 (class vrp_jt_simplifier): ...to this. Provide pass specific
5252 (class vrp_jt_state): New.
5253 (vrp_jump_threader_simplifier::simplify): Rename...
5254 (vrp_jt_simplifier::simplify): ...to this. Inline code from
5255 what used to be the base class.
5256 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
5259 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
5262 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
5265 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
5266 Richard Biener <rguenther@suse.de>
5268 * match.pd (negation simplifications): Implement some negation
5269 folding transformations from fold-const.c's fold_negate_expr.
5270 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
5271 argument, to control whether the op should be simplified prior
5272 to looking up/assigning a value number.
5273 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
5274 (vn_nary_simplify): Likewise.
5275 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
5276 now call vn_nary_build_or_lookup_1 disabling simplification.
5278 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
5280 PR tree-optimization/102087
5281 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5282 Update bound/cmp/control for niter.
5284 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5286 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5287 Move check for non-empty BB here.
5288 (fur_source::register_outgoing_edges): ...from here.
5290 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5292 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
5293 Remove call to improve_range_with_equivs.
5294 (path_range_query::improve_range_with_equivs): Remove
5295 * gimple-range-path.h: Remove improve_range_with_equivs.
5297 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5299 * config/i386/avx512fp16intrin.h:
5300 (_mm512_mask_blend_ph): New intrinsic.
5301 (_mm512_permutex2var_ph): Ditto.
5302 (_mm512_permutexvar_ph): Ditto.
5303 * config/i386/avx512fp16vlintrin.h:
5304 (_mm256_mask_blend_ph): New intrinsic.
5305 (_mm256_permutex2var_ph): Ditto.
5306 (_mm256_permutexvar_ph): Ditto.
5307 (_mm_mask_blend_ph): Ditto.
5308 (_mm_permutex2var_ph): Ditto.
5309 (_mm_permutexvar_ph): Ditto.
5311 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5313 * config/i386/avx512fp16intrin.h: Add new intrinsics.
5314 (_mm512_conj_pch): New intrinsic.
5315 (_mm512_mask_conj_pch): Ditto.
5316 (_mm512_maskz_conj_pch): Ditto.
5317 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
5318 (_mm256_conj_pch): New intrinsic.
5319 (_mm256_mask_conj_pch): Ditto.
5320 (_mm256_maskz_conj_pch): Ditto.
5321 (_mm_conj_pch): Ditto.
5322 (_mm_mask_conj_pch): Ditto.
5323 (_mm_maskz_conj_pch): Ditto.
5325 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5327 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
5328 (_mm512_reduce_add_ph): New intrinsic.
5329 (_mm512_reduce_mul_ph): Ditto.
5330 (_mm512_reduce_min_ph): Ditto.
5331 (_mm512_reduce_max_ph): Ditto.
5332 * config/i386/avx512fp16vlintrin.h
5333 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
5334 (_mm256_reduce_add_ph): New intrinsic.
5335 (_mm256_reduce_mul_ph): Ditto.
5336 (_mm256_reduce_min_ph): Ditto.
5337 (_mm256_reduce_max_ph): Ditto.
5338 (_mm_reduce_add_ph): Ditto.
5339 (_mm_reduce_mul_ph): Ditto.
5340 (_mm_reduce_min_ph): Ditto.
5341 (_mm_reduce_max_ph): Ditto.
5343 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5345 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
5346 __m128h_u): New typedef.
5347 (_mm512_load_ph): New intrinsic.
5348 (_mm256_load_ph): Ditto.
5349 (_mm_load_ph): Ditto.
5350 (_mm512_loadu_ph): Ditto.
5351 (_mm256_loadu_ph): Ditto.
5352 (_mm_loadu_ph): Ditto.
5353 (_mm512_store_ph): Ditto.
5354 (_mm256_store_ph): Ditto.
5355 (_mm_store_ph): Ditto.
5356 (_mm512_storeu_ph): Ditto.
5357 (_mm256_storeu_ph): Ditto.
5358 (_mm_storeu_ph): Ditto.
5359 (_mm512_abs_ph): Ditto.
5360 * config/i386/avx512fp16vlintrin.h
5361 (_mm_abs_ph): Ditto.
5362 (_mm256_abs_ph): Ditto.
5364 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5366 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
5368 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5371 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
5372 is actually a full copy of the source operand into the target.
5373 Don't emit a strict low part move if source and target mode match.
5375 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5377 PR middle-end/102415
5378 * omp-expand.c (expand_omp_single): If region->exit is NULL,
5379 assert region->entry is GIMPLE_OMP_SCOPE region and return.
5381 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5383 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
5384 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
5385 arguments from 2 to 3.
5386 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
5387 allocate clause allocator and print align if present.
5388 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
5389 for omp_default_mem_alloc if align modifier is present. If align
5390 modifier is present, use TREE_LIST to encode both allocator and
5392 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
5393 Handle align modifier on allocator clause if present.
5395 2021-09-22 liuhongt <hongtao.liu@intel.com>
5397 * config/i386/i386.md (define_attr "isa"): Add
5399 (define_attr "enabled"): Correspond fma_or_avx512vl to
5400 TARGET_FMA || TARGET_AVX512VL.
5401 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
5406 2021-09-22 liuhongt <hongtao.liu@intel.com>
5408 * config/i386/i386.md (cstorehf3): New define_expand.
5410 2021-09-22 liuhongt <hongtao.liu@intel.com>
5412 * config/i386/i386.md (<rounding_insn>hf2): New expander.
5413 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
5414 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
5415 Extend from VF_128 to VFH_128.
5417 2021-09-22 liuhongt <hongtao.liu@intel.com>
5419 * config/i386/i386-features.c (i386-features.c): Handle
5421 * config/i386/i386.md (sqrthf2): New expander.
5422 (*sqrthf2): New define_insn.
5423 * config/i386/sse.md
5424 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
5427 2021-09-22 liuhongt <hongtao.liu@intel.com>
5429 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
5431 (_mm_mask3_fcmadd_sch): Likewise.
5432 (_mm_maskz_fcmadd_sch): Likewise.
5433 (_mm_fcmadd_sch): Likewise.
5434 (_mm_mask_fmadd_sch): Likewise.
5435 (_mm_mask3_fmadd_sch): Likewise.
5436 (_mm_maskz_fmadd_sch): Likewise.
5437 (_mm_fmadd_sch): Likewise.
5438 (_mm_mask_fcmadd_round_sch): Likewise.
5439 (_mm_mask3_fcmadd_round_sch): Likewise.
5440 (_mm_maskz_fcmadd_round_sch): Likewise.
5441 (_mm_fcmadd_round_sch): Likewise.
5442 (_mm_mask_fmadd_round_sch): Likewise.
5443 (_mm_mask3_fmadd_round_sch): Likewise.
5444 (_mm_maskz_fmadd_round_sch): Likewise.
5445 (_mm_fmadd_round_sch): Likewise.
5446 (_mm_fcmul_sch): Likewise.
5447 (_mm_mask_fcmul_sch): Likewise.
5448 (_mm_maskz_fcmul_sch): Likewise.
5449 (_mm_fmul_sch): Likewise.
5450 (_mm_mask_fmul_sch): Likewise.
5451 (_mm_maskz_fmul_sch): Likewise.
5452 (_mm_fcmul_round_sch): Likewise.
5453 (_mm_mask_fcmul_round_sch): Likewise.
5454 (_mm_maskz_fcmul_round_sch): Likewise.
5455 (_mm_fmul_round_sch): Likewise.
5456 (_mm_mask_fmul_round_sch): Likewise.
5457 (_mm_maskz_fmul_round_sch): Likewise.
5458 * config/i386/i386-builtin.def: Add corresponding new builtins.
5459 * config/i386/sse.md
5460 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
5461 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
5462 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
5464 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
5465 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
5467 * config/i386/subst.md (mask_scalarcz_name): New.
5468 (mask_scalarc_name): Ditto.
5469 (mask_scalarc_operand3): Ditto.
5470 (mask_scalarcz_operand4): Ditto.
5471 (round_scalarcz_name): Ditto.
5472 (round_scalarc_mask_operand3): Ditto.
5473 (round_scalarcz_mask_operand4): Ditto.
5474 (round_scalarc_mask_op3): Ditto.
5475 (round_scalarcz_mask_op4): Ditto.
5476 (round_scalarcz_constraint): Ditto.
5477 (round_scalarcz_nimm_predicate): Ditto.
5478 (mask_scalarcz): Ditto.
5479 (mask_scalarc): Ditto.
5480 (round_scalarcz): Ditto.
5482 2021-09-22 liuhongt <hongtao.liu@intel.com>
5484 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
5486 (_mm512_mask_fcmadd_pch): Likewise.
5487 (_mm512_mask3_fcmadd_pch): Likewise.
5488 (_mm512_maskz_fcmadd_pch): Likewise.
5489 (_mm512_fmadd_pch): Likewise.
5490 (_mm512_mask_fmadd_pch): Likewise.
5491 (_mm512_mask3_fmadd_pch): Likewise.
5492 (_mm512_maskz_fmadd_pch): Likewise.
5493 (_mm512_fcmadd_round_pch): Likewise.
5494 (_mm512_mask_fcmadd_round_pch): Likewise.
5495 (_mm512_mask3_fcmadd_round_pch): Likewise.
5496 (_mm512_maskz_fcmadd_round_pch): Likewise.
5497 (_mm512_fmadd_round_pch): Likewise.
5498 (_mm512_mask_fmadd_round_pch): Likewise.
5499 (_mm512_mask3_fmadd_round_pch): Likewise.
5500 (_mm512_maskz_fmadd_round_pch): Likewise.
5501 (_mm512_fcmul_pch): Likewise.
5502 (_mm512_mask_fcmul_pch): Likewise.
5503 (_mm512_maskz_fcmul_pch): Likewise.
5504 (_mm512_fmul_pch): Likewise.
5505 (_mm512_mask_fmul_pch): Likewise.
5506 (_mm512_maskz_fmul_pch): Likewise.
5507 (_mm512_fcmul_round_pch): Likewise.
5508 (_mm512_mask_fcmul_round_pch): Likewise.
5509 (_mm512_maskz_fcmul_round_pch): Likewise.
5510 (_mm512_fmul_round_pch): Likewise.
5511 (_mm512_mask_fmul_round_pch): Likewise.
5512 (_mm512_maskz_fmul_round_pch): Likewise.
5513 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
5515 (_mm_mask_fmadd_pch): Likewise.
5516 (_mm_mask3_fmadd_pch): Likewise.
5517 (_mm_maskz_fmadd_pch): Likewise.
5518 (_mm256_fmadd_pch): Likewise.
5519 (_mm256_mask_fmadd_pch): Likewise.
5520 (_mm256_mask3_fmadd_pch): Likewise.
5521 (_mm256_maskz_fmadd_pch): Likewise.
5522 (_mm_fcmadd_pch): Likewise.
5523 (_mm_mask_fcmadd_pch): Likewise.
5524 (_mm_mask3_fcmadd_pch): Likewise.
5525 (_mm_maskz_fcmadd_pch): Likewise.
5526 (_mm256_fcmadd_pch): Likewise.
5527 (_mm256_mask_fcmadd_pch): Likewise.
5528 (_mm256_mask3_fcmadd_pch): Likewise.
5529 (_mm256_maskz_fcmadd_pch): Likewise.
5530 (_mm_fmul_pch): Likewise.
5531 (_mm_mask_fmul_pch): Likewise.
5532 (_mm_maskz_fmul_pch): Likewise.
5533 (_mm256_fmul_pch): Likewise.
5534 (_mm256_mask_fmul_pch): Likewise.
5535 (_mm256_maskz_fmul_pch): Likewise.
5536 (_mm_fcmul_pch): Likewise.
5537 (_mm_mask_fcmul_pch): Likewise.
5538 (_mm_maskz_fcmul_pch): Likewise.
5539 (_mm256_fcmul_pch): Likewise.
5540 (_mm256_mask_fcmul_pch): Likewise.
5541 (_mm256_maskz_fcmul_pch): Likewise.
5542 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
5543 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
5544 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
5545 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
5546 * config/i386/i386-builtin.def: Add new builtins.
5547 * config/i386/i386-expand.c: Handle new builtin types.
5548 * config/i386/subst.md (SUBST_CV): New.
5549 (maskc_name): Ditto.
5550 (maskc_operand3): Ditto.
5552 (sdc_maskz_name): Ditto.
5553 (sdc_mask_op4): Ditto.
5554 (sdc_mask_op5): Ditto.
5555 (sdc_mask_mode512bit_condition): Ditto.
5557 (round_maskc_operand3): Ditto.
5558 (round_sdc_mask_operand4): Ditto.
5559 (round_maskc_op3): Ditto.
5560 (round_sdc_mask_op4): Ditto.
5561 (round_saeonly_sdc_mask_operand5): Ditto.
5562 * config/i386/sse.md (unspec): Add complex fma unspecs.
5563 (avx512fmaskcmode): New.
5564 (UNSPEC_COMPLEX_F_C_MA): Ditto.
5565 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
5566 (complexopname): Ditto.
5567 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
5568 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
5569 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
5571 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
5572 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
5574 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
5576 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
5577 rs6000-density-size-threshold, rs6000-density-penalty,
5578 rs6000-density-load-pct-threshold,
5579 rs6000-density-load-num-threshold): New parameter.
5580 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
5581 corresponding parameters.
5583 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5585 * gimple-range-path.cc (path_range_query::defined_outside_path):
5587 (path_range_query::range_on_path_entry): New.
5588 (path_range_query::internal_range_of_expr): Resolve unknowns
5590 (path_range_query::improve_range_with_equivs): New.
5591 (path_range_query::ssa_range_in_phi): Resolve unknowns with
5593 * gimple-range-path.h (class path_range_query): Add
5594 defined_outside_path, range_on_path_entry, and
5595 improve_range_with_equivs.
5597 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5599 * gimple-range-path.cc (path_range_query::add_to_imports): New.
5600 (path_range_query::add_copies_to_imports): New.
5601 (path_range_query::precompute_ranges): Call
5602 add_copies_to_imports.
5603 * gimple-range-path.h (class path_range_query): Add prototypes
5604 for add_copies_to_imports and add_to_imports.
5606 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5608 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5609 Remove useless code.
5611 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5613 * gimple-range-fold.h (class fur_source): Make oracle protected.
5614 * gimple-range-path.cc (path_range_query::path_range_query): Add
5615 resolve argument. Initialize oracle.
5616 (path_range_query::~path_range_query): Delete oracle.
5617 (path_range_query::range_of_stmt): Adapt to use relations.
5618 (path_range_query::precompute_ranges): Pre-compute relations.
5619 (class jt_fur_source): New
5620 (jt_fur_source::jt_fur_source): New.
5621 (jt_fur_source::register_relation): New.
5622 (jt_fur_source::query_relation): New.
5623 (path_range_query::precompute_relations): New.
5624 (path_range_query::precompute_phi_relations): New.
5625 * gimple-range-path.h (path_range_query): Add resolve argument.
5626 Add oracle, precompute_relations, precompute_phi_relations.
5627 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
5628 resolve argument to solver.
5630 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5632 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5633 Rename postfold_gcond_edges to register_outgoing_edges and
5635 (fold_using_range::postfold_gcond_edges): Rename...
5636 (fur_source::register_outgoing_edges): ...to this.
5637 * gimple-range-fold.h (postfold_gcond_edges): Rename to
5638 register_outgoing_edges and move to fur_source.
5640 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5642 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
5643 dom_info_available_p.
5645 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5647 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
5648 and quick_grow_cleared instead of safe_grow_cleared.
5650 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
5653 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
5656 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
5658 * configure.ac: Detect when the assembler supports new-style
5659 architecture extensions.
5660 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
5661 the full CPU string if the assembler can grok it.
5662 (arm_rewrite_march): Likewise but for the architecture.
5663 * config.in: Regenerate.
5664 * configure: Regenerate.
5666 2021-09-21 Richard Biener <rguenther@suse.de>
5668 PR tree-optimization/102421
5669 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
5670 adjust alignment info.
5672 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
5674 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
5675 bitfield stream out.
5677 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
5679 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
5680 undefined edges, apply an equivalence if appropriate.
5681 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
5682 UNDEFINED if EDGE_EXECUTABLE is not set.
5683 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
5684 as EXECUTABLE upon startup.
5685 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
5686 EDGE_EXECUTABLE set.
5687 * vr-values.c (set_and_propagate_unexecutable): New.
5688 (simplify_using_ranges::fold_cond): Call set_and_propagate.
5689 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
5690 * vr-values.h: Add prototype.
5692 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
5694 * value-relation.cc (equiv_oracle::register_initial_def): New.
5695 (equiv_oracle::register_relation): Call register_initial_def.
5696 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
5697 (relation_oracle::register_stmt): Check def block of PHI arguments.
5698 * value-relation.h (equiv_oracle): Add new prototypes.
5700 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
5702 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5703 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
5704 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
5705 __ROUNDING_MATH__ according to their corresponding flags.
5706 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
5707 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
5708 and __ROUNDING_MATH__.
5710 2021-09-20 Richard Biener <rguenther@suse.de>
5712 * tree-vect-stmts.c (vectorizable_load): Use the vectype
5715 2021-09-20 Richard Biener <rguenther@suse.de>
5717 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
5718 Do not compute alignment of the vectorized access here.
5720 2021-09-20 Richard Biener <rguenther@suse.de>
5722 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
5723 Store -1 for runtime alias peeling iterations.
5725 2021-09-20 Richard Biener <rguenther@suse.de>
5727 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
5729 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
5731 * input.c (string_concat_db::record_string_concatenation)
5732 (string_concat_db::get_string_concatenation): Skip for
5733 'RESERVED_LOCATION_P'.
5735 2021-09-20 Richard Biener <rguenther@suse.de>
5737 PR tree-optimization/65206
5738 * tree-data-ref.h (struct data_reference): Add alt_indices,
5740 * tree-data-ref.c (free_data_ref): Release alt_indices.
5741 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
5742 (create_data_ref): Adjust.
5743 (initialize_data_dependence_relation): Split into head
5744 and tail. When the base objects fail to match up try
5745 again with pointer-based analysis of indices.
5746 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
5747 not compare the lazily computed alternate set of indices.
5749 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
5751 * gcc.c: Test for execute OK when we find the
5752 programs for assembler linker and dsymutil and those
5753 were specified at configure-time.
5755 2021-09-19 Martin Sebor <msebor@redhat.com>
5757 PR middle-end/102403
5758 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
5759 Correct a function pre/postcondition.
5761 2021-09-19 Martin Sebor <msebor@redhat.com>
5763 PR middle-end/102243
5764 * tree-ssa-strlen.c (get_range): Handle null cfun.
5766 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
5768 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
5769 unwinder shim as a convenience library.
5771 2021-09-19 Andrew Pinski <apinski@marvell.com>
5773 * doc/install.texi: Add note about
5774 binutils 2.35 is required for LTO usage.
5776 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
5778 * tree-ssa-threadbackward.c
5779 (back_threader_registry::register_path): Use push_edge.
5780 * tree-ssa-threadedge.c
5781 (jump_threader::thread_around_empty_blocks): Same.
5782 (jump_threader::thread_through_normal_block): Same.
5783 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
5785 * tree-ssa-threadupdate.c
5786 (jt_path_registry::allocate_thread_edge): Remove.
5787 (jt_path_registry::push_edge): New.
5788 (dump_jump_thread_path): Make static.
5789 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
5792 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
5794 * gimple-range-path.cc (path_range_query::path_range_query): Add
5796 (path_range_query::dump): Remove extern declaration of dump_ranger.
5797 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
5798 * gimple-range-trace.h (dump_ranger): Add prototype.
5800 2021-09-19 John Ericson <git@JohnEricson.me>
5802 * gcc.c (find_a_program): New function, factored out of...
5803 (find_a_file): Here.
5804 (execute): Use find_a_program when looking for programs rather
5807 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
5809 * config/avr/avr-mcus.def: Add atmega324pb.
5810 * doc/avr-mmcu.texi: Corresponding changes.
5812 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
5815 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
5816 the same transformations as to @1. For comparisons against NaN,
5817 don't check HONOR_SNANS but confirm that neither operand is a
5820 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
5822 * attribs.c (make_unique_name): Delete.
5823 * attribs.h (make_unique_name): Delete.
5825 2021-09-19 Andrew Pinski <apinski@marvell.com>
5827 * lra-constraints.c (check_and_process_move): Assert
5828 that dclass and sclass are greater than or equal to NO_REGS.
5830 2021-09-18 Jakub Jelinek <jakub@redhat.com>
5832 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
5833 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
5834 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
5836 2021-09-18 liuhongt <hongtao.liu@intel.com>
5838 * config/i386/i386-features.c (remove_partial_avx_dependency):
5839 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
5840 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
5842 2021-09-18 Jakub Jelinek <jakub@redhat.com>
5844 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
5845 if file/namespace scope variable doesn't have predetermined sharing,
5846 treat it as if there was default(none).
5848 2021-09-18 liuhongt <hongtao.liu@intel.com>
5850 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
5852 (_mm_mask_fmadd_sh): Likewise.
5853 (_mm_mask3_fmadd_sh): Likewise.
5854 (_mm_maskz_fmadd_sh): Likewise.
5855 (_mm_fmadd_round_sh): Likewise.
5856 (_mm_mask_fmadd_round_sh): Likewise.
5857 (_mm_mask3_fmadd_round_sh): Likewise.
5858 (_mm_maskz_fmadd_round_sh): Likewise.
5859 (_mm_fnmadd_sh): Likewise.
5860 (_mm_mask_fnmadd_sh): Likewise.
5861 (_mm_mask3_fnmadd_sh): Likewise.
5862 (_mm_maskz_fnmadd_sh): Likewise.
5863 (_mm_fnmadd_round_sh): Likewise.
5864 (_mm_mask_fnmadd_round_sh): Likewise.
5865 (_mm_mask3_fnmadd_round_sh): Likewise.
5866 (_mm_maskz_fnmadd_round_sh): Likewise.
5867 (_mm_fmsub_sh): Likewise.
5868 (_mm_mask_fmsub_sh): Likewise.
5869 (_mm_mask3_fmsub_sh): Likewise.
5870 (_mm_maskz_fmsub_sh): Likewise.
5871 (_mm_fmsub_round_sh): Likewise.
5872 (_mm_mask_fmsub_round_sh): Likewise.
5873 (_mm_mask3_fmsub_round_sh): Likewise.
5874 (_mm_maskz_fmsub_round_sh): Likewise.
5875 (_mm_fnmsub_sh): Likewise.
5876 (_mm_mask_fnmsub_sh): Likewise.
5877 (_mm_mask3_fnmsub_sh): Likewise.
5878 (_mm_maskz_fnmsub_sh): Likewise.
5879 (_mm_fnmsub_round_sh): Likewise.
5880 (_mm_mask_fnmsub_round_sh): Likewise.
5881 (_mm_mask3_fnmsub_round_sh): Likewise.
5882 (_mm_maskz_fnmsub_round_sh): Likewise.
5883 * config/i386/i386-builtin-types.def
5884 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
5885 * config/i386/i386-builtin.def: Add new builtins.
5886 * config/i386/i386-expand.c: Handle new builtin type.
5887 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
5888 Ajdust to support FP16.
5889 (fmai_vmfmsub_<mode><round_name>): Ditto.
5890 (fmai_vmfnmadd_<mode><round_name>): Ditto.
5891 (fmai_vmfnmsub_<mode><round_name>): Ditto.
5892 (*fmai_fmadd_<mode>): Ditto.
5893 (*fmai_fmsub_<mode>): Ditto.
5894 (*fmai_fnmadd_<mode><round_name>): Ditto.
5895 (*fmai_fnmsub_<mode><round_name>): Ditto.
5896 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
5897 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
5898 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
5899 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
5900 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
5901 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
5902 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
5903 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
5904 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
5905 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
5906 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
5907 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
5908 adjust to support FP16.
5909 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
5910 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
5911 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
5914 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
5916 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
5918 (maskload<mode><avx512fmaskmodelower>): Ditto.
5919 (maskstore<mode><avx512fmaskmodelower>): Ditto.
5921 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
5923 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
5925 (ix86_expand_copysign): Ditto.
5926 (ix86_expand_xorsign): Ditto.
5927 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
5929 (ix86_build_signbit_mask): Ditto.
5930 (ix86_can_change_mode_class): Ditto.
5931 * config/i386/i386.md
5932 (SSEMODEF): Add HFmode.
5933 (ssevecmodef): Ditto.
5934 (<code>hf2): New define_expand.
5935 (*<code>hf2_1): New define_insn_and_split.
5936 (copysign<mode>): Extend to support HFmode under AVX512FP16.
5937 (xorsign<mode>): Ditto.
5938 * config/i386/sse.md (VFB): New mode iterator.
5939 (VFB_128_256): Ditto.
5941 (sseintvecmode2): Support HF vector mode.
5942 (<code><mode>2): Use new mode iterator.
5943 (*<code><mode>2): Ditto.
5944 (copysign<mode>3): Ditto.
5945 (xorsign<mode>3): Ditto.
5946 (<code><mode>3<mask_name>): Ditto.
5947 (<code><mode>3<mask_name>): Ditto.
5948 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
5949 (<sse>_andnot<mode>3<mask_name>): Ditto.
5950 (*<code><mode>3<mask_name>): Ditto.
5951 (*<code><mode>3<mask_name>): Ditto.
5953 2021-09-18 liuhongt <hongtao.liu@intel.com>
5955 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
5957 (_mm512_mask3_fmadd_ph): Likewise.
5958 (_mm512_maskz_fmadd_ph): Likewise.
5959 (_mm512_fmadd_round_ph): Likewise.
5960 (_mm512_mask_fmadd_round_ph): Likewise.
5961 (_mm512_mask3_fmadd_round_ph): Likewise.
5962 (_mm512_maskz_fmadd_round_ph): Likewise.
5963 (_mm512_fnmadd_ph): Likewise.
5964 (_mm512_mask_fnmadd_ph): Likewise.
5965 (_mm512_mask3_fnmadd_ph): Likewise.
5966 (_mm512_maskz_fnmadd_ph): Likewise.
5967 (_mm512_fnmadd_round_ph): Likewise.
5968 (_mm512_mask_fnmadd_round_ph): Likewise.
5969 (_mm512_mask3_fnmadd_round_ph): Likewise.
5970 (_mm512_maskz_fnmadd_round_ph): Likewise.
5971 (_mm512_fmsub_ph): Likewise.
5972 (_mm512_mask_fmsub_ph): Likewise.
5973 (_mm512_mask3_fmsub_ph): Likewise.
5974 (_mm512_maskz_fmsub_ph): Likewise.
5975 (_mm512_fmsub_round_ph): Likewise.
5976 (_mm512_mask_fmsub_round_ph): Likewise.
5977 (_mm512_mask3_fmsub_round_ph): Likewise.
5978 (_mm512_maskz_fmsub_round_ph): Likewise.
5979 (_mm512_fnmsub_ph): Likewise.
5980 (_mm512_mask_fnmsub_ph): Likewise.
5981 (_mm512_mask3_fnmsub_ph): Likewise.
5982 (_mm512_maskz_fnmsub_ph): Likewise.
5983 (_mm512_fnmsub_round_ph): Likewise.
5984 (_mm512_mask_fnmsub_round_ph): Likewise.
5985 (_mm512_mask3_fnmsub_round_ph): Likewise.
5986 (_mm512_maskz_fnmsub_round_ph): Likewise.
5987 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
5989 (_mm256_mask_fmadd_ph): Likewise.
5990 (_mm256_mask3_fmadd_ph): Likewise.
5991 (_mm256_maskz_fmadd_ph): Likewise.
5992 (_mm_fmadd_ph): Likewise.
5993 (_mm_mask_fmadd_ph): Likewise.
5994 (_mm_mask3_fmadd_ph): Likewise.
5995 (_mm_maskz_fmadd_ph): Likewise.
5996 (_mm256_fnmadd_ph): Likewise.
5997 (_mm256_mask_fnmadd_ph): Likewise.
5998 (_mm256_mask3_fnmadd_ph): Likewise.
5999 (_mm256_maskz_fnmadd_ph): Likewise.
6000 (_mm_fnmadd_ph): Likewise.
6001 (_mm_mask_fnmadd_ph): Likewise.
6002 (_mm_mask3_fnmadd_ph): Likewise.
6003 (_mm_maskz_fnmadd_ph): Likewise.
6004 (_mm256_fmsub_ph): Likewise.
6005 (_mm256_mask_fmsub_ph): Likewise.
6006 (_mm256_mask3_fmsub_ph): Likewise.
6007 (_mm256_maskz_fmsub_ph): Likewise.
6008 (_mm_fmsub_ph): Likewise.
6009 (_mm_mask_fmsub_ph): Likewise.
6010 (_mm_mask3_fmsub_ph): Likewise.
6011 (_mm_maskz_fmsub_ph): Likewise.
6012 (_mm256_fnmsub_ph): Likewise.
6013 (_mm256_mask_fnmsub_ph): Likewise.
6014 (_mm256_mask3_fnmsub_ph): Likewise.
6015 (_mm256_maskz_fnmsub_ph): Likewise.
6016 (_mm_fnmsub_ph): Likewise.
6017 (_mm_mask_fnmsub_ph): Likewise.
6018 (_mm_mask3_fnmsub_ph): Likewise.
6019 (_mm_maskz_fnmsub_ph): Likewise.
6020 * config/i386/i386-builtin.def: Add corresponding new builtins.
6021 * config/i386/sse.md
6022 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
6023 support HF vector modes.
6024 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
6026 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
6027 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
6028 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
6029 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
6030 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
6031 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
6032 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
6034 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
6035 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
6036 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
6037 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
6038 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
6039 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
6041 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
6042 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
6043 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
6044 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
6045 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
6046 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
6047 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
6049 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
6050 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
6051 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
6052 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
6053 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
6055 2021-09-18 liuhongt <hongtao.liu@intel.com>
6057 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
6059 (_mm512_mask_fmaddsub_ph): Likewise.
6060 (_mm512_mask3_fmaddsub_ph): Likewise.
6061 (_mm512_maskz_fmaddsub_ph): Likewise.
6062 (_mm512_fmaddsub_round_ph): Likewise.
6063 (_mm512_mask_fmaddsub_round_ph): Likewise.
6064 (_mm512_mask3_fmaddsub_round_ph): Likewise.
6065 (_mm512_maskz_fmaddsub_round_ph): Likewise.
6066 (_mm512_mask_fmsubadd_ph): Likewise.
6067 (_mm512_mask3_fmsubadd_ph): Likewise.
6068 (_mm512_maskz_fmsubadd_ph): Likewise.
6069 (_mm512_fmsubadd_round_ph): Likewise.
6070 (_mm512_mask_fmsubadd_round_ph): Likewise.
6071 (_mm512_mask3_fmsubadd_round_ph): Likewise.
6072 (_mm512_maskz_fmsubadd_round_ph): Likewise.
6073 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
6075 (_mm256_mask_fmaddsub_ph): Likewise.
6076 (_mm256_mask3_fmaddsub_ph): Likewise.
6077 (_mm256_maskz_fmaddsub_ph): Likewise.
6078 (_mm_fmaddsub_ph): Likewise.
6079 (_mm_mask_fmaddsub_ph): Likewise.
6080 (_mm_mask3_fmaddsub_ph): Likewise.
6081 (_mm_maskz_fmaddsub_ph): Likewise.
6082 (_mm256_fmsubadd_ph): Likewise.
6083 (_mm256_mask_fmsubadd_ph): Likewise.
6084 (_mm256_mask3_fmsubadd_ph): Likewise.
6085 (_mm256_maskz_fmsubadd_ph): Likewise.
6086 (_mm_fmsubadd_ph): Likewise.
6087 (_mm_mask_fmsubadd_ph): Likewise.
6088 (_mm_mask3_fmsubadd_ph): Likewise.
6089 (_mm_maskz_fmsubadd_ph): Likewise.
6090 * config/i386/i386-builtin.def: Add corresponding new builtins.
6091 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
6092 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
6093 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
6095 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
6097 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
6098 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
6099 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
6101 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
6102 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
6104 2021-09-18 liuhongt <hongtao.liu@intel.com>
6107 * config/i386/i386.c (ix86_print_operand): Handle
6108 V8HF/V16HF/V32HFmode.
6109 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
6110 * config/i386/sse.md (avx512bcst): Remove.
6112 2021-09-17 Martin Sebor <msebor@redhat.com>
6114 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
6115 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
6116 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
6117 (check_defs): Add comment.
6118 (can_skip_redundant_opnd): Update comment.
6119 (compute_uninit_opnds_pos): Adjust to namespace change.
6120 (find_pdom): Move to gimple-predicate-analysis.cc.
6122 (struct uninit_undef_val_t): New.
6123 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
6124 (find_control_equiv_block): Same.
6125 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
6126 (MAX_SWITCH_CASES): Same.
6127 (compute_control_dep_chain): Same.
6128 (find_uninit_use): Use predicate analyzer.
6129 (struct pred_info): Move to gimple-predicate-analysis.
6130 (convert_control_dep_chain_into_preds): Same.
6131 (find_predicates): Same.
6132 (collect_phi_def_edges): Same.
6133 (warn_uninitialized_phi): Use predicate analyzer.
6134 (find_def_preds): Move to gimple-predicate-analysis.
6135 (dump_pred_info): Same.
6136 (dump_pred_chain): Same.
6137 (dump_predicates): Same.
6138 (destroy_predicate_vecs): Remove.
6139 (execute_late_warn_uninitialized): New.
6140 (get_cmp_code): Move to gimple-predicate-analysis.
6141 (is_value_included_in): Same.
6142 (value_sat_pred_p): Same.
6143 (find_matching_predicate_in_rest_chains): Same.
6144 (is_use_properly_guarded): Same.
6145 (prune_uninit_phi_opnds): Same.
6146 (find_var_cmp_const): Same.
6147 (use_pred_not_overlap_with_undef_path_pred): Same.
6148 (pred_equal_p): Same.
6149 (is_neq_relop_p): Same.
6150 (is_neq_zero_form_p): Same.
6151 (pred_expr_equal_p): Same.
6152 (is_pred_expr_subset_of): Same.
6153 (is_pred_chain_subset_of): Same.
6154 (is_included_in): Same.
6155 (is_superset_of): Same.
6157 (simplify_pred): Same.
6158 (simplify_preds_2): Same.
6159 (simplify_preds_3): Same.
6160 (simplify_preds_4): Same.
6161 (simplify_preds): Same.
6163 (push_to_worklist): Same.
6164 (get_pred_info_from_cmp): Same.
6165 (is_degenerated_phi): Same.
6166 (normalize_one_pred_1): Same.
6167 (normalize_one_pred): Same.
6168 (normalize_one_pred_chain): Same.
6169 (normalize_preds): Same.
6170 (can_one_predicate_be_invalidated_p): Same.
6171 (can_chain_union_be_invalidated_p): Same.
6172 (uninit_uses_cannot_happen): Same.
6173 (pass_late_warn_uninitialized::execute): Define.
6174 * gimple-predicate-analysis.cc: New file.
6175 * gimple-predicate-analysis.h: New file.
6177 2021-09-17 Julian Brown <julian@codesourcery.com>
6179 * config/gcn/gcn.c (gimple.h): Include.
6180 (gcn_fork_join): Emit barrier for worker-level joins.
6181 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
6182 writes_gang_private bitmap parameter. Set bit for blocks
6183 containing gang-private variable writes.
6184 (worker_single_simple): Don't emit barrier after predicated block.
6185 (worker_single_copy): Don't emit barrier if we're not broadcasting
6186 anything and the block contains no gang-private writes.
6187 (neuter_worker_single): Don't predicate blocks that only contain
6188 NOPs or internal marker functions. Pass has_gang_private_write
6189 argument to worker_single_copy.
6190 (oacc_do_neutering): Add writes_gang_private bitmap handling.
6192 2021-09-17 Julian Brown <julian@codesourcery.com>
6194 * config/gcn/gcn-protos.h
6195 (gcn_goacc_create_worker_broadcast_record): Update prototype.
6196 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
6197 preallocated block of LDS memory. Do not cache/share decls for
6198 reduction temporaries between invocations.
6199 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
6200 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
6201 and return temporary LDS space at that offset. Return pointer in
6203 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
6205 (ACC_LDS_SIZE): Define as acc_lds_size.
6206 (gcn_init_machine_status): Don't initialise lds_allocated,
6207 lds_allocs, reduc_decls fields of machine function struct.
6208 (gcn_option_override): Handle default size for gang-private
6209 variables and -mgang-private-size option.
6210 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
6211 initialising M0_REG.
6212 (gcn_shared_mem_layout): New function.
6213 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
6214 gang_private_hwm variable.
6215 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
6216 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
6217 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
6218 * config/gcn/gcn.opt (gang_private_size_opt): New global.
6219 (mgang-private-size=): New option.
6220 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
6222 * doc/tm.texi: Regenerate.
6223 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
6225 (build_sender_ref): Handle sender_decl being pointer.
6226 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
6227 parameters. Pass placement argument to
6228 create_worker_broadcast_record hook invocations. Handle
6229 sender_decl being pointer and isolate_broadcasts inserting extra
6231 (blk_offset_map_t): Add typedef.
6232 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
6233 preallocated range to worker_single_copy call.
6234 (dfs_broadcast_reachable_1): New function.
6235 (idx_decl_pair_t, used_range_vec_t): New typedefs.
6236 (sort_size_descending): New function.
6237 (addr_range): New class.
6238 (splay_tree_compare_addr_range, splay_tree_free_key)
6239 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
6240 (execute_omp_oacc_neuter_broadcast): Rename to...
6241 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
6242 parameters. Arrange layout of shared memory for broadcast
6244 (execute_omp_oacc_neuter_broadcast): New function.
6245 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
6246 handling from here. Enable pass for all OpenACC routines in order
6247 to call shared memory-layout hook.
6248 * target.def (create_worker_broadcast_record): Add OFFSET
6250 (shared_mem_layout): New hook.
6252 2021-09-17 Julian Brown <julian@codesourcery.com>
6253 Thomas Schwinge <thomas@codesourcery.com>
6255 * omp-oacc-neuter-broadcast.cc
6256 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
6258 (execute_omp_oacc_neuter_broadcast): Adjust.
6260 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6262 * value-relation.cc (class equiv_chain): Move to header file.
6263 (path_oracle::path_oracle): New.
6264 (path_oracle::~path_oracle): New.
6265 (path_oracle::register_relation): New.
6266 (path_oracle::query_relation): New.
6267 (path_oracle::reset_path): New.
6268 (path_oracle::dump): New.
6269 * value-relation.h (class equiv_chain): Move to here.
6270 (class path_oracle): New.
6272 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6274 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
6276 * gimple-range-fold.cc (fur_depend::register_relation): Use
6277 register_stmt/edge routines.
6278 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
6279 (equiv_oracle::equiv_oracle): Create self equivalence cache.
6280 (equiv_oracle::~equiv_oracle): Release same.
6281 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
6282 are no equivalences.
6283 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
6284 (equiv_oracle::register_relation): Rename from register_equiv.
6285 (relation_chain_head::find_relation): Relocate from dom_oracle.
6286 (relation_oracle::register_stmt): New.
6287 (relation_oracle::register_edge): New.
6288 (dom_oracle::*): Rename from relation_oracle.
6289 (dom_oracle::register_relation): Adjust to call equiv_oracle.
6290 (dom_oracle::set_one_relation): Split from register_relation.
6291 (dom_oracle::register_transitives): Consolidate 2 methods.
6292 (dom_oracle::find_relation_block): Move core to relation_chain.
6293 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
6294 * value-relation.h (class relation_oracle): New pure virtual base.
6295 (class equiv_oracle): Inherit from relation_oracle and adjust.
6296 (class dom_oracle): Rename from old relation_oracle and adjust.
6298 2021-09-17 Martin Sebor <msebor@redhat.com>
6300 PR middle-end/102200
6301 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
6302 (handle_min_max_size): Change argument. Store original SSA_NAME for
6303 operands to potentially distinct (sub)objects.
6304 (compute_objsize_r): Adjust call to the above.
6306 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6308 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
6309 (rs6000_new_builtin_vectorized_function): New function.
6310 (rs6000_new_builtin_md_vectorized_function): Likewise.
6311 (rs6000_builtin_vectorized_function): Call
6312 rs6000_new_builtin_vectorized_function.
6313 (rs6000_builtin_md_vectorized_function): Call
6314 rs6000_new_builtin_md_vectorized_function.
6316 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6318 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
6319 (ASSEMBLE_PAIR): Likewise.
6320 (BUILD_ACC): Likewise.
6321 (DISASSEMBLE_ACC): Likewise.
6322 (DISASSEMBLE_PAIR): Likewise.
6323 (PMXVBF16GER2): Likewise.
6324 (PMXVBF16GER2NN): Likewise.
6325 (PMXVBF16GER2NP): Likewise.
6326 (PMXVBF16GER2PN): Likewise.
6327 (PMXVBF16GER2PP): Likewise.
6328 (PMXVF16GER2): Likewise.
6329 (PMXVF16GER2NN): Likewise.
6330 (PMXVF16GER2NP): Likewise.
6331 (PMXVF16GER2PN): Likewise.
6332 (PMXVF16GER2PP): Likewise.
6333 (PMXVF32GER): Likewise.
6334 (PMXVF32GERNN): Likewise.
6335 (PMXVF32GERNP): Likewise.
6336 (PMXVF32GERPN): Likewise.
6337 (PMXVF32GERPP): Likewise.
6338 (PMXVF64GER): Likewise.
6339 (PMXVF64GERNN): Likewise.
6340 (PMXVF64GERNP): Likewise.
6341 (PMXVF64GERPN): Likewise.
6342 (PMXVF64GERPP): Likewise.
6343 (PMXVI16GER2): Likewise.
6344 (PMXVI16GER2PP): Likewise.
6345 (PMXVI16GER2S): Likewise.
6346 (PMXVI16GER2SPP): Likewise.
6347 (PMXVI4GER8): Likewise.
6348 (PMXVI4GER8PP): Likewise.
6349 (PMXVI8GER4): Likewise.
6350 (PMXVI8GER4PP): Likewise.
6351 (PMXVI8GER4SPP): Likewise.
6352 (XVBF16GER2): Likewise.
6353 (XVBF16GER2NN): Likewise.
6354 (XVBF16GER2NP): Likewise.
6355 (XVBF16GER2PN): Likewise.
6356 (XVBF16GER2PP): Likewise.
6357 (XVF16GER2): Likewise.
6358 (XVF16GER2NN): Likewise.
6359 (XVF16GER2NP): Likewise.
6360 (XVF16GER2PN): Likewise.
6361 (XVF16GER2PP): Likewise.
6362 (XVF32GER): Likewise.
6363 (XVF32GERNN): Likewise.
6364 (XVF32GERNP): Likewise.
6365 (XVF32GERPN): Likewise.
6366 (XVF32GERPP): Likewise.
6367 (XVF64GER): Likewise.
6368 (XVF64GERNN): Likewise.
6369 (XVF64GERNP): Likewise.
6370 (XVF64GERPN): Likewise.
6371 (XVF64GERPP): Likewise.
6372 (XVI16GER2): Likewise.
6373 (XVI16GER2PP): Likewise.
6374 (XVI16GER2S): Likewise.
6375 (XVI16GER2SPP): Likewise.
6376 (XVI4GER8): Likewise.
6377 (XVI4GER8PP): Likewise.
6378 (XVI8GER4): Likewise.
6379 (XVI8GER4PP): Likewise.
6380 (XVI8GER4SPP): Likewise.
6381 (XXMFACC): Likewise.
6382 (XXMTACC): Likewise.
6383 (XXSETACCZ): Likewise.
6384 (ASSEMBLE_PAIR_V): Likewise.
6385 (BUILD_PAIR): Likewise.
6386 (DISASSEMBLE_PAIR_V): Likewise.
6389 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
6390 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
6391 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
6392 (parse_bif_attrs): Handle ismmaint.
6393 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
6394 (write_bif_static_init): Handle ismmaint.
6396 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6398 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
6400 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
6401 (rs6000_new_builtin_valid_without_lhs): New function.
6402 (rs6000_gimple_fold_new_mma_builtin): Likewise.
6403 (rs6000_gimple_fold_new_builtin): Likewise.
6405 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
6407 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
6408 Destruct stale Value objects.
6409 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6412 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
6415 * match.pd (shift optimizations): Disable recent sign-changing
6416 optimization for shifts by zero, these will be folded later.
6418 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6420 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
6421 [power9] to [always].
6423 2021-09-17 Richard Biener <rguenther@suse.de>
6425 * tree-vect-stmts.c (vectorizable_load): Do not frob
6428 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6430 * config/i386/i386-features.c (remove_partial_avx_dependency):
6431 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
6432 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
6434 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
6436 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6437 * config/i386/i386.md (SSE FP to FP splitters): Replace
6438 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
6439 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
6440 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
6441 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
6442 * config/i386/x86-tune.def
6443 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
6444 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6446 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6449 * config/i386/i386-features.c (remove_partial_avx_dependency):
6450 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
6451 before generating vxorps.
6453 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6455 * config/i386/i386-options.c (processor_cost_table): Use
6456 tremont_cost for Tremont.
6457 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
6458 (tremont_memset): Likewise.
6459 (tremont_cost): Likewise.
6460 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
6463 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6465 * common/config/i386/i386-common.c: Use Haswell scheduling model
6467 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
6469 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
6471 (ix86_adjust_cost): Handle Tremont.
6472 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
6474 (X86_TUNE_USE_LEAVE): Likewise.
6475 (X86_TUNE_PUSH_MEMORY): Likewise.
6476 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
6477 (X86_TUNE_USE_CLTD): Likewise.
6478 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
6479 (X86_TUNE_AVOID_MFENCE): Likewise.
6480 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
6481 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
6482 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
6483 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
6484 (X86_TUNE_OPT_AGU): Likewise.
6485 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
6486 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
6487 (X86_TUNE_EXPAND_ABS): Likewise.
6488 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
6489 (X86_TUNE_SLOW_PSHUFB): Likewise.
6491 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
6493 PR rtl-optimization/102306
6494 * combine.c (try_combine): Abort the combination if we are about to
6495 duplicate volatile references.
6497 2021-09-17 liuhongt <hongtao.liu@intel.com>
6499 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
6501 (_mm256_undefined_ph): Likewise.
6502 (_mm512_undefined_ph): Likewise.
6503 (_mm_cvtsh_h): Likewise.
6504 (_mm256_cvtsh_h): Likewise.
6505 (_mm512_cvtsh_h): Likewise.
6506 (_mm512_castph_ps): Likewise.
6507 (_mm512_castph_pd): Likewise.
6508 (_mm512_castph_si512): Likewise.
6509 (_mm512_castph512_ph128): Likewise.
6510 (_mm512_castph512_ph256): Likewise.
6511 (_mm512_castph128_ph512): Likewise.
6512 (_mm512_castph256_ph512): Likewise.
6513 (_mm512_zextph128_ph512): Likewise.
6514 (_mm512_zextph256_ph512): Likewise.
6515 (_mm512_castps_ph): Likewise.
6516 (_mm512_castpd_ph): Likewise.
6517 (_mm512_castsi512_ph): Likewise.
6518 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
6520 (_mm256_castph_ps): Likewise.
6521 (_mm_castph_pd): Likewise.
6522 (_mm256_castph_pd): Likewise.
6523 (_mm_castph_si128): Likewise.
6524 (_mm256_castph_si256): Likewise.
6525 (_mm_castps_ph): Likewise.
6526 (_mm256_castps_ph): Likewise.
6527 (_mm_castpd_ph): Likewise.
6528 (_mm256_castpd_ph): Likewise.
6529 (_mm_castsi128_ph): Likewise.
6530 (_mm256_castsi256_ph): Likewise.
6531 (_mm256_castph256_ph128): Likewise.
6532 (_mm256_castph128_ph256): Likewise.
6533 (_mm256_zextph128_ph256): Likewise.
6535 2021-09-17 liuhongt <hongtao.liu@intel.com>
6537 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
6539 (_mm_mask_cvtsh_ss): Likewise.
6540 (_mm_maskz_cvtsh_ss): Likewise.
6541 (_mm_cvtsh_sd): Likewise.
6542 (_mm_mask_cvtsh_sd): Likewise.
6543 (_mm_maskz_cvtsh_sd): Likewise.
6544 (_mm_cvt_roundsh_ss): Likewise.
6545 (_mm_mask_cvt_roundsh_ss): Likewise.
6546 (_mm_maskz_cvt_roundsh_ss): Likewise.
6547 (_mm_cvt_roundsh_sd): Likewise.
6548 (_mm_mask_cvt_roundsh_sd): Likewise.
6549 (_mm_maskz_cvt_roundsh_sd): Likewise.
6550 (_mm_cvtss_sh): Likewise.
6551 (_mm_mask_cvtss_sh): Likewise.
6552 (_mm_maskz_cvtss_sh): Likewise.
6553 (_mm_cvtsd_sh): Likewise.
6554 (_mm_mask_cvtsd_sh): Likewise.
6555 (_mm_maskz_cvtsd_sh): Likewise.
6556 (_mm_cvt_roundss_sh): Likewise.
6557 (_mm_mask_cvt_roundss_sh): Likewise.
6558 (_mm_maskz_cvt_roundss_sh): Likewise.
6559 (_mm_cvt_roundsd_sh): Likewise.
6560 (_mm_mask_cvt_roundsd_sh): Likewise.
6561 (_mm_maskz_cvt_roundsd_sh): Likewise.
6562 * config/i386/i386-builtin-types.def
6563 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
6564 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
6565 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
6566 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
6567 * config/i386/i386-builtin.def: Add corrresponding new builtins.
6568 * config/i386/i386-expand.c: Handle new builtin types.
6569 * config/i386/sse.md (VF48_128): New mode iterator.
6570 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
6572 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
6575 2021-09-17 liuhongt <hongtao.liu@intel.com>
6577 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
6579 (_mm512_mask_cvtph_pd): Likewise.
6580 (_mm512_maskz_cvtph_pd): Likewise.
6581 (_mm512_cvt_roundph_pd): Likewise.
6582 (_mm512_mask_cvt_roundph_pd): Likewise.
6583 (_mm512_maskz_cvt_roundph_pd): Likewise.
6584 (_mm512_cvtxph_ps): Likewise.
6585 (_mm512_mask_cvtxph_ps): Likewise.
6586 (_mm512_maskz_cvtxph_ps): Likewise.
6587 (_mm512_cvtx_roundph_ps): Likewise.
6588 (_mm512_mask_cvtx_roundph_ps): Likewise.
6589 (_mm512_maskz_cvtx_roundph_ps): Likewise.
6590 (_mm512_cvtxps_ph): Likewise.
6591 (_mm512_mask_cvtxps_ph): Likewise.
6592 (_mm512_maskz_cvtxps_ph): Likewise.
6593 (_mm512_cvtx_roundps_ph): Likewise.
6594 (_mm512_mask_cvtx_roundps_ph): Likewise.
6595 (_mm512_maskz_cvtx_roundps_ph): Likewise.
6596 (_mm512_cvtpd_ph): Likewise.
6597 (_mm512_mask_cvtpd_ph): Likewise.
6598 (_mm512_maskz_cvtpd_ph): Likewise.
6599 (_mm512_cvt_roundpd_ph): Likewise.
6600 (_mm512_mask_cvt_roundpd_ph): Likewise.
6601 (_mm512_maskz_cvt_roundpd_ph): Likewise.
6602 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
6604 (_mm_mask_cvtph_pd): Likewise.
6605 (_mm_maskz_cvtph_pd): Likewise.
6606 (_mm256_cvtph_pd): Likewise.
6607 (_mm256_mask_cvtph_pd): Likewise.
6608 (_mm256_maskz_cvtph_pd): Likewise.
6609 (_mm_cvtxph_ps): Likewise.
6610 (_mm_mask_cvtxph_ps): Likewise.
6611 (_mm_maskz_cvtxph_ps): Likewise.
6612 (_mm256_cvtxph_ps): Likewise.
6613 (_mm256_mask_cvtxph_ps): Likewise.
6614 (_mm256_maskz_cvtxph_ps): Likewise.
6615 (_mm_cvtxps_ph): Likewise.
6616 (_mm_mask_cvtxps_ph): Likewise.
6617 (_mm_maskz_cvtxps_ph): Likewise.
6618 (_mm256_cvtxps_ph): Likewise.
6619 (_mm256_mask_cvtxps_ph): Likewise.
6620 (_mm256_maskz_cvtxps_ph): Likewise.
6621 (_mm_cvtpd_ph): Likewise.
6622 (_mm_mask_cvtpd_ph): Likewise.
6623 (_mm_maskz_cvtpd_ph): Likewise.
6624 (_mm256_cvtpd_ph): Likewise.
6625 (_mm256_mask_cvtpd_ph): Likewise.
6626 (_mm256_maskz_cvtpd_ph): Likewise.
6627 * config/i386/i386-builtin.def: Add corresponding new builtins.
6628 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6629 * config/i386/i386-expand.c: Handle new builtin types.
6630 * config/i386/sse.md
6631 (VF4_128_8_256): New.
6632 (VF48H_AVX512VL): Ditto.
6633 (ssePHmode): Add HF vector modes.
6634 (castmode): Add new convertable modes.
6637 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
6638 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
6639 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
6640 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
6641 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
6642 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
6643 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
6645 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
6646 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
6647 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
6648 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
6650 2021-09-17 liuhongt <hongtao.liu@intel.com>
6652 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
6654 (_mm_cvttsh_u32): Likewise.
6655 (_mm_cvtt_roundsh_i32): Likewise.
6656 (_mm_cvtt_roundsh_u32): Likewise.
6657 (_mm_cvttsh_i64): Likewise.
6658 (_mm_cvttsh_u64): Likewise.
6659 (_mm_cvtt_roundsh_i64): Likewise.
6660 (_mm_cvtt_roundsh_u64): Likewise.
6661 * config/i386/i386-builtin.def: Add corresponding new builtins.
6662 * config/i386/sse.md
6663 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
6666 2021-09-17 liuhongt <hongtao.liu@intel.com>
6668 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
6670 (_mm512_mask_cvttph_epi32): Likewise.
6671 (_mm512_maskz_cvttph_epi32): Likewise.
6672 (_mm512_cvtt_roundph_epi32): Likewise.
6673 (_mm512_mask_cvtt_roundph_epi32): Likewise.
6674 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
6675 (_mm512_cvttph_epu32): Likewise.
6676 (_mm512_mask_cvttph_epu32): Likewise.
6677 (_mm512_maskz_cvttph_epu32): Likewise.
6678 (_mm512_cvtt_roundph_epu32): Likewise.
6679 (_mm512_mask_cvtt_roundph_epu32): Likewise.
6680 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
6681 (_mm512_cvttph_epi64): Likewise.
6682 (_mm512_mask_cvttph_epi64): Likewise.
6683 (_mm512_maskz_cvttph_epi64): Likewise.
6684 (_mm512_cvtt_roundph_epi64): Likewise.
6685 (_mm512_mask_cvtt_roundph_epi64): Likewise.
6686 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
6687 (_mm512_cvttph_epu64): Likewise.
6688 (_mm512_mask_cvttph_epu64): Likewise.
6689 (_mm512_maskz_cvttph_epu64): Likewise.
6690 (_mm512_cvtt_roundph_epu64): Likewise.
6691 (_mm512_mask_cvtt_roundph_epu64): Likewise.
6692 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
6693 (_mm512_cvttph_epi16): Likewise.
6694 (_mm512_mask_cvttph_epi16): Likewise.
6695 (_mm512_maskz_cvttph_epi16): Likewise.
6696 (_mm512_cvtt_roundph_epi16): Likewise.
6697 (_mm512_mask_cvtt_roundph_epi16): Likewise.
6698 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
6699 (_mm512_cvttph_epu16): Likewise.
6700 (_mm512_mask_cvttph_epu16): Likewise.
6701 (_mm512_maskz_cvttph_epu16): Likewise.
6702 (_mm512_cvtt_roundph_epu16): Likewise.
6703 (_mm512_mask_cvtt_roundph_epu16): Likewise.
6704 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
6705 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
6707 (_mm_mask_cvttph_epi32): Likewise.
6708 (_mm_maskz_cvttph_epi32): Likewise.
6709 (_mm256_cvttph_epi32): Likewise.
6710 (_mm256_mask_cvttph_epi32): Likewise.
6711 (_mm256_maskz_cvttph_epi32): Likewise.
6712 (_mm_cvttph_epu32): Likewise.
6713 (_mm_mask_cvttph_epu32): Likewise.
6714 (_mm_maskz_cvttph_epu32): Likewise.
6715 (_mm256_cvttph_epu32): Likewise.
6716 (_mm256_mask_cvttph_epu32): Likewise.
6717 (_mm256_maskz_cvttph_epu32): Likewise.
6718 (_mm_cvttph_epi64): Likewise.
6719 (_mm_mask_cvttph_epi64): Likewise.
6720 (_mm_maskz_cvttph_epi64): Likewise.
6721 (_mm256_cvttph_epi64): Likewise.
6722 (_mm256_mask_cvttph_epi64): Likewise.
6723 (_mm256_maskz_cvttph_epi64): Likewise.
6724 (_mm_cvttph_epu64): Likewise.
6725 (_mm_mask_cvttph_epu64): Likewise.
6726 (_mm_maskz_cvttph_epu64): Likewise.
6727 (_mm256_cvttph_epu64): Likewise.
6728 (_mm256_mask_cvttph_epu64): Likewise.
6729 (_mm256_maskz_cvttph_epu64): Likewise.
6730 (_mm_cvttph_epi16): Likewise.
6731 (_mm_mask_cvttph_epi16): Likewise.
6732 (_mm_maskz_cvttph_epi16): Likewise.
6733 (_mm256_cvttph_epi16): Likewise.
6734 (_mm256_mask_cvttph_epi16): Likewise.
6735 (_mm256_maskz_cvttph_epi16): Likewise.
6736 (_mm_cvttph_epu16): Likewise.
6737 (_mm_mask_cvttph_epu16): Likewise.
6738 (_mm_maskz_cvttph_epu16): Likewise.
6739 (_mm256_cvttph_epu16): Likewise.
6740 (_mm256_mask_cvttph_epu16): Likewise.
6741 (_mm256_maskz_cvttph_epu16): Likewise.
6742 * config/i386/i386-builtin.def: Add new builtins.
6743 * config/i386/sse.md
6744 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
6746 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
6747 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
6748 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
6749 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
6751 2021-09-17 liuhongt <hongtao.liu@intel.com>
6753 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
6754 (_mm_cvtsh_u32): Likewise.
6755 (_mm_cvt_roundsh_i32): Likewise.
6756 (_mm_cvt_roundsh_u32): Likewise.
6757 (_mm_cvtsh_i64): Likewise.
6758 (_mm_cvtsh_u64): Likewise.
6759 (_mm_cvt_roundsh_i64): Likewise.
6760 (_mm_cvt_roundsh_u64): Likewise.
6761 (_mm_cvti32_sh): Likewise.
6762 (_mm_cvtu32_sh): Likewise.
6763 (_mm_cvt_roundi32_sh): Likewise.
6764 (_mm_cvt_roundu32_sh): Likewise.
6765 (_mm_cvti64_sh): Likewise.
6766 (_mm_cvtu64_sh): Likewise.
6767 (_mm_cvt_roundi64_sh): Likewise.
6768 (_mm_cvt_roundu64_sh): Likewise.
6769 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6770 * config/i386/i386-builtin.def: Add corresponding new builtins.
6771 * config/i386/i386-expand.c (ix86_expand_round_builtin):
6772 Handle new builtin types.
6773 * config/i386/sse.md
6774 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
6776 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
6777 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
6779 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
6781 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
6782 (altivec_resolve_new_overloaded_builtin): New forward decl.
6783 (rs6000_new_builtin_type_compatible): New function.
6784 (altivec_resolve_overloaded_builtin): Call
6785 altivec_resolve_new_overloaded_builtin.
6786 (altivec_build_new_resolved_builtin): New function.
6787 (altivec_resolve_new_overloaded_builtin): Likewise.
6788 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
6790 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
6791 name of rs6000_new_builtin_is_supported.
6793 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
6795 * config/i386/i386-protos.h (ix86_decompose_address):
6796 Change return type to bool.
6797 * config/i386/i386.c (ix86_decompose_address): Ditto.
6799 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
6802 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
6803 Added 'build/' to target, use build/%.o rule.
6804 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
6805 '$(build_exeext)' to target and 'build/' for the *.o files.
6806 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
6809 2021-09-16 Martin Jambor <mjambor@suse.cz>
6811 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
6812 the body has been removed. Remove trailing whitespace.
6814 2021-09-16 Richard Biener <rguenther@suse.de>
6816 PR middle-end/102360
6817 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
6818 of non-memory more robust.
6820 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6822 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
6823 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
6824 (leon5_adjust_cost): Increase cost of store with data dependency
6825 on ALU instruction and FPU anti-dependencies.
6826 (sparc_option_override): Add LEON5 costs
6827 (sparc_adjust_cost): Add LEON5 cost adjustments
6828 * config/sparc/sparc.h: Add LEON5
6829 * config/sparc/sparc.md: Include LEON5 scheduling information
6830 * config/sparc/sparc.opt: Add LEON5
6831 * doc/invoke.texi: Add LEON5
6832 * config/sparc/leon5.md: New file.
6834 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6836 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
6837 sensitive sequence for B2BST errata workaround.
6839 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6841 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
6842 functions with atomic instruction in the UT700 errata workaround.
6844 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6846 * config/sparc/sparc.c (next_active_non_empty_insn): New function
6847 that returns next active non empty assembly instruction.
6848 (sparc_do_work_around_errata): Use new function.
6850 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6852 * config/sparc/sparc.c (store_insn_p): Add predicate for store
6854 (load_insn_p): Add predicate for load attributes.
6855 (sparc_do_work_around_errata): Use new predicates.
6857 2021-09-16 Andreas Larsson <andreas@gaisler.com>
6859 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
6862 2021-09-16 Martin Liska <mliska@suse.cz>
6864 * config/mips/netbsd.h: Fix typo in name of a macro.
6866 2021-09-16 liuhongt <hongtao.liu@intel.com>
6868 PR middle-end/102080
6869 * match.pd: Check mask type when doing cond_op related gimple
6871 * tree.c (is_truth_type_for): New function.
6872 * tree.h (is_truth_type_for): New declaration.
6874 2021-09-16 liuhongt <hongtao.liu@intel.com>
6876 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
6878 (_mm512_mask_cvtepi32_ph): Likewise.
6879 (_mm512_maskz_cvtepi32_ph): Likewise.
6880 (_mm512_cvt_roundepi32_ph): Likewise.
6881 (_mm512_mask_cvt_roundepi32_ph): Likewise.
6882 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
6883 (_mm512_cvtepu32_ph): Likewise.
6884 (_mm512_mask_cvtepu32_ph): Likewise.
6885 (_mm512_maskz_cvtepu32_ph): Likewise.
6886 (_mm512_cvt_roundepu32_ph): Likewise.
6887 (_mm512_mask_cvt_roundepu32_ph): Likewise.
6888 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
6889 (_mm512_cvtepi64_ph): Likewise.
6890 (_mm512_mask_cvtepi64_ph): Likewise.
6891 (_mm512_maskz_cvtepi64_ph): Likewise.
6892 (_mm512_cvt_roundepi64_ph): Likewise.
6893 (_mm512_mask_cvt_roundepi64_ph): Likewise.
6894 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
6895 (_mm512_cvtepu64_ph): Likewise.
6896 (_mm512_mask_cvtepu64_ph): Likewise.
6897 (_mm512_maskz_cvtepu64_ph): Likewise.
6898 (_mm512_cvt_roundepu64_ph): Likewise.
6899 (_mm512_mask_cvt_roundepu64_ph): Likewise.
6900 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
6901 (_mm512_cvtepi16_ph): Likewise.
6902 (_mm512_mask_cvtepi16_ph): Likewise.
6903 (_mm512_maskz_cvtepi16_ph): Likewise.
6904 (_mm512_cvt_roundepi16_ph): Likewise.
6905 (_mm512_mask_cvt_roundepi16_ph): Likewise.
6906 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
6907 (_mm512_cvtepu16_ph): Likewise.
6908 (_mm512_mask_cvtepu16_ph): Likewise.
6909 (_mm512_maskz_cvtepu16_ph): Likewise.
6910 (_mm512_cvt_roundepu16_ph): Likewise.
6911 (_mm512_mask_cvt_roundepu16_ph): Likewise.
6912 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
6913 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
6915 (_mm_mask_cvtepi32_ph): Likewise.
6916 (_mm_maskz_cvtepi32_ph): Likewise.
6917 (_mm256_cvtepi32_ph): Likewise.
6918 (_mm256_mask_cvtepi32_ph): Likewise.
6919 (_mm256_maskz_cvtepi32_ph): Likewise.
6920 (_mm_cvtepu32_ph): Likewise.
6921 (_mm_mask_cvtepu32_ph): Likewise.
6922 (_mm_maskz_cvtepu32_ph): Likewise.
6923 (_mm256_cvtepu32_ph): Likewise.
6924 (_mm256_mask_cvtepu32_ph): Likewise.
6925 (_mm256_maskz_cvtepu32_ph): Likewise.
6926 (_mm_cvtepi64_ph): Likewise.
6927 (_mm_mask_cvtepi64_ph): Likewise.
6928 (_mm_maskz_cvtepi64_ph): Likewise.
6929 (_mm256_cvtepi64_ph): Likewise.
6930 (_mm256_mask_cvtepi64_ph): Likewise.
6931 (_mm256_maskz_cvtepi64_ph): Likewise.
6932 (_mm_cvtepu64_ph): Likewise.
6933 (_mm_mask_cvtepu64_ph): Likewise.
6934 (_mm_maskz_cvtepu64_ph): Likewise.
6935 (_mm256_cvtepu64_ph): Likewise.
6936 (_mm256_mask_cvtepu64_ph): Likewise.
6937 (_mm256_maskz_cvtepu64_ph): Likewise.
6938 (_mm_cvtepi16_ph): Likewise.
6939 (_mm_mask_cvtepi16_ph): Likewise.
6940 (_mm_maskz_cvtepi16_ph): Likewise.
6941 (_mm256_cvtepi16_ph): Likewise.
6942 (_mm256_mask_cvtepi16_ph): Likewise.
6943 (_mm256_maskz_cvtepi16_ph): Likewise.
6944 (_mm_cvtepu16_ph): Likewise.
6945 (_mm_mask_cvtepu16_ph): Likewise.
6946 (_mm_maskz_cvtepu16_ph): Likewise.
6947 (_mm256_cvtepu16_ph): Likewise.
6948 (_mm256_mask_cvtepu16_ph): Likewise.
6949 (_mm256_maskz_cvtepu16_ph): Likewise.
6950 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6951 * config/i386/i386-builtin.def: Add corresponding new builtins.
6952 * config/i386/i386-expand.c
6953 (ix86_expand_args_builtin): Handle new builtin types.
6954 (ix86_expand_round_builtin): Ditto.
6955 * config/i386/i386-modes.def: Declare V2HF and V6HF.
6956 * config/i386/sse.md (VI2H_AVX512VL): New.
6958 (sseintvecmode): Add HF vector modes.
6959 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
6961 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
6962 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
6963 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
6964 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
6965 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
6966 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
6967 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
6968 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
6969 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
6970 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
6971 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
6973 2021-09-16 liuhongt <hongtao.liu@intel.com>
6975 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
6977 (_mm512_mask_cvtph_epi32): Likewise.
6978 (_mm512_maskz_cvtph_epi32): Likewise.
6979 (_mm512_cvt_roundph_epi32): Likewise.
6980 (_mm512_mask_cvt_roundph_epi32): Likewise.
6981 (_mm512_maskz_cvt_roundph_epi32): Likewise.
6982 (_mm512_cvtph_epu32): Likewise.
6983 (_mm512_mask_cvtph_epu32): Likewise.
6984 (_mm512_maskz_cvtph_epu32): Likewise.
6985 (_mm512_cvt_roundph_epu32): Likewise.
6986 (_mm512_mask_cvt_roundph_epu32): Likewise.
6987 (_mm512_maskz_cvt_roundph_epu32): Likewise.
6988 (_mm512_cvtph_epi64): Likewise.
6989 (_mm512_mask_cvtph_epi64): Likewise.
6990 (_mm512_maskz_cvtph_epi64): Likewise.
6991 (_mm512_cvt_roundph_epi64): Likewise.
6992 (_mm512_mask_cvt_roundph_epi64): Likewise.
6993 (_mm512_maskz_cvt_roundph_epi64): Likewise.
6994 (_mm512_cvtph_epu64): Likewise.
6995 (_mm512_mask_cvtph_epu64): Likewise.
6996 (_mm512_maskz_cvtph_epu64): Likewise.
6997 (_mm512_cvt_roundph_epu64): Likewise.
6998 (_mm512_mask_cvt_roundph_epu64): Likewise.
6999 (_mm512_maskz_cvt_roundph_epu64): Likewise.
7000 (_mm512_cvtph_epi16): Likewise.
7001 (_mm512_mask_cvtph_epi16): Likewise.
7002 (_mm512_maskz_cvtph_epi16): Likewise.
7003 (_mm512_cvt_roundph_epi16): Likewise.
7004 (_mm512_mask_cvt_roundph_epi16): Likewise.
7005 (_mm512_maskz_cvt_roundph_epi16): Likewise.
7006 (_mm512_cvtph_epu16): Likewise.
7007 (_mm512_mask_cvtph_epu16): Likewise.
7008 (_mm512_maskz_cvtph_epu16): Likewise.
7009 (_mm512_cvt_roundph_epu16): Likewise.
7010 (_mm512_mask_cvt_roundph_epu16): Likewise.
7011 (_mm512_maskz_cvt_roundph_epu16): Likewise.
7012 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
7014 (_mm_mask_cvtph_epi32): Likewise.
7015 (_mm_maskz_cvtph_epi32): Likewise.
7016 (_mm256_cvtph_epi32): Likewise.
7017 (_mm256_mask_cvtph_epi32): Likewise.
7018 (_mm256_maskz_cvtph_epi32): Likewise.
7019 (_mm_cvtph_epu32): Likewise.
7020 (_mm_mask_cvtph_epu32): Likewise.
7021 (_mm_maskz_cvtph_epu32): Likewise.
7022 (_mm256_cvtph_epu32): Likewise.
7023 (_mm256_mask_cvtph_epu32): Likewise.
7024 (_mm256_maskz_cvtph_epu32): Likewise.
7025 (_mm_cvtph_epi64): Likewise.
7026 (_mm_mask_cvtph_epi64): Likewise.
7027 (_mm_maskz_cvtph_epi64): Likewise.
7028 (_mm256_cvtph_epi64): Likewise.
7029 (_mm256_mask_cvtph_epi64): Likewise.
7030 (_mm256_maskz_cvtph_epi64): Likewise.
7031 (_mm_cvtph_epu64): Likewise.
7032 (_mm_mask_cvtph_epu64): Likewise.
7033 (_mm_maskz_cvtph_epu64): Likewise.
7034 (_mm256_cvtph_epu64): Likewise.
7035 (_mm256_mask_cvtph_epu64): Likewise.
7036 (_mm256_maskz_cvtph_epu64): Likewise.
7037 (_mm_cvtph_epi16): Likewise.
7038 (_mm_mask_cvtph_epi16): Likewise.
7039 (_mm_maskz_cvtph_epi16): Likewise.
7040 (_mm256_cvtph_epi16): Likewise.
7041 (_mm256_mask_cvtph_epi16): Likewise.
7042 (_mm256_maskz_cvtph_epi16): Likewise.
7043 (_mm_cvtph_epu16): Likewise.
7044 (_mm_mask_cvtph_epu16): Likewise.
7045 (_mm_maskz_cvtph_epu16): Likewise.
7046 (_mm256_cvtph_epu16): Likewise.
7047 (_mm256_mask_cvtph_epu16): Likewise.
7048 (_mm256_maskz_cvtph_epu16): Likewise.
7049 * config/i386/i386-builtin-types.def: Add new builtin types.
7050 * config/i386/i386-builtin.def: Add new builtins.
7051 * config/i386/i386-expand.c
7052 (ix86_expand_args_builtin): Handle new builtin types.
7053 (ix86_expand_round_builtin): Ditto.
7054 * config/i386/sse.md (sseintconvert): New.
7056 (UNSPEC_US_FIX_NOTRUNC): Ditto.
7057 (sseintconvertsignprefix): Ditto.
7058 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
7061 2021-09-16 liuhongt <hongtao.liu@intel.com>
7063 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
7065 (_mm_cvtsi128_si16): Likewise.
7066 (_mm_mask_load_sh): Likewise.
7067 (_mm_maskz_load_sh): Likewise.
7068 (_mm_mask_store_sh): Likewise.
7069 (_mm_move_sh): Likewise.
7070 (_mm_mask_move_sh): Likewise.
7071 (_mm_maskz_move_sh): Likewise.
7072 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7073 * config/i386/i386-builtin.def: Add corresponding new builtins.
7074 * config/i386/i386-expand.c
7075 (ix86_expand_special_args_builtin): Handle new builtin types.
7076 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
7077 * config/i386/sse.md (VI2F): New mode iterator.
7078 (vec_set<mode>_0): Use new mode iterator.
7079 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
7080 (avx512f_store<mode>_mask): Ditto.
7082 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
7084 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
7086 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
7088 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
7089 Proceed if no symbol summary or the symbol alias flag is false.
7091 2021-09-15 Jakub Jelinek <jakub@redhat.com>
7095 * varasm.c (output_constructor_regular_field): Instead of assertion
7096 that array_size_for_constructor result is equal to size of
7097 TREE_TYPE (local->val) in bytes, assert that the type size is greater
7098 or equal to array_size_for_constructor result and use type size as
7101 2021-09-15 Martin Liska <mliska@suse.cz>
7104 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
7106 2021-09-15 Martin Liska <mliska@suse.cz>
7109 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
7110 Check that we have a symbol summary for a symbol.
7112 2021-09-15 Richard Biener <rguenther@suse.de>
7115 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
7116 to inherit from elfos.h
7118 2021-09-15 liuhongt <hongtao.liu@intel.com>
7121 * config/i386/i386-expand.c
7122 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
7124 (ix86_expand_vector_set): Use blendw instead of pinsrw.
7125 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
7126 AVX512FP16 which supports 16bit vector load.
7127 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
7129 (avx512bw_interleave_high<mode><mask_name>): .. this, and
7130 extend to V32HFmode.
7131 (avx2_interleave_highv16hi<mask_name>): Rename to ..
7132 (avx2_interleave_high<mode><mask_name>): .. this, and extend
7134 (vec_interleave_highv8hi<mask_name>): Rename to ..
7135 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
7136 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
7138 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
7139 this, and extend to V32HFmode.
7140 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
7141 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
7142 (vec_interleave_lowv8hi<mask_name>): Rename to ..
7143 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
7144 (sse4_1_pblendw): Rename to ..
7145 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
7146 (avx2_pblendph): New define_expand.
7147 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
7148 sseintmodesuffix instead of ssemodesuffix.
7149 (blendsuf): New mode attr.
7151 2021-09-15 Richard Biener <rguenther@suse.de>
7153 * tree-vectorizer.h (dr_misalignment): Move out of line.
7154 (dr_target_alignment): New.
7155 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
7156 (set_dr_target_alignment): New.
7157 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
7158 * tree-vect-data-refs.c (dr_misalignment): Compute and
7159 return the group members misalignment.
7160 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
7161 (vect_analyze_data_refs_alignment): Compute alignment only
7162 for the first element of a DR group.
7163 (vect_slp_analyze_node_alignment): Likewise.
7165 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
7167 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
7168 * config/i386/avx512fp16vlintrin.h: Likewise.
7169 * config/i386/i386-builtin.def: Adjust builtin name and
7170 enumeration to match AVX512F style.
7172 2021-09-15 Richard Biener <rguenther@suse.de>
7174 PR tree-optimization/102318
7175 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
7176 previous change and do the mode conversion separately from
7177 the sign conversion.
7179 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
7180 Peter Cordes <peter@cordes.ca>
7183 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
7184 (*vec_extract<mode><ssescalarmodelower>_valign): Output
7185 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
7188 2021-09-15 Richard Biener <rguenther@suse.de>
7190 * config.gcc: Remove vax-*-openbsd* configuration.
7192 2021-09-15 Richard Biener <rguenther@suse.de>
7194 * config.gcc: Remove m68k-openbsd.
7196 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
7199 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
7201 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
7203 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
7204 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
7205 (*mma_xxsetaccz): Delete.
7206 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
7207 Use UNSPECV_MMA_XXSETACCZ. Update comment.
7208 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
7210 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
7212 * Makefile.in: Remove variables related to applying no-PIE
7213 to the exes on $build.
7214 * configure: Regenerate.
7215 * configure.ac: Remove configuration related to applying
7216 no-PIE to the exes on $build.
7218 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
7220 * config/arc/arc.md (doloop_end): Add missing mode.
7221 (loop_end): Likewise.
7223 2021-09-14 Jakub Jelinek <jakub@redhat.com>
7225 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
7226 it to recursive calls, for depth above 7 just gimplify or return.
7227 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
7228 __builtin_clear_padding and TARGET_EXPR.
7229 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
7231 2021-09-14 liuhongt <hongtao.liu@intel.com>
7233 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
7235 (_mm_mask_fpclass_sh_mask): Likewise.
7236 (_mm512_mask_fpclass_ph_mask): Likewise.
7237 (_mm512_fpclass_ph_mask): Likewise.
7238 (_mm_getexp_sh): Likewise.
7239 (_mm_mask_getexp_sh): Likewise.
7240 (_mm_maskz_getexp_sh): Likewise.
7241 (_mm512_getexp_ph): Likewise.
7242 (_mm512_mask_getexp_ph): Likewise.
7243 (_mm512_maskz_getexp_ph): Likewise.
7244 (_mm_getexp_round_sh): Likewise.
7245 (_mm_mask_getexp_round_sh): Likewise.
7246 (_mm_maskz_getexp_round_sh): Likewise.
7247 (_mm512_getexp_round_ph): Likewise.
7248 (_mm512_mask_getexp_round_ph): Likewise.
7249 (_mm512_maskz_getexp_round_ph): Likewise.
7250 (_mm_getmant_sh): Likewise.
7251 (_mm_mask_getmant_sh): Likewise.
7252 (_mm_maskz_getmant_sh): Likewise.
7253 (_mm512_getmant_ph): Likewise.
7254 (_mm512_mask_getmant_ph): Likewise.
7255 (_mm512_maskz_getmant_ph): Likewise.
7256 (_mm_getmant_round_sh): Likewise.
7257 (_mm_mask_getmant_round_sh): Likewise.
7258 (_mm_maskz_getmant_round_sh): Likewise.
7259 (_mm512_getmant_round_ph): Likewise.
7260 (_mm512_mask_getmant_round_ph): Likewise.
7261 (_mm512_maskz_getmant_round_ph): Likewise.
7262 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
7264 (_mm_fpclass_ph_mask): Likewise.
7265 (_mm256_mask_fpclass_ph_mask): Likewise.
7266 (_mm256_fpclass_ph_mask): Likewise.
7267 (_mm256_getexp_ph): Likewise.
7268 (_mm256_mask_getexp_ph): Likewise.
7269 (_mm256_maskz_getexp_ph): Likewise.
7270 (_mm_getexp_ph): Likewise.
7271 (_mm_mask_getexp_ph): Likewise.
7272 (_mm_maskz_getexp_ph): Likewise.
7273 (_mm256_getmant_ph): Likewise.
7274 (_mm256_mask_getmant_ph): Likewise.
7275 (_mm256_maskz_getmant_ph): Likewise.
7276 (_mm_getmant_ph): Likewise.
7277 (_mm_mask_getmant_ph): Likewise.
7278 (_mm_maskz_getmant_ph): Likewise.
7279 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7280 * config/i386/i386-builtin.def: Add corresponding new builtins.
7281 * config/i386/i386-expand.c
7282 (ix86_expand_args_builtin): Handle new builtin types.
7283 (ix86_expand_round_builtin): Ditto.
7284 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
7285 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
7286 to support HF vector modes.
7287 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
7289 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
7290 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
7291 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
7292 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7295 2021-09-14 liuhongt <hongtao.liu@intel.com>
7297 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
7299 (_mm512_mask_reduce_ph): Likewise.
7300 (_mm512_maskz_reduce_ph): Likewise.
7301 (_mm512_reduce_round_ph): Likewise.
7302 (_mm512_mask_reduce_round_ph): Likewise.
7303 (_mm512_maskz_reduce_round_ph): Likewise.
7304 (_mm_reduce_sh): Likewise.
7305 (_mm_mask_reduce_sh): Likewise.
7306 (_mm_maskz_reduce_sh): Likewise.
7307 (_mm_reduce_round_sh): Likewise.
7308 (_mm_mask_reduce_round_sh): Likewise.
7309 (_mm_maskz_reduce_round_sh): Likewise.
7310 (_mm512_roundscale_ph): Likewise.
7311 (_mm512_mask_roundscale_ph): Likewise.
7312 (_mm512_maskz_roundscale_ph): Likewise.
7313 (_mm512_roundscale_round_ph): Likewise.
7314 (_mm512_mask_roundscale_round_ph): Likewise.
7315 (_mm512_maskz_roundscale_round_ph): Likewise.
7316 (_mm_roundscale_sh): Likewise.
7317 (_mm_mask_roundscale_sh): Likewise.
7318 (_mm_maskz_roundscale_sh): Likewise.
7319 (_mm_roundscale_round_sh): Likewise.
7320 (_mm_mask_roundscale_round_sh): Likewise.
7321 (_mm_maskz_roundscale_round_sh): Likewise.
7322 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
7324 (_mm_mask_reduce_ph): Likewise.
7325 (_mm_maskz_reduce_ph): Likewise.
7326 (_mm256_reduce_ph): Likewise.
7327 (_mm256_mask_reduce_ph): Likewise.
7328 (_mm256_maskz_reduce_ph): Likewise.
7329 (_mm_roundscale_ph): Likewise.
7330 (_mm_mask_roundscale_ph): Likewise.
7331 (_mm_maskz_roundscale_ph): Likewise.
7332 (_mm256_roundscale_ph): Likewise.
7333 (_mm256_mask_roundscale_ph): Likewise.
7334 (_mm256_maskz_roundscale_ph): Likewise.
7335 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7336 * config/i386/i386-builtin.def: Add corresponding new builtins.
7337 * config/i386/i386-expand.c
7338 (ix86_expand_args_builtin): Handle new builtin types.
7339 (ix86_expand_round_builtin): Ditto.
7340 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
7342 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
7343 ... this, and adjust for round operands.
7344 (reduces<mode><mask_scalar_name>): Likewise, with ...
7345 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
7347 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
7348 Adjust for HF vector modes.
7349 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
7351 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
7353 2021-09-14 liuhongt <hongtao.liu@intel.com>
7355 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
7357 (_mm512_mask_rcp_ph): Likewise.
7358 (_mm512_maskz_rcp_ph): Likewise.
7359 (_mm_rcp_sh): Likewise.
7360 (_mm_mask_rcp_sh): Likewise.
7361 (_mm_maskz_rcp_sh): Likewise.
7362 (_mm512_scalef_ph): Likewise.
7363 (_mm512_mask_scalef_ph): Likewise.
7364 (_mm512_maskz_scalef_ph): Likewise.
7365 (_mm512_scalef_round_ph): Likewise.
7366 (_mm512_mask_scalef_round_ph): Likewise.
7367 (_mm512_maskz_scalef_round_ph): Likewise.
7368 (_mm_scalef_sh): Likewise.
7369 (_mm_mask_scalef_sh): Likewise.
7370 (_mm_maskz_scalef_sh): Likewise.
7371 (_mm_scalef_round_sh): Likewise.
7372 (_mm_mask_scalef_round_sh): Likewise.
7373 (_mm_maskz_scalef_round_sh): Likewise.
7374 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
7376 (_mm256_rcp_ph): Likewise.
7377 (_mm_mask_rcp_ph): Likewise.
7378 (_mm256_mask_rcp_ph): Likewise.
7379 (_mm_maskz_rcp_ph): Likewise.
7380 (_mm256_maskz_rcp_ph): Likewise.
7381 (_mm_scalef_ph): Likewise.
7382 (_mm256_scalef_ph): Likewise.
7383 (_mm_mask_scalef_ph): Likewise.
7384 (_mm256_mask_scalef_ph): Likewise.
7385 (_mm_maskz_scalef_ph): Likewise.
7386 (_mm256_maskz_scalef_ph): Likewise.
7387 * config/i386/i386-builtin.def: Add new builtins.
7388 * config/i386/sse.md (VFH_AVX512VL): New.
7389 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
7390 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
7391 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
7392 Adjust to support HF vector modes.
7393 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
7395 2021-09-14 liuhongt <hongtao.liu@intel.com>
7397 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
7399 (_mm512_mask_sqrt_ph): Likewise.
7400 (_mm512_maskz_sqrt_ph): Likewise.
7401 (_mm512_sqrt_round_ph): Likewise.
7402 (_mm512_mask_sqrt_round_ph): Likewise.
7403 (_mm512_maskz_sqrt_round_ph): Likewise.
7404 (_mm512_rsqrt_ph): Likewise.
7405 (_mm512_mask_rsqrt_ph): Likewise.
7406 (_mm512_maskz_rsqrt_ph): Likewise.
7407 (_mm_rsqrt_sh): Likewise.
7408 (_mm_mask_rsqrt_sh): Likewise.
7409 (_mm_maskz_rsqrt_sh): Likewise.
7410 (_mm_sqrt_sh): Likewise.
7411 (_mm_mask_sqrt_sh): Likewise.
7412 (_mm_maskz_sqrt_sh): Likewise.
7413 (_mm_sqrt_round_sh): Likewise.
7414 (_mm_mask_sqrt_round_sh): Likewise.
7415 (_mm_maskz_sqrt_round_sh): Likewise.
7416 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
7417 (_mm256_sqrt_ph): Likewise.
7418 (_mm_mask_sqrt_ph): Likewise.
7419 (_mm256_mask_sqrt_ph): Likewise.
7420 (_mm_maskz_sqrt_ph): Likewise.
7421 (_mm256_maskz_sqrt_ph): Likewise.
7422 (_mm_rsqrt_ph): Likewise.
7423 (_mm256_rsqrt_ph): Likewise.
7424 (_mm_mask_rsqrt_ph): Likewise.
7425 (_mm256_mask_rsqrt_ph): Likewise.
7426 (_mm_maskz_rsqrt_ph): Likewise.
7427 (_mm256_maskz_rsqrt_ph): Likewise.
7428 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7429 * config/i386/i386-builtin.def: Add corresponding new builtins.
7430 * config/i386/i386-expand.c
7431 (ix86_expand_args_builtin): Handle new builtins.
7432 (ix86_expand_round_builtin): Ditto.
7433 * config/i386/sse.md (VF_AVX512FP16VL): New.
7434 (sqrt<mode>2): Adjust for HF vector modes.
7435 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
7436 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
7438 (<sse>_rsqrt<mode>2<mask_name>): New.
7439 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
7441 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7444 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
7445 'RESERVED_LOCATION_P' locations.
7446 * warning-control.cc (get_nowarn_spec, suppress_warning)
7447 (copy_warning): Likewise.
7449 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7451 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
7452 (typedef key_type_t): ... this. Remove.
7453 (nowarn_map): Document.
7454 * diagnostic-spec.c (nowarn_map): Likewise.
7455 * warning-control.cc (convert_to_key): Evolve functions into...
7456 (get_location): ... these. Adjust all users.
7458 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7460 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
7462 2021-09-13 Jason Merrill <jason@redhat.com>
7464 * params.opt: Add destructive-interference-size and
7465 constructive-interference-size.
7466 * doc/invoke.texi: Document them.
7467 * config/aarch64/aarch64.c (aarch64_override_options_internal):
7469 * config/arm/arm.c (arm_option_override): Set them.
7470 * config/i386/i386-options.c (ix86_option_override_internal):
7473 2021-09-13 Martin Liska <mliska@suse.cz>
7474 H.J. Lu <hjl.tools@gmail.com>
7477 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
7478 for x86-64 micro levels for __builtin_cpu_supports.
7479 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
7480 Add priorities for the micro-arch levels.
7481 (enum processor_features): Add new features.
7482 * common/config/i386/i386-isas.h: Add micro-arch features.
7483 * config/i386/i386-builtins.c (get_builtin_code_for_version):
7484 Support the micro-arch levels by callsing
7485 __builtin_cpu_supports.
7486 * doc/extend.texi: Document that the levels are support by
7487 __builtin_cpu_supports.
7489 2021-09-13 Andrew Pinski <apinski@marvell.com>
7492 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
7494 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
7495 (aarch64_general_gimple_fold_builtin): Likewise.
7497 2021-09-13 Andrew Pinski <apinski@marvell.com>
7499 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
7500 to the Unsupported targets list.
7501 Remove support for m32r-*-linux* and m32rle-*-linux*.
7502 * config/m32r/linux.h: Removed.
7503 * config/m32r/t-linux: Removed.
7505 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7508 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
7509 register index for SVE predicate modes.
7511 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7513 * tree-ssa-threadbackward.c
7514 (back_threader_profitability::profitable_path_p): Remove FSM
7516 (back_threader_registry::register_path): Same.
7517 * tree-ssa-threadedge.c
7518 (jump_threader::simplify_control_stmt_condition): Same.
7519 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
7520 Add backedge_threads argument.
7521 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
7522 backedge_threads argument.
7523 (back_jt_path_registry::back_jt_path_registry): Same.
7524 (dump_jump_thread_path): Adjust for FSM removal.
7525 (back_jt_path_registry::rewire_first_differing_edge): Same.
7526 (back_jt_path_registry::adjust_paths_after_duplication): Same.
7527 (back_jt_path_registry::update_cfg): Same.
7528 (jt_path_registry::register_jump_thread): Same.
7529 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
7531 (class back_jt_path_registry): Add backedge_threads to
7534 2021-09-13 Martin Liska <mliska@suse.cz>
7537 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
7539 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
7542 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
7543 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
7544 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
7545 TARGET_AVX256_STORE_BY_PIECES instead of
7546 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
7547 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7548 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
7549 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7550 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
7551 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
7553 2021-09-13 liuhongt <hongtao.liu@intel.com>
7556 * expmed.c (extract_bit_field_using_extv): Use
7557 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
7559 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7561 * Makefile.in (OBJS): Add value-pointer-equiv.o.
7562 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
7563 value-pointer-equiv.*.
7564 (ssa_equiv_stack::ssa_equiv_stack): Same.
7565 (ssa_equiv_stack::enter): Same.
7566 (ssa_equiv_stack::leave): Same.
7567 (ssa_equiv_stack::push_replacement): Same.
7568 (ssa_equiv_stack::get_replacement): Same.
7569 (is_pointer_ssa): Same.
7570 (class pointer_equiv_analyzer): Same.
7571 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7572 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
7573 (pointer_equiv_analyzer::set_global_equiv): Same.
7574 (pointer_equiv_analyzer::set_cond_equiv): Same.
7575 (pointer_equiv_analyzer::get_equiv): Same.
7576 (pointer_equiv_analyzer::enter): Same.
7577 (pointer_equiv_analyzer::leave): Same.
7578 (pointer_equiv_analyzer::get_equiv_expr): Same.
7579 (pta_valueize): Same.
7580 (pointer_equiv_analyzer::visit_stmt): Same.
7581 (pointer_equiv_analyzer::visit_edge): Same.
7582 (hybrid_folder::value_of_expr): Same.
7583 (hybrid_folder::value_on_edge): Same.
7584 * value-pointer-equiv.cc: New file.
7585 * value-pointer-equiv.h: New file.
7587 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7590 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
7591 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
7593 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7596 * config/arm/arm.md (movmisaligndi): New define_expand.
7597 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
7599 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7602 * emit-rtl.c (gen_highpart): Use adjust_address to handle
7603 MEM rather than calling simplify_gen_subreg.
7605 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
7607 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
7608 define into a block.
7610 2021-09-13 Richard Biener <rguenther@suse.de>
7612 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
7613 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
7614 change based on DARWIN_PREFER_DWARF not being defined.
7616 2021-09-13 Richard Biener <rguenther@suse.de>
7618 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
7619 to inherit from elfos.h
7621 2021-09-13 Richard Biener <rguenther@suse.de>
7623 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
7625 2021-09-13 Richard Biener <rguenther@suse.de>
7627 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
7628 override, pick up DWARF2_DEBUG define from elfos.h
7630 2021-09-13 Richard Biener <rguenther@suse.de>
7632 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
7635 2021-09-13 Richard Biener <rguenther@suse.de>
7637 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
7640 2021-09-13 Richard Biener <rguenther@suse.de>
7642 * config/i386/cygming.h: Always default to DWARF2 debugging.
7643 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
7645 * doc/install.texi: Document binutils 2.16 as minimum
7646 requirement for mingw.
7648 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
7650 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
7651 nstmts, nloads and extra_ctor_cost.
7652 (rs6000_density_test): Add load density related heuristics. Do
7653 extra costing on vector construction statements if need.
7654 (rs6000_init_cost): Init new members.
7655 (rs6000_update_target_cost_per_stmt): New function.
7656 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
7657 rs6000_update_target_cost_per_stmt and call it.
7659 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
7661 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
7662 (rs6000_init_cost): Adjust.
7664 2021-09-13 liuhongt <hongtao.liu@intel.com>
7666 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
7667 (UNSPEC_XORSIGN): Ditto.
7669 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
7671 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
7672 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
7675 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
7677 * tree-ssa-threadbackward.c (class back_threader_registry): Use
7678 back_jt_path_registry.
7679 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
7680 fwd_jt_path_registry.
7681 * tree-ssa-threadedge.h (class jump_threader): Same..
7682 * tree-ssa-threadupdate.c
7683 (jump_thread_path_registry::jump_thread_path_registry): Rename...
7684 (jt_path_registry::jt_path_registry): ...to this.
7685 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
7686 (jt_path_registry::~jt_path_registry): ...this.
7687 (fwd_jt_path_registry::fwd_jt_path_registry): New.
7688 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
7689 (jump_thread_path_registry::allocate_thread_edge): Rename...
7690 (jt_path_registry::allocate_thread_edge): ...to this.
7691 (jump_thread_path_registry::allocate_thread_path): Rename...
7692 (jt_path_registry::allocate_thread_path): ...to this.
7693 (jump_thread_path_registry::lookup_redirection_data): Rename...
7694 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
7695 (jump_thread_path_registry::thread_block_1): Rename...
7696 (fwd_jt_path_registry::thread_block_1): ...to this.
7697 (jump_thread_path_registry::thread_block): Rename...
7698 (fwd_jt_path_registry::thread_block): ...to this.
7699 (jt_path_registry::thread_through_loop_header): Rename...
7700 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
7701 (jump_thread_path_registry::mark_threaded_blocks): Rename...
7702 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
7703 (jump_thread_path_registry::debug_path): Rename...
7704 (jt_path_registry::debug_path): ...to this.
7705 (jump_thread_path_registry::dump): Rename...
7706 (jt_path_registry::debug): ...to this.
7707 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
7708 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
7709 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
7710 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
7711 (jump_thread_path_registry::duplicate_thread_path): Rename...
7712 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
7713 drop ill-formed candidates.
7714 (jump_thread_path_registry::remove_jump_threads_including): Rename...
7715 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
7716 (jt_path_registry::thread_through_all_blocks): New.
7717 (back_jt_path_registry::update_cfg): New.
7718 (fwd_jt_path_registry::update_cfg): New.
7719 (jump_thread_path_registry::register_jump_thread): Rename...
7720 (jt_path_registry::register_jump_thread): ...to this.
7721 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7723 (class jt_path_registry): ...here.
7724 (class fwd_jt_path_registry): New.
7725 (class back_jt_path_registry): New.
7727 2021-09-10 liuhongt <hongtao.liu@intel.com>
7730 2021-09-01 liuhongt <hongtao.liu@intel.com>
7732 * emit-rtl.c (validate_subreg): Get rid of all float-int
7735 2021-09-10 Jakub Jelinek <jakub@redhat.com>
7737 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
7738 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
7739 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
7740 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
7741 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
7742 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
7743 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
7744 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
7746 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
7747 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
7749 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
7751 * tree.h (OMP_ATOMIC_WEAK): Define.
7752 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
7753 fail memory order being encoded in the same enum and also print
7754 fail clause if present.
7755 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
7756 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
7757 handle pre_p == NULL case as a test mode that only returns value
7758 but doesn't change gimplify nor change anything otherwise, adjust
7759 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
7760 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
7761 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
7763 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
7764 COND_EXPR rhs. Set weak flag on gimple load/store for
7766 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
7767 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
7769 (expand_omp_atomic_cas): New function.
7770 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
7772 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
7773 using expand_omp_atomic_cas.
7775 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
7776 Michael Matz <matz@suse.de>
7778 * tree-pass.h (PROP_loop_opts_done): New.
7779 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
7780 Intersect with global range.
7781 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
7782 * tree-ssa-threadbackward.c
7783 (back_threader_profitability::profitable_path_p): Disable
7784 threading through latches until after loop optimizations have run.
7786 2021-09-10 David Faust <david.faust@oracle.com>
7788 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
7791 2021-09-10 David Faust <david.faust@oracle.com>
7793 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
7794 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
7795 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
7796 (bpf_expand_cbranch): New function.
7797 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
7799 (zero_extendsidi2): Only use mov32 instruction if it is available.
7800 (SIM mode iterator): Conditionalize support for SI mode.
7801 (JM mode iterator): New.
7802 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
7803 (*branch_on_di): Update name, use new JM iterator.
7804 * config/bpf/bpf.opt: (mjmpext): New option.
7808 (bpf_isa): New enum.
7810 2021-09-10 David Faust <david.faust@oracle.com>
7812 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
7813 for register-to-register extensions.
7814 (zero_extendqidi2): Likewise.
7816 2021-09-10 Richard Biener <rguenther@suse.de>
7818 PR middle-end/102273
7819 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
7821 2021-09-10 Richard Biener <rguenther@suse.de>
7823 PR middle-end/102269
7824 * gimplify.c (is_var_need_auto_init): Empty types do not need
7827 2021-09-10 Richard Biener <rguenther@suse.de>
7829 * configure.ac (--with-stabs): Remove.
7830 * configure: Regenerate.
7831 * doc/install.texi: Remove --with-stabs documentation.
7833 2021-09-10 liuhongt <hongtao.liu@intel.com>
7835 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
7837 (_mm512_mask_cmp_ph_mask): Likewise.
7838 (_mm512_cmp_round_ph_mask): Likewise.
7839 (_mm512_mask_cmp_round_ph_mask): Likewise.
7840 (_mm_cmp_sh_mask): Likewise.
7841 (_mm_mask_cmp_sh_mask): Likewise.
7842 (_mm_cmp_round_sh_mask): Likewise.
7843 (_mm_mask_cmp_round_sh_mask): Likewise.
7844 (_mm_comieq_sh): Likewise.
7845 (_mm_comilt_sh): Likewise.
7846 (_mm_comile_sh): Likewise.
7847 (_mm_comigt_sh): Likewise.
7848 (_mm_comige_sh): Likewise.
7849 (_mm_comineq_sh): Likewise.
7850 (_mm_ucomieq_sh): Likewise.
7851 (_mm_ucomilt_sh): Likewise.
7852 (_mm_ucomile_sh): Likewise.
7853 (_mm_ucomigt_sh): Likewise.
7854 (_mm_ucomige_sh): Likewise.
7855 (_mm_ucomineq_sh): Likewise.
7856 (_mm_comi_round_sh): Likewise.
7857 (_mm_comi_sh): Likewise.
7858 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
7859 (_mm_mask_cmp_ph_mask): Likewise.
7860 (_mm256_cmp_ph_mask): Likewise.
7861 (_mm256_mask_cmp_ph_mask): Likewise.
7862 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7863 * config/i386/i386-builtin.def: Add corresponding new builtins.
7864 * config/i386/i386-expand.c
7865 (ix86_expand_args_builtin): Handle new builtin types.
7866 (ix86_expand_round_builtin): Ditto.
7867 * config/i386/i386.md (ssevecmode): Add HF mode.
7868 (MODEFH): New mode iterator.
7869 * config/i386/sse.md
7870 (V48H_AVX512VL): New mode iterator to support HF vector modes.
7871 Ajdust corresponding description.
7872 (ssecmpintprefix): New.
7873 (VI12_AVX512VL): Adjust to support HF vector modes.
7874 (cmp_imm_predicate): Likewise.
7875 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
7877 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
7878 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
7879 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
7881 2021-09-10 liuhongt <hongtao.liu@intel.com>
7883 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
7884 (_mm512_mask_max_ph): Likewise.
7885 (_mm512_maskz_max_ph): Likewise.
7886 (_mm512_min_ph): Likewise.
7887 (_mm512_mask_min_ph): Likewise.
7888 (_mm512_maskz_min_ph): Likewise.
7889 (_mm512_max_round_ph): Likewise.
7890 (_mm512_mask_max_round_ph): Likewise.
7891 (_mm512_maskz_max_round_ph): Likewise.
7892 (_mm512_min_round_ph): Likewise.
7893 (_mm512_mask_min_round_ph): Likewise.
7894 (_mm512_maskz_min_round_ph): Likewise.
7895 (_mm_max_sh): Likewise.
7896 (_mm_mask_max_sh): Likewise.
7897 (_mm_maskz_max_sh): Likewise.
7898 (_mm_min_sh): Likewise.
7899 (_mm_mask_min_sh): Likewise.
7900 (_mm_maskz_min_sh): Likewise.
7901 (_mm_max_round_sh): Likewise.
7902 (_mm_mask_max_round_sh): Likewise.
7903 (_mm_maskz_max_round_sh): Likewise.
7904 (_mm_min_round_sh): Likewise.
7905 (_mm_mask_min_round_sh): Likewise.
7906 (_mm_maskz_min_round_sh): Likewise.
7907 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
7908 (_mm256_max_ph): Likewise.
7909 (_mm_mask_max_ph): Likewise.
7910 (_mm256_mask_max_ph): Likewise.
7911 (_mm_maskz_max_ph): Likewise.
7912 (_mm256_maskz_max_ph): Likewise.
7913 (_mm_min_ph): Likewise.
7914 (_mm256_min_ph): Likewise.
7915 (_mm_mask_min_ph): Likewise.
7916 (_mm256_mask_min_ph): Likewise.
7917 (_mm_maskz_min_ph): Likewise.
7918 (_mm256_maskz_min_ph): Likewise.
7919 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7920 * config/i386/i386-builtin.def: Add corresponding new builtins.
7921 * config/i386/i386-expand.c
7922 (ix86_expand_args_builtin): Handle new builtin types.
7923 * config/i386/sse.md
7924 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
7925 support HF vector modes.
7926 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
7927 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
7929 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
7931 * config/i386/subst.md (round_saeonly_mode512bit_condition):
7932 Adjust for HF vector modes.
7934 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
7936 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
7937 (_mm_mask_add_sh): Likewise.
7938 (_mm_maskz_add_sh): Likewise.
7939 (_mm_sub_sh): Likewise.
7940 (_mm_mask_sub_sh): Likewise.
7941 (_mm_maskz_sub_sh): Likewise.
7942 (_mm_mul_sh): Likewise.
7943 (_mm_mask_mul_sh): Likewise.
7944 (_mm_maskz_mul_sh): Likewise.
7945 (_mm_div_sh): Likewise.
7946 (_mm_mask_div_sh): Likewise.
7947 (_mm_maskz_div_sh): Likewise.
7948 (_mm_add_round_sh): Likewise.
7949 (_mm_mask_add_round_sh): Likewise.
7950 (_mm_maskz_add_round_sh): Likewise.
7951 (_mm_sub_round_sh): Likewise.
7952 (_mm_mask_sub_round_sh): Likewise.
7953 (_mm_maskz_sub_round_sh): Likewise.
7954 (_mm_mul_round_sh): Likewise.
7955 (_mm_mask_mul_round_sh): Likewise.
7956 (_mm_maskz_mul_round_sh): Likewise.
7957 (_mm_div_round_sh): Likewise.
7958 (_mm_mask_div_round_sh): Likewise.
7959 (_mm_maskz_div_round_sh): Likewise.
7960 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7961 * config/i386/i386-builtin.def: Add corresponding new builtins.
7962 * config/i386/i386-expand.c
7963 (ix86_expand_round_builtin): Handle new builtins.
7964 * config/i386/sse.md (VF_128): Change description.
7965 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
7966 Adjust to support HF vector modes.
7967 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
7970 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
7972 * config/i386/i386-expand.c
7973 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
7974 * config/i386/i386.c
7975 (ix86_preferred_simd_mode): Handle HF mode.
7976 * config/i386/sse.md (V_256H): New mode iterator.
7977 (avx_vextractf128<mode>): Use it.
7978 (VEC_INIT_MODE): Align vector HFmode condition to vector
7979 HImodes since there're no real HF instruction used.
7980 (VEC_INIT_HALF_MODE): Ditto.
7982 (VIHF_AVX512BW): Ditto.
7983 (*vec_extracthf): Ditto.
7984 (VEC_EXTRACT_MODE): Ditto.
7986 2021-09-10 Richard Biener <rguenther@suse.de>
7989 * config/dbx.h: Remove.
7990 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
7991 * config/lynx.h: Likewise.
7993 2021-09-10 liuhongt <hongtao.liu@intel.com>
7995 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
7996 right into ANDNOT + AND + IOR, using paradoxical subregs.
7997 (ix86_split_copysign_const): Remove.
7998 (ix86_split_copysign_var): Ditto.
7999 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
8000 (ix86_split_copysign_var): Ditto.
8001 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
8002 (@copysign<mode>3_var): Ditto.
8004 2021-09-09 qing zhao <qing.zhao@oracle.com>
8006 * builtins.c (expand_builtin_memset): Make external visible.
8007 * builtins.h (expand_builtin_memset): Declare extern.
8008 * common.opt (ftrivial-auto-var-init=): New option.
8009 * doc/extend.texi: Document the uninitialized attribute.
8010 * doc/invoke.texi: Document -ftrivial-auto-var-init.
8011 * flag-types.h (enum auto_init_type): New enumerated type
8013 * gimple-fold.c (clear_padding_type): Add one new parameter.
8014 (clear_padding_union): Likewise.
8015 (clear_padding_emit_loop): Likewise.
8016 (clear_type_padding_in_mask): Likewise.
8017 (gimple_fold_builtin_clear_padding): Handle this new parameter.
8018 * gimplify.c (gimple_add_init_for_auto_var): New function.
8019 (gimple_add_padding_init_for_auto_var): New function.
8020 (is_var_need_auto_init): New function.
8021 (gimplify_decl_expr): Add initialization to automatic variables per
8023 (gimplify_call_expr): Add one new parameter for call to
8024 __builtin_clear_padding.
8025 (gimplify_init_constructor): Add padding initialization in the end.
8026 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
8027 (expand_DEFERRED_INIT): New function.
8028 * internal-fn.def (DEFERRED_INIT): New internal function.
8029 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
8030 * tree-sra.c (generate_subtree_deferred_init): New function.
8031 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
8032 calls to .DEFERRED_INIT.
8033 (sra_modify_deferred_init): New function.
8034 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
8035 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
8036 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
8038 (check_defs): Likewise.
8039 (warn_uninitialized_vars): Likewise.
8040 * tree-ssa.c (ssa_undefined_value_p): Likewise.
8041 * tree.c (build_common_builtin_nodes): Build tree node for
8042 BUILT_IN_CLEAR_PADDING when needed.
8044 2021-09-09 Richard Biener <rguenther@suse.de>
8046 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
8049 2021-09-09 Richard Biener <rguenther@suse.de>
8051 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
8052 DOM walk from get_loop_body_in_dom_order using a worklist
8055 2021-09-09 liuhongt <hongtao.liu@intel.com>
8057 * config.gcc: Add avx512fp16vlintrin.h.
8058 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
8059 (_mm512_mask_add_ph): Likewise.
8060 (_mm512_maskz_add_ph): Likewise.
8061 (_mm512_sub_ph): Likewise.
8062 (_mm512_mask_sub_ph): Likewise.
8063 (_mm512_maskz_sub_ph): Likewise.
8064 (_mm512_mul_ph): Likewise.
8065 (_mm512_mask_mul_ph): Likewise.
8066 (_mm512_maskz_mul_ph): Likewise.
8067 (_mm512_div_ph): Likewise.
8068 (_mm512_mask_div_ph): Likewise.
8069 (_mm512_maskz_div_ph): Likewise.
8070 (_mm512_add_round_ph): Likewise.
8071 (_mm512_mask_add_round_ph): Likewise.
8072 (_mm512_maskz_add_round_ph): Likewise.
8073 (_mm512_sub_round_ph): Likewise.
8074 (_mm512_mask_sub_round_ph): Likewise.
8075 (_mm512_maskz_sub_round_ph): Likewise.
8076 (_mm512_mul_round_ph): Likewise.
8077 (_mm512_mask_mul_round_ph): Likewise.
8078 (_mm512_maskz_mul_round_ph): Likewise.
8079 (_mm512_div_round_ph): Likewise.
8080 (_mm512_mask_div_round_ph): Likewise.
8081 (_mm512_maskz_div_round_ph): Likewise.
8082 * config/i386/avx512fp16vlintrin.h: New header.
8083 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
8084 Add new builtin types.
8085 * config/i386/i386-builtin.def: Add corresponding builtins.
8086 * config/i386/i386-expand.c
8087 (ix86_expand_args_builtin): Handle new builtin types.
8088 (ix86_expand_round_builtin): Likewise.
8089 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
8090 * config/i386/sse.md (VFH): New mode_iterator.
8092 (avx512fmaskmode): Add HF vector modes.
8093 (avx512fmaskhalfmode): Likewise.
8094 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
8096 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
8097 (mul<mode>3<mask_name><round_name>): Likewise.
8098 (*mul<mode>3<mask_name><round_name>): Likewise.
8099 (div<mode>3): Likewise.
8100 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
8101 * config/i386/subst.md (SUBST_V): Add HF vector modes.
8102 (SUBST_A): Likewise.
8103 (round_mode512bit_condition): Adjust for V32HFmode.
8105 2021-09-09 liuhongt <hongtao.liu@intel.com>
8108 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
8109 (reduc_plus_scal_v4sf): .. this, New define_expand.
8110 (reduc_plus_scal_v2df): .. and this, New define_expand.
8112 2021-09-09 liuhongt <hongtao.liu@intel.com>
8115 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
8118 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
8121 * doc/trouble.texi (Copy Assignment): Fix description of
8122 behaviour and fix code in example.
8124 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
8127 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
8128 r11 instead of r12 for restoring CR.
8130 2021-09-08 Jakub Jelinek <jakub@redhat.com>
8131 liuhongt <hongtao.liu@intel.com>
8134 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
8135 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
8136 into AND with mask and XOR, using paradoxical subregs.
8137 (ix86_split_xorsign): Remove.
8138 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
8140 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
8142 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
8144 2021-09-08 Jakub Jelinek <jakub@redhat.com>
8147 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
8148 operands[2], emit abs<mode>2 instead.
8149 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
8150 first alternative even for avx, add another alternative with
8151 =&Yv <- 0, Yv, Yvm constraints.
8152 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
8153 to op1, emit vpandn instead.
8155 2021-09-08 liuhongt <hongtao.liu@intel.com>
8157 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
8158 (_mm256_set_ph): Likewise.
8159 (_mm512_set_ph): Likewise.
8160 (_mm_setr_ph): Likewise.
8161 (_mm256_setr_ph): Likewise.
8162 (_mm512_setr_ph): Likewise.
8163 (_mm_set1_ph): Likewise.
8164 (_mm256_set1_ph): Likewise.
8165 (_mm512_set1_ph): Likewise.
8166 (_mm_setzero_ph): Likewise.
8167 (_mm256_setzero_ph): Likewise.
8168 (_mm512_setzero_ph): Likewise.
8169 (_mm_set_sh): Likewise.
8170 (_mm_load_sh): Likewise.
8171 (_mm_store_sh): Likewise.
8172 * config/i386/i386-builtin-types.def (V8HF): New type.
8173 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
8174 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
8175 Support vector HFmodes.
8176 (ix86_expand_vector_init_one_nonzero): Likewise.
8177 (ix86_expand_vector_init_one_var): Likewise.
8178 (ix86_expand_vector_init_interleave): Likewise.
8179 (ix86_expand_vector_init_general): Likewise.
8180 (ix86_expand_vector_set): Likewise.
8181 (ix86_expand_vector_extract): Likewise.
8182 (ix86_expand_vector_init_concat): Likewise.
8183 (ix86_expand_sse_movcc): Handle vector HFmodes.
8184 (ix86_expand_vector_set_var): Ditto.
8185 * config/i386/i386-modes.def: Add HF vector modes in comment.
8186 * config/i386/i386.c (classify_argument): Add HF vector modes.
8187 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
8188 (ix86_vector_mode_supported_p): Likewise.
8189 (ix86_set_reg_reg_cost): Handle vector HFmode.
8190 (ix86_get_ssemov): Handle vector HFmode.
8191 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
8193 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
8195 (function_arg_advance_32): Ditto.
8196 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
8197 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
8198 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
8199 (VALID_SSE2_REG_VHF_MODE): New.
8200 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
8201 (SSE_REG_MODE_P): Add vector HFmode.
8202 * config/i386/i386.md (mode): Add HF vector modes.
8203 (MODE_SIZE): Likewise.
8204 (ssemodesuffix): Add ph suffix for HF vector modes.
8205 * config/i386/sse.md (VFH_128): New mode iterator.
8206 (VMOVE): Adjust for HF vector modes.
8208 (V_256_512): Likewise.
8210 (avx512fmaskmode): Likewise.
8211 (shuffletype): Likewise.
8212 (sseinsnmode): Likewise.
8213 (ssedoublevecmode): Likewise.
8214 (ssehalfvecmode): Likewise.
8215 (ssehalfvecmodelower): Likewise.
8216 (ssePScmode): Likewise.
8217 (ssescalarmode): Likewise.
8218 (ssescalarmodelower): Likewise.
8219 (sseintprefix): Likewise.
8221 (bcstscalarsuff): Likewise.
8222 (xtg_mode): Likewise.
8223 (VI12HF_AVX512VL): New mode_iterator.
8224 (VF_AVX512FP16): Likewise.
8226 (VIHF_256): Likewise.
8227 (VIHF_AVX512BW): Likewise.
8228 (V16_256): Likewise.
8229 (V32_512): Likewise.
8230 (sseintmodesuffix): New mode_attr.
8231 (sse): Add scalar and vector HFmodes.
8232 (ssescalarmode): Add vector HFmode mapping.
8233 (ssescalarmodesuffix): Add sh suffix for HFmode.
8234 (*<sse>_vm<insn><mode>3): Use VFH_128.
8235 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
8236 (*ieee_<ieee_maxmin><mode>3): Likewise.
8237 (<avx512>_blendm<mode>): New define_insn.
8238 (vec_setv8hf): New define_expand.
8239 (vec_set<mode>_0): New define_insn for HF vector set.
8240 (*avx512fp16_movsh): Likewise.
8241 (avx512fp16_movsh): Likewise.
8242 (vec_extract_lo_v32hi): Rename to ...
8243 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
8245 (vec_extract_hi_v32hi): Likewise.
8246 (vec_extract_hi_<mode>): Likewise.
8247 (vec_extract_lo_v16hi): Likewise.
8248 (vec_extract_lo_<mode>): Likewise.
8249 (vec_extract_hi_v16hi): Likewise.
8250 (vec_extract_hi_<mode>): Likewise.
8251 (vec_set_hi_v16hi): Likewise.
8252 (vec_set_hi_<mode>): Likewise.
8253 (vec_set_lo_v16hi): Likewise.
8254 (vec_set_lo_<mode>): Likewise.
8255 (*vec_extract<mode>_0): New define_insn_and_split for HF
8257 (*vec_extracthf): New define_insn.
8258 (VEC_EXTRACT_MODE): Add HF vector modes.
8259 (PINSR_MODE): Add V8HF.
8260 (sse2p4_1): Likewise.
8261 (pinsr_evex_isa): Likewise.
8262 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
8263 insert for V8HFmode.
8264 (pbroadcast_evex_isa): Add HF vector modes.
8265 (AVX2_VEC_DUP_MODE): Likewise.
8266 (VEC_INIT_MODE): Likewise.
8267 (VEC_INIT_HALF_MODE): Likewise.
8268 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
8270 (avx2_pbroadcast<mode>_1): Likewise.
8271 (<avx512>_vec_dup<mode>_1): Likewise.
8272 (<avx512>_vec_dup<mode><mask_name>): Likewise.
8273 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
8276 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
8277 H.J. Lu <hongjiu.lu@intel.com>
8278 Liu Hongtao <hongtao.liu@intel.com>
8279 Wang Hongyu <hongyu.wang@intel.com>
8280 Xu Dianhong <dianhong.xu@intel.com>
8282 * common/config/i386/cpuinfo.h (get_available_features):
8283 Detect FEATURE_AVX512FP16.
8284 * common/config/i386/i386-common.c
8285 (OPTION_MASK_ISA_AVX512FP16_SET,
8286 OPTION_MASK_ISA_AVX512FP16_UNSET,
8287 OPTION_MASK_ISA2_AVX512FP16_SET,
8288 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
8289 (OPTION_MASK_ISA2_AVX512BW_UNSET,
8290 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
8291 (ix86_handle_option): Handle -mavx512fp16.
8292 * common/config/i386/i386-cpuinfo.h (enum processor_features):
8293 Add FEATURE_AVX512FP16.
8294 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
8295 * config.gcc: Add avx512fp16intrin.h.
8296 * config/i386/avx512fp16intrin.h: New intrinsic header.
8297 * config/i386/cpuid.h: Add bit_AVX512FP16.
8298 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
8299 * config/i386/i386-builtins.c: Support _Float16 type for i386
8301 (ix86_register_float16_builtin_type): New function.
8302 (ix86_float16_type_node): New.
8303 * config/i386/i386-c.c (ix86_target_macros_internal): Define
8305 * config/i386/i386-expand.c (ix86_expand_branch): Support
8307 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
8308 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8309 (ix86_expand_fp_movcc): Ditto.
8310 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
8311 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
8312 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
8313 * config/i386/i386.c (ix86_get_ssemov): Use
8314 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
8315 (ix86_get_excess_precision): Use
8316 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
8318 (sse_store_index): Use SFmode cost for HFmode cost.
8319 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
8320 GPR cost for HFmode.
8321 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
8322 (ix86_mangle_type): Add manlging for _Float16 type.
8323 (inline_secondary_memory_needed): No memory is needed for
8324 16bit movement between gpr and sse reg under
8326 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
8327 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8328 (ix86_division_cost): Ditto.
8329 (ix86_rtx_costs): Ditto.
8330 (ix86_add_stmt_cost): Ditto.
8331 (ix86_optab_supported_p): Ditto.
8332 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
8333 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
8334 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
8335 * config/i386/i386.md (mode): Add HFmode.
8336 (MODE_SIZE): Add HFmode.
8337 (isa): Add avx512fp16.
8338 (enabled): Handle avx512fp16.
8339 (ssemodesuffix): Add sh suffix for HFmode.
8340 (comm): Add mult, div.
8341 (plusminusmultdiv): New code iterator.
8342 (insn): Add mult, div.
8343 (*movhf_internal): Adjust for avx512fp16 instruction.
8344 (*movhi_internal): Ditto.
8345 (*cmpi<unord>hf): New define_insn for HFmode.
8346 (*ieee_s<ieee_maxmin>hf3): Likewise.
8347 (extendhf<mode>2): Likewise.
8348 (trunc<mode>hf2): Likewise.
8349 (float<floatunssuffix><mode>hf2): Likewise.
8350 (*<insn>hf): Likewise.
8351 (cbranchhf4): New expander.
8352 (movhfcc): Likewise.
8353 (<insn>hf3): Likewise.
8356 * config/i386/i386.opt: Add mavx512fp16.
8357 * config/i386/immintrin.h: Include avx512fp16intrin.h.
8358 * doc/invoke.texi: Add mavx512fp16.
8359 * doc/extend.texi: Add avx512fp16 Usage Notes.
8361 2021-09-08 liuhongt <hongtao.liu@intel.com>
8363 * common.opt: Support -fexcess-precision=16.
8364 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
8365 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
8366 EXCESS_PRECISION_TYPE_FLOAT16.
8367 * config/arm/arm.c (arm_excess_precision): Ditto.
8368 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
8369 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
8370 when EXCESS_PRECISION_TYPE_FLOAT16.
8371 * config/s390/s390.c (s390_excess_precision): Ditto.
8372 * coretypes.h (enum excess_precision_type): Add
8373 EXCESS_PRECISION_TYPE_FLOAT16.
8374 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
8375 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
8376 * doc/extend.texi (Half-Precision): Document
8377 -fexcess-precision=16.
8378 * flag-types.h (enum excess_precision): Add
8379 EXCESS_PRECISION_FLOAT16.
8380 * target.def (excess_precision): Update document.
8381 * tree.c (excess_precision_type): Set excess_precision_type to
8382 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
8384 2021-09-08 liuhongt <hongtao.liu@intel.com>
8386 * doc/extend.texi: (@node Floating Types): Adjust the wording.
8387 (@node Half-Precision): Ditto.
8389 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
8392 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
8393 'CONST_INT_P (src)' to the condition of the block that tries to
8394 eliminate literal when loading integer contant.
8396 2021-09-07 David Faust <david.faust@oracle.com>
8398 * doc/extend.texi (BPF Type Attributes) New node.
8399 Document new preserve_access_index attribute.
8400 Document new preserve_access_index builtin.
8401 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
8403 2021-09-07 David Faust <david.faust@oracle.com>
8405 * config/bpf/bpf.c: Adjust includes.
8406 (bpf_handle_preserve_access_index_attribute): New function.
8407 (bpf_attribute_table): Use it here.
8408 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
8409 (bpf_option_override): Handle "-mco-re" option.
8410 (bpf_asm_init_sections): New.
8411 (TARGET_ASM_INIT_SECTIONS): Redefine.
8412 (bpf_file_end): New.
8413 (TARGET_ASM_FILE_END): Redefine.
8414 (bpf_init_builtins): Add "__builtin_preserve_access_index".
8415 (bpf_core_compute, bpf_core_get_index): New.
8416 (is_attr_preserve_access): New.
8417 (bpf_expand_builtin): Handle new builtins.
8418 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
8419 (bpf_core_walk): New.
8420 (bpf_resolve_overloaded_builtin): New.
8421 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
8423 (pass_bpf_core_attr): New RTL pass.
8424 * config/bpf/bpf-passes.def: New file.
8425 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
8426 * config/bpf/coreout.c: New file.
8427 * config/bpf/coreout.h: Likewise.
8428 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
8429 (coreout.o): New rule.
8430 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
8431 * config.gcc (bpf): Add coreout.h to extra_headers.
8432 Add coreout.o to extra_objs.
8433 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
8435 2021-09-07 David Faust <david.faust@oracle.com>
8437 * btfout.c (get_btf_id): Function is no longer static.
8438 * ctfc.h: Expose it here.
8440 2021-09-07 David Faust <david.faust@oracle.com>
8442 * ctfc.c (ctf_lookup_tree_type): New function.
8445 2021-09-07 David Faust <david.faust@oracle.com>
8447 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
8448 * ctfc.h: Analogous change.
8450 2021-09-07 David Faust <david.faust@oracle.com>
8452 * dwarf2out.c (lookup_type_die): Function is no longer static.
8453 * dwarf2out.h: Expose it here.
8455 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8457 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
8458 (ctf_debug_early_finish): New definition.
8459 (ctf_debug_finish): Likewise.
8460 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
8461 (ctf_debug_early_finish): New declaration.
8462 (ctf_debug_finish): Likewise.
8463 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
8464 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
8466 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8468 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
8469 support when compiling for CO-RE.
8470 * config/bpf/bpf.opt: Add new command line option -mco-re.
8472 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8474 * flag-types.h (enum debug_info_type): Add new enum
8475 DINFO_TYPE_BTF_WITH_CORE.
8476 (BTF_WITH_CORE_DEBUG): New bitmask.
8477 * flags.h (btf_with_core_debuginfo_p): New declaration.
8478 * opts.c (btf_with_core_debuginfo_p): New definition.
8480 2021-09-07 Jason Merrill <jason@redhat.com>
8482 * tree.h (error_operand_p): Change to inline function.
8484 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
8486 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
8487 (empty_block_with_phis_p): ...this.
8488 (potentially_threadable_block): Same.
8489 (jump_threader::thread_through_normal_block): Same.
8491 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
8494 * dwarf2out.c (mark_base_types): New overloaded function.
8495 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
8496 as the compilation unit, and call move_marked_base_types afterward.
8498 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
8501 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
8503 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
8505 2021-09-07 Richard Biener <rguenther@suse.de>
8507 PR tree-optimization/102226
8508 * tree-vect-loop.c (vect_transform_cycle_phi): Record
8509 the converted value for the epilogue PHI use.
8511 2021-09-07 Martin Liska <mliska@suse.cz>
8513 PR gcov-profile/80223
8514 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
8515 options, do not inline when no_profile_instrument_function
8516 attributes are different in early inliner. It's fine to inline
8517 it after PGO instrumentation.
8519 2021-09-07 Richard Biener <rguenther@suse.de>
8521 PR tree-optimization/101555
8522 * tree-ssa-pre.c (translate_vuse_through_block): Do not
8523 perform an alias walk to determine the validity of the
8524 mem at the start of the block which is already guaranteed
8525 by means of prune_clobbered_mems.
8526 (phi_translate_1): Pass edge to translate_vuse_through_block.
8528 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
8531 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
8532 (remainder<mode>3): Likewise.
8534 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
8536 * config/mips/mips.c (mips_file_start): add .module for
8539 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
8541 * wide-int.cc (wi::clz): Reorder tests to ensure the result
8542 is zero for all negative values.
8544 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
8546 * doc/invoke.texi (-foffload-options): Fix @opindex.
8548 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
8551 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
8552 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
8553 source alternative for AVX.
8555 2021-09-06 liuhongt <hongtao.liu@intel.com>
8557 PR middle-end/102182
8558 * optabs.c (expand_fix): Add from1 to avoid from being
8561 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
8563 * dwarf2out.c (modified_type_die): Deal with all array types earlier
8564 and use local variable consistently throughout the function.
8566 2021-09-06 Jakub Jelinek <jakub@redhat.com>
8568 PR tree-optimization/102207
8569 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8570 were promoted from signed to wider unsigned type.
8572 2021-09-06 Andrew Pinski <apinski@marvell.com>
8574 PR tree-optimization/63184
8575 * match.pd: Add simplification of pointer_diff of two pointer_plus
8576 with addr_expr in the first operand of each pointer_plus.
8577 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
8578 in the first operand of each pointer_plus.
8580 2021-09-06 Richard Biener <rguenther@suse.de>
8582 PR tree-optimization/102176
8583 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
8585 (vect_bb_slp_scalar_cost): Use the computed set of
8586 vectorized scalar stmts instead of relying on the out-of-date
8587 and not accurate PURE_SLP_STMT.
8588 (vect_bb_vectorization_profitable_p): Compute the set
8589 of vectorized scalar stmts.
8591 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8593 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
8594 GIMPLE_COND special casing.
8595 (path_range_query::range_defined_in_block): Use range_of_stmt
8596 instead of calling fold_range directly.
8598 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8600 * gimple-range-path.cc (path_range_query::range_of_expr): Set
8601 m_undefined_path when appropriate.
8602 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
8603 (path_range_query::unreachable_path_p): New.
8604 (path_range_query::precompute_ranges): Set m_undefined_path.
8605 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
8606 (path_range_query::internal_range_of_expr): New.
8607 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
8608 Use unreachable_path_p.
8610 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8612 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8613 Remove argument and call find_taken_edge.
8614 (back_threader::resolve_phi): Do not calculate taken edge before
8615 calling maybe_register_path.
8616 (back_threader::find_paths_to_names): Same.
8618 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
8620 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
8621 * config/h8300/testcompare.md (store_c): Update name, use new
8623 (store_neg_c, store_shifted_c): New patterns.
8625 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
8628 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
8629 instead of r12 for CR save, in all cases.
8631 2021-09-03 Andrew Pinski <apinski@marvell.com>
8633 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
8634 Handle error_mark_node as the type of the type_decl.
8636 2021-09-03 Andrew Pinski <apinski@marvell.com>
8638 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
8640 (aarch64_simd_types): Likewise.
8641 (aarch64_simd_intOI_type_node): Likewise.
8642 (aarch64_simd_intCI_type_node): Likewise.
8643 (aarch64_simd_intXI_type_node): Likewise.
8644 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
8645 (aarch64_fp16_ptr_type_node): Likewise.
8646 (aarch64_bf16_type_node): Likewise.
8647 (aarch64_bf16_ptr_type_node): Likewise.
8649 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8651 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
8653 (minus_op1_op2_relation_effect): ...here.
8654 (class operator_pointer_diff): New.
8655 (operator_pointer_diff::op1_op2_relation_effect): Call
8656 minus_op1_op2_relation_effect.
8657 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
8659 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8661 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
8662 Add may_peel_loop_headers.
8663 (back_threader_registry::thread_through_all_blocks): Same.
8664 (try_thread_blocks): Pass may_peel_loop_headers argument.
8665 (pass_early_thread_jumps::execute): Same.
8667 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8669 * tree-ssa-threadedge.c (has_phis_p): New.
8670 (forwarder_block_p): New.
8671 (potentially_threadable_block): Call forwarder_block_p.
8672 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
8673 (jump_threader::thread_through_normal_block): Call
8676 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8678 * tree-ssa-threadbackward.c (back_threader::dump): New.
8679 (back_threader::debug): New.
8680 (back_threader_profitability::profitable_path_p): Dump blocks
8681 even if we are bailing early.
8683 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8685 * tree-ssa-threadupdate.c (cancel_thread): New.
8686 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
8687 (jump_thread_path_registry::mark_threaded_blocks): Same.
8688 (jump_thread_path_registry::register_jump_thread): Same.
8690 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8692 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
8693 which objects are available.
8694 (jt_state::pop): Same.
8695 (jt_state::register_equiv): Same.
8696 (jt_state::register_equivs_on_edge): Same.
8698 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8700 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
8701 Move pop until after a thread is registered.
8703 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8705 * tree-ssa-threadupdate.c (debug): New.
8707 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8709 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
8710 (push_dump_file::~push_dump_file): New.
8711 (dump_ranger): Change dump_file temporarily while dumping
8713 * gimple-range-trace.h (class push_dump_file): New.
8715 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8717 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
8718 (dump_ranger): Dump function name.
8720 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8722 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8723 Adjust for non-null.
8724 (path_range_query::adjust_for_non_null_uses): New.
8725 (path_range_query::precompute_ranges): Call
8726 adjust_for_non_null_uses.
8727 * gimple-range-path.h: Add m_non_null and
8728 adjust_for_non_null_uses.
8730 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8732 * gimple-range-path.cc (path_range_query::dump): Dump path
8734 (path_range_query::precompute_ranges): Dump entire path.
8736 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8738 * value-relation.cc (relation_oracle::debug): New.
8739 * value-relation.h (relation_oracle::debug): New.
8741 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8743 * tree-ssa-loop-ch.c: Remove unnecessary include file.
8745 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8747 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
8748 Skip statements with no defining BB.
8749 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8750 Do not get confused by statements with no defining BB.
8752 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8754 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
8755 gimple_range_adjustment. Add support for constants.
8756 (adjust_realpart_expr): New.
8757 (gimple_range_adjustment): Move IMAGPART_EXPR code to
8758 adjust_imagpart_expr.
8759 * range-op.cc (integral_table::integral_table): Add entry for
8762 2021-09-03 Jakub Jelinek <jakub@redhat.com>
8764 * omp-expand.c (expand_omp_atomic_pipeline): Use
8765 IFN_ATOMIC_COMPARE_EXCHANGE instead of
8766 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
8769 2021-09-03 Jakub Jelinek <jakub@redhat.com>
8772 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
8773 that is false if DECL_BIT_FIELD.
8774 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
8775 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
8776 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
8777 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
8778 of DECL_FIELD_ABI_IGNORED.
8779 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
8780 SET_DECL_FIELD_ABI_IGNORED instead of writing to
8781 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
8782 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
8783 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
8784 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
8786 2021-09-03 liuhongt <hongtao.liu@intel.com>
8789 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
8790 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
8791 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
8793 2021-09-02 Martin Sebor <msebor@redhat.com>
8795 PR tree-optimization/17506
8797 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
8799 2021-09-02 Richard Biener <rguenther@suse.de>
8801 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
8802 fix for PR78185 and continue processing when leaving
8805 2021-09-02 Jakub Jelinek <jakub@redhat.com>
8807 PR tree-optimization/99591
8808 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8811 2021-09-02 Richard Biener <rguenther@suse.de>
8814 2021-09-02 Richard Biener <rguenther@suse.de>
8816 PR tree-optimization/102155
8817 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
8818 over a part of the RPO array and do not recurse here.
8819 Dump blocks marked as always executed.
8820 (fill_always_executed_in): Walk over the RPO array and
8821 process loops whose header we run into.
8822 (loop_invariant_motion_in_fun): Compute the first RPO
8823 using rev_post_order_and_mark_dfs_back_seme in iteration
8824 order and pass that to fill_always_executed_in.
8826 2021-09-02 liuhongt <hongtao.liu@intel.com>
8828 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
8829 * config/i386/i386.c (enum x86_64_reg_class): Add
8831 (merge_classes): Handle X86_64_SSEHF_CLASS.
8832 (examine_argument): Ditto.
8833 (construct_container): Ditto.
8834 (classify_argument): Ditto, and set HFmode/HCmode to
8836 (function_value_32): Return _FLoat16/Complex Float16 by
8838 (function_value_64): Return _Float16/Complex Float16 by SSE
8840 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
8841 (ix86_secondary_reload): Require gpr as intermediate register
8842 to store _Float16 from sse register when sse4 is not
8844 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
8846 (ix86_scalar_mode_supported_p): Ditto.
8847 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
8848 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
8849 (VALID_INT_MODE_P): Add HFmode and HCmode.
8850 * config/i386/i386.md (*pushhf_rex64): New define_insn.
8852 (*movhf_internal): Ditto.
8853 * doc/extend.texi (Half-Precision Floating Point): Documemt
8856 2021-09-02 Richard Biener <rguenther@suse.de>
8858 PR tree-optimization/102155
8859 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
8860 over a part of the RPO array and do not recurse here.
8861 Dump blocks marked as always executed.
8862 (fill_always_executed_in): Walk over the RPO array and
8863 process loops whose header we run into.
8864 (loop_invariant_motion_in_fun): Compute the first RPO
8865 using rev_post_order_and_mark_dfs_back_seme in iteration
8866 order and pass that to fill_always_executed_in.
8868 2021-09-02 YunQiang Su <syq@debian.org>
8871 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
8873 * config/mips/mips.c (mips_module_isa_name): New.
8874 mips_file_start: add .module mipsREV to all asm output
8876 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
8878 PR tree-optimization/102152
8879 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
8880 comparison to a scalar comparison before calling
8881 update_stmt_if_modified.
8883 2021-09-01 Andrew Pinski <apinski@marvell.com>
8886 * config/aarch64/aarch64.c (aarch64_expand_setmem):
8887 Check STRICT_ALIGNMENT before creating an overlapping
8890 2021-09-01 Martin Sebor <msebor@redhat.com>
8892 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
8893 (check_access): Pass additional argument.
8894 (check_memop_access): Remove template and make a member function.
8895 (maybe_check_dealloc_call): Make a pass_waccess member function.
8896 (class pass_waccess): Add, rename, and remove members.
8897 (pass_waccess::pass_waccess): Adjust to name change.
8898 (pass_waccess::~pass_waccess): Same.
8899 (check_alloca): Make a member function.
8900 (check_alloc_size_call): Same.
8901 (check_strcat): Same.
8902 (check_strncat): Same.
8903 (check_stxcpy): Same.
8904 (check_stxncpy): Same.
8905 (check_strncmp): Same.
8906 (maybe_warn_rdwr_sizes): Rename...
8907 (pass_waccess::maybe_check_access_sizes): ...to this.
8908 (pass_waccess::check_call): Adjust to name changes.
8909 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
8911 (pass_waccess::execute): Adjust to name changes.
8912 * gimple-ssa-warn-access.h (check_memop_access): Remove.
8913 * pointer-query.cc (access_ref::phi): Handle null pointer.
8914 (access_ref::inform_access): Same.
8915 (pointer_query::put_ref): Modify a cached value, not a copy of it.
8916 (pointer_query::dump): New function.
8917 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
8919 * pointer-query.h (pointer_query::dump): Declare.
8920 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
8921 (dump_strlen_info): Use function query.
8922 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
8924 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
8926 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
8928 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
8930 * doc/extend.texi: Document unavailable attribute.
8931 * print-tree.c (print_node): Handle unavailable attribute.
8932 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
8933 * tree.c (error_unavailable_use): New.
8934 * tree.h (TREE_UNAVAILABLE): New.
8935 (error_unavailable_use): New.
8937 2021-09-01 Jakub Jelinek <jakub@redhat.com>
8939 PR tree-optimization/102124
8940 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
8941 MINUS_EXPR, if itype is unsigned with smaller precision than type,
8942 add an extra cast to signed variant of itype to ensure sign-extension.
8944 2021-09-01 Martin Liska <mliska@suse.cz>
8946 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
8947 edges and rather use colors for TRUE and FALSE edges.
8949 2021-09-01 Richard Biener <rguenther@suse.de>
8951 PR tree-optimization/93491
8952 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
8953 after processing the stmt itself. Do not consider
8954 pure functions possibly not returning. Properly avoid
8955 adding possibly trapping calls to EXP_GEN when there's
8956 a preceeding possibly not returning call.
8957 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
8960 2021-09-01 Richard Biener <rguenther@suse.de>
8962 PR tree-optimization/102139
8963 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
8964 type to record a std::pair of the stmt-info and the innermost
8966 (dr_vec_info::group): New member.
8967 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
8968 (vect_compute_data_ref_alignment): Verify the recorded
8969 base alignment can be used.
8970 (data_ref_pair): Remove.
8971 (dr_group_sort_cmp): Adjust.
8972 (vect_analyze_data_ref_accesses): Store the group-ID in the
8973 dr_vec_info and operate on a vector of dr_vec_infos.
8975 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
8977 * read-md.c (md_reader::handle_enum): support value assignation.
8978 * doc/md.texi: record define_c_enum value assignation support.
8980 2021-09-01 Jakub Jelinek <jakub@redhat.com>
8982 PR tree-optimization/102141
8983 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
8984 argument. If false, emit stmts after gsi instead of before, and
8986 (bswap_replace): Adjust callers. When converting output of bswap,
8987 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
8990 2021-09-01 liuhongt <hongtao.liu@intel.com>
8992 * emit-rtl.c (validate_subreg): Get rid of all float-int
8995 2021-09-01 liuhongt <hongtao.liu@intel.com>
8998 2021-08-30 liuhongt <hongtao.liu@intel.com>
9000 * expmed.c (extract_bit_field_1): Make sure we're playing with
9001 integral modes before call extract_integral_bit_field.
9002 (extract_integral_bit_field): Add a parameter of type
9003 scalar_int_mode which corresponds to of tmode.
9004 And call extract_and_convert_fixed_bit_field instead of
9005 extract_fixed_bit_field and convert_extracted_bit_field.
9006 (extract_and_convert_fixed_bit_field): New function, it's a
9007 combination of extract_fixed_bit_field and
9008 convert_extracted_bit_field.
9010 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
9012 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
9014 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
9016 * omp-general.h (omp_is_reference): Rename to...
9017 (omp_privatize_by_reference): ... this. Adjust all users...
9018 * omp-general.c: ... here, ...
9019 * gimplify.c: ... here, ...
9020 * omp-expand.c: ... here, ...
9021 * omp-low.c: ... here.
9023 2021-08-31 Martin Sebor <msebor@redhat.com>
9025 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
9026 pointer element for equality to zero, not that of the cotaining
9029 2021-08-31 Martin Sebor <msebor@redhat.com>
9031 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
9033 2021-08-31 Martin Sebor <msebor@redhat.com>
9035 * function.h (function): Add comments.
9036 (get_range_query): Same. Add attribute returns nonnull.
9038 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9040 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
9041 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
9042 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
9044 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
9046 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
9048 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
9049 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
9050 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
9051 (m32c-*-elf*, m32c-*-rtems*): Likewise.
9052 * config/cris/cris.h (DBX_NO_XREFS): Remove.
9053 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
9054 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
9055 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
9056 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
9057 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
9059 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
9061 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
9062 allowed on target constructs and only with particular other clauses.
9063 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
9065 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
9066 structs when 'ancestor' is used.
9067 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
9069 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
9070 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
9072 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9074 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
9075 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9077 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
9078 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
9079 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
9080 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
9081 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
9082 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9083 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
9084 would be paradoxical.
9086 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9088 * combine.c (combine_simplify_rtx): Avoid converting an explicit
9089 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
9090 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9092 2021-08-31 Richard Biener <rguenther@suse.de>
9094 PR tree-optimization/102142
9095 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
9096 condition under which to unset the visited flag.
9098 2021-08-31 Richard Biener <rguenther@suse.de>
9100 PR middle-end/102129
9101 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
9102 possibly trapping expressions across calls.
9104 2021-08-31 Jakub Jelinek <jakub@redhat.com>
9106 PR tree-optimization/102134
9107 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
9108 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
9111 2021-08-31 Andrew Pinski <apinski@marvell.com>
9114 * collect-utils.c (setup_signals): New declaration.
9115 * collect-utils.h (setup_signals): New function.
9116 * collect2.c (handler): Delete.
9117 (main): Instead of manually setting up the signals,
9118 just call setup_signals.
9119 * lto-wrapper.c (main): Likewise.
9121 2021-08-31 Andrew Pinski <apinski@marvell.com>
9124 * config/i386/i386-protos.h (x86_output_aligned_bss):
9125 Change align argument to unsigned type.
9126 (x86_elf_aligned_decl_common): Likewise.
9127 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
9128 (x86_output_aligned_bss): Likewise.
9130 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9132 * config/mips/mips.c (mips_module_isa_name): New.
9133 mips_file_start: add .module mipsREV to all asm output
9135 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9137 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
9138 use enum instead of int for 'isa' member.
9139 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
9140 config/mips/netbsd.h: replace hardcoded numbers with enum.
9142 2021-08-31 liuhongt <hongtao.liu@intel.com>
9144 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
9145 define_split to define_insn_and_split.
9146 (*avx2_eq<mode>3): Removed.
9147 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
9148 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
9149 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
9151 (*avx2_gt<mode>3): Removed.
9152 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
9153 define_insn to define_expand, and adjust pattern.
9154 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
9156 2021-08-30 David Malcolm <dmalcolm@redhat.com>
9159 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
9161 2021-08-30 Jason Merrill <jason@redhat.com>
9163 * doc/invoke.texi: Document -Wmissing-requires.
9165 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9167 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
9168 TARGET_EXTRA_BUILTINS guard.
9170 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9172 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
9173 initialization of V2DI_type_node and unsigned_V2DI_type_node.
9175 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9177 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
9178 decl when new_builtins_are_live.
9179 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
9182 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
9184 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
9185 OPTION_MASK_P10_FUSION_2STORE.
9186 (POWERPC_MASKS): Likewise.
9187 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
9188 store fusion for Power10.
9189 (is_fusable_store): New.
9190 (power10_sched_reorder): Likewise.
9191 (rs6000_sched_reorder): Do Power10 specific reordering.
9192 (rs6000_sched_reorder2): Likewise.
9193 * config/rs6000/rs6000.opt: Add new option.
9195 2021-08-30 Richard Biener <rguenther@suse.de>
9197 PR tree-optimization/102128
9198 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9199 Move scanning for if-converted scalar code to the caller
9200 and instead delay clearing the visited flag for profitable
9202 (vect_slp_region): Cost all subgraphs before scheduling.
9203 For if-converted BB vectorization scan for scalar COND_EXPRs
9204 and do not vectorize if any found and the cost model is
9207 2021-08-30 Richard Biener <rguenther@suse.de>
9209 * common.opt (fexceptions): Mark
9210 EnabledBy(fnon-call-exceptions).
9211 * doc/invoke.texi (fnon-call-exceptions): Document this
9212 enables -fexceptions.
9214 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
9216 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
9217 is not already defined.
9219 2021-08-30 liuhongt <hongtao.liu@intel.com>
9221 * expmed.c (extract_bit_field_1): Make sure we're playing with
9222 integral modes before call extract_integral_bit_field.
9223 (extract_integral_bit_field): Add a parameter of type
9224 scalar_int_mode which corresponds to of tmode.
9225 And call extract_and_convert_fixed_bit_field instead of
9226 extract_fixed_bit_field and convert_extracted_bit_field.
9227 (extract_and_convert_fixed_bit_field): New function, it's a
9228 combination of extract_fixed_bit_field and
9229 convert_extracted_bit_field.
9231 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9233 * config/darwin.c (darwin_libc_has_function): Do not run
9234 the checks for x86 or modern Darwin. Make sure that there
9235 is a value set for darwin_macosx_version_min before testing.
9237 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9239 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
9241 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
9243 * ipa-modref-tree.h (modref_access_node::merge): Break out
9244 logic combining offsets and logic merging ranges to ...
9245 (modref_access_node::combined_offsets): ... here
9246 (modref_access_node::update2): ... here
9247 (modref_access_node::closer_pair_p): New member function.
9248 (modref_access_node::forced_merge): New member function.
9249 (modre_ref_node::insert): Do merging when table is full.
9251 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
9254 * config.gcc: MIPS: use N64 ABI by default if the triple end
9255 with -gnuabi64, which is used by Debian since 2013.
9257 2021-08-28 Alexandre Oliva <oliva@adacore.com>
9259 * ipa-modref.c (analyze_function): Skip debug stmts.
9260 * tree-inline.c (estimate_num_insn): Consider builtins even
9261 without a cgraph_node.
9263 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9265 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
9266 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
9267 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
9268 * config/h8300/h8300.md (eqne): New code iterator.
9269 (geultu, geultu_to_c): Similarly.
9270 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
9271 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
9272 (cmp<mode>_c): New pattern
9274 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9276 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
9278 (dom_opt_dom_walker::optimize_stmt): Use it.
9280 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9282 * config/darwin.c (finalize_ctors): Add a section-start linker-
9284 (finalize_dtors): Likewise.
9285 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
9287 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
9289 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
9290 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
9291 old initialization logic when new builtins are enabled.
9292 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
9293 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
9294 (write_init_file): Likewise.
9296 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9298 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
9299 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
9300 (amdgcn-* | gcn-*) Likewise.
9302 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
9304 * symbol-summary.h: Added missing template keyword.
9306 2021-08-27 Richard Biener <rguenther@suse.de>
9308 PR tree-optimization/45178
9309 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
9310 infinite loops without exit do not mark control dependent
9311 edges of the latch necessary.
9313 2021-08-27 konglin1 <lingling.kong@intel.com>
9316 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
9318 (<avx512>scattersi<mode>): Likewise.
9319 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
9320 (*avx512f_scatterdi<VI48F:mode>): Likewise
9322 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9324 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
9325 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
9326 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
9327 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
9328 P10_BUILTIN_PEXTD on Power10.
9330 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9332 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
9333 signedness for some Power10 bifs.
9335 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
9338 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
9339 computed alignment if the entire struct has attribute packed.
9341 2021-08-27 liuhongt <hongtao.liu@intel.com>
9345 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
9346 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
9347 IX86_BUILTIN_SHUFPD256， IX86_BUILTIN_SHUFPS，
9348 IX86_BUILTIN_SHUFPS256.
9349 (ix86_masked_all_ones): New function.
9351 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9353 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
9354 (conditional moves with memory inputs splitters): Ditto.
9355 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
9357 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9359 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
9360 search after merging.
9362 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9364 * config/rs6000/rs6000-overload.def: Add remaining overloads.
9366 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9368 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
9370 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9372 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
9373 crypto, and htm stanzas.
9375 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9377 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
9379 2021-08-26 Martin Sebor <msebor@redhat.com>
9381 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
9382 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
9383 (warn_uninitialized_vars): Same. Reduce visibility of locals.
9384 (warn_uninitialized_phi): Same.
9386 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9388 * tree-ssa-ccp.c (get_individual_bits): Helper function to
9389 extract the individual bits from a widest_int constant (mask).
9390 (gray_code_bit_flips): New read-only table for effiently
9391 enumerating permutations/combinations of bits.
9392 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
9393 by unknown counts that are guaranteed less than the target
9394 precision and four or fewer unknown bits by enumeration.
9395 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
9396 enumeration under the same conditions. Handle remaining
9397 shifts as a mask based upon the minimum possible shift value.
9399 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9400 Richard Biener <rguenther@suse.de>
9402 * match.pd (shift transformations): Remove a redundant
9403 !POINTER_TYPE_P check.
9405 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9408 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
9409 Set all_regs to true in the call to replace_rtx.
9411 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9413 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
9414 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
9415 hiting --param modref-max-refs limit.
9416 (modref_tree:insert_base): Be smarter when hitting
9417 --param modref-max-bases limit. Add new parameter REF.
9418 (modref_tree:insert): Update.
9419 (modref_tree:merge): Update.
9420 * ipa-modref.c (read_modref_records): Update.
9422 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9424 * params.opt: (modref-max-adjustments): Add full stop.
9426 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9428 * ipa-modref-tree.h (modref_ref_node::verify): New member
9430 (modref_ref_node::insert): Use it.
9431 (modref_ref_node::try_mere_with): Fix off by one error.
9433 2021-08-26 Martin Liska <mliska@suse.cz>
9434 Stefan Kneifel <stefan.kneifel@bluewin.ch>
9436 * cgraph.h (create_version_clone_with_body): Add new parameter.
9437 * cgraphclones.c: Likewise.
9438 * multiple_target.c (create_dispatcher_calls): Do not use
9440 (create_target_clone): Likewise here.
9442 2021-08-26 Jonathan Yong <10walls@gmail.com>
9444 * doc/extend.texi: Add note about reserved priorities
9445 to the constructor attribute.
9447 2021-08-25 Martin Sebor <msebor@redhat.com>
9449 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
9450 range table header alone.
9451 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
9453 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
9455 * doc/invoke.texi: Document --param modref-max-adjustments.
9456 * ipa-modref-tree.c (test_insert_search_collapse): Update.
9457 (test_merge): Update.
9458 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
9459 (modref_access_node::operator==): Fix handling of access ranges.
9460 (modref_access_node::contains): Constify parameter; handle also
9461 mismatched parm offsets.
9462 (modref_access_node::update): New function.
9463 (modref_access_node::merge): New function.
9464 (unspecified_modref_access_node): Update constructor.
9465 (modref_ref_node::insert_access): Add record_adjustments parameter;
9467 (modref_ref_node::try_merge_with): New private function.
9468 (modref_tree::insert): New record_adjustments parameter.
9469 (modref_tree::merge): New record_adjustments parameter.
9470 (modref_tree::copy_from): Update.
9471 * ipa-modref.c (dump_access): Dump adjustments field.
9472 (get_access): Update constructor.
9473 (record_access): Update call of insert.
9474 (record_access_lto): Update call of insert.
9475 (merge_call_side_effects): Add record_adjustments parameter.
9476 (get_access_for_fnspec): Update.
9477 (process_fnspec): Update.
9478 (analyze_call): Update.
9479 (analyze_function): Update.
9480 (read_modref_records): Update.
9481 (ipa_merge_modref_summary_after_inlining): Update.
9482 (propagate_unknown_call): Update.
9483 (modref_propagate_in_scc): Update.
9484 * params.opt (param-max-modref-adjustments=): New.
9486 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
9488 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
9490 (xxspltiw_v4si): Use vecperm type attribute.
9491 (xxspltiw_v4si_inst): Use vecperm type attribute.
9492 (xxspltiw_v4sf_inst): Likewise.
9493 (xxspltidp_v2df): Use vecperm type attribute. Use
9494 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
9495 (xxspltidp_v2df_inst): Likewise.
9496 (xxsplti32dx_v4si): Use vecperm type attribute.
9497 (xxsplti32dx_v4si_inst): Likewise.
9498 (xxsplti32dx_v4sf_inst): Likewise.
9499 (xxblend_<mode>): Likewise.
9500 (xxpermx): Likewise.
9501 (xxpermx_inst): Likewise.
9504 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
9507 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
9508 * diagnostic.c (diagnostic_initialize_input_context): New function.
9509 * diagnostic.h (diagnostic_initialize_input_context): Declare.
9510 * input.c (default_charset_callback): New function.
9511 (file_cache::initialize_input_context): New function.
9512 (file_cache_slot::create): Added ability to convert the input
9513 according to the input context.
9514 (file_cache::file_cache): Initialize the new input context.
9515 (class file_cache_slot): Added new m_alloc_offset member.
9516 (file_cache_slot::file_cache_slot): Initialize the new member.
9517 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
9518 (file_cache_slot::maybe_grow): Likewise.
9519 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
9520 (file_cache_slot::get_next_line): Likewise.
9521 * input.h (class file_cache): Added input context member.
9523 2021-08-25 Richard Biener <rguenther@suse.de>
9525 PR tree-optimization/102046
9526 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
9527 update ->any_pattern when swapping operands.
9529 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
9532 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
9533 (ix86_decompose_address): Remove retval check for ASHIFT,
9534 allow non-canonical zero extend if AND mask covers ASHIFT
9536 (ix86_legitimate_address_p): Adjust condition for decompose.
9537 (ix86_rtx_costs): Adjust cost for lea with non-canonical
9539 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
9541 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
9543 PR tree-optimization/101145
9544 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
9546 (number_of_iterations_lt): Invoke above function.
9547 (adjust_cond_for_loop_until_wrap):
9548 Merge to number_of_iterations_until_wrap.
9549 (number_of_iterations_cond): Update invokes for
9550 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
9552 2021-08-25 konglin1 <lingling.kong@intel.com>
9555 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
9557 (_mm512_mask_fpclass_ps_mask): Ditto.
9559 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
9561 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
9562 (vec_unpacku_hi_v8hi): Likewise.
9563 (vec_unpacku_lo_v16qi): Likewise.
9564 (vec_unpacku_lo_v8hi): Likewise.
9565 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
9566 (vec_unpacku_lo_<VP_small_lc>): Likewise.
9568 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
9570 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
9571 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9572 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9573 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
9575 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9577 PR middle-end/102031
9578 * simplify-rtx.c (simplify_truncation): When comparing precisions
9579 use "subreg_prec" variable, not "subreg_mode".
9581 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9583 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
9586 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9588 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
9589 various pointer type nodes.
9590 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
9591 values for various pointer types.
9592 (ptr_V16QI_type_node): New macro.
9593 (ptr_V1TI_type_node): New macro.
9594 (ptr_V2DI_type_node): New macro.
9595 (ptr_V2DF_type_node): New macro.
9596 (ptr_V4SI_type_node): New macro.
9597 (ptr_V4SF_type_node): New macro.
9598 (ptr_V8HI_type_node): New macro.
9599 (ptr_unsigned_V16QI_type_node): New macro.
9600 (ptr_unsigned_V1TI_type_node): New macro.
9601 (ptr_unsigned_V8HI_type_node): New macro.
9602 (ptr_unsigned_V4SI_type_node): New macro.
9603 (ptr_unsigned_V2DI_type_node): New macro.
9604 (ptr_bool_V16QI_type_node): New macro.
9605 (ptr_bool_V8HI_type_node): New macro.
9606 (ptr_bool_V4SI_type_node): New macro.
9607 (ptr_bool_V2DI_type_node): New macro.
9608 (ptr_bool_V1TI_type_node): New macro.
9609 (ptr_pixel_type_node): New macro.
9610 (ptr_intQI_type_node): New macro.
9611 (ptr_uintQI_type_node): New macro.
9612 (ptr_intHI_type_node): New macro.
9613 (ptr_uintHI_type_node): New macro.
9614 (ptr_intSI_type_node): New macro.
9615 (ptr_uintSI_type_node): New macro.
9616 (ptr_intDI_type_node): New macro.
9617 (ptr_uintDI_type_node): New macro.
9618 (ptr_intTI_type_node): New macro.
9619 (ptr_uintTI_type_node): New macro.
9620 (ptr_long_integer_type_node): New macro.
9621 (ptr_long_unsigned_type_node): New macro.
9622 (ptr_float_type_node): New macro.
9623 (ptr_double_type_node): New macro.
9624 (ptr_long_double_type_node): New macro.
9625 (ptr_dfloat64_type_node): New macro.
9626 (ptr_dfloat128_type_node): New macro.
9627 (ptr_ieee128_type_node): New macro.
9628 (ptr_ibm128_type_node): New macro.
9629 (ptr_vector_pair_type_node): New macro.
9630 (ptr_vector_quad_type_node): New macro.
9631 (ptr_long_long_integer_type_node): New macro.
9632 (ptr_long_long_unsigned_type_node): New macro.
9634 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9636 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
9637 and power9-64 stanzas.
9639 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9640 Tom de Vries <tdevries@suse.de>
9642 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
9643 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
9644 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
9645 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
9646 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
9647 * config/nvptx/nvptx-c.c: New source file.
9648 (nvptx_cpu_cpp_builtins): Move implementation here.
9650 2021-08-24 Martin Sebor <msebor@redhat.com>
9652 PR middle-end/101600
9653 PR middle-end/101977
9654 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
9655 the phrasing of a warning.
9656 (check_access): Use the remaining size after subtracting any offset
9657 rather than the whole object size.
9658 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
9659 clear for any nonnull PHI argument.
9660 (compute_objsize): Clear argument.
9662 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9664 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
9666 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9668 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
9671 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
9673 * value-relation.cc (rr_transitive_table): New.
9674 (relation_transitive): New.
9675 (value_relation::swap): Remove.
9676 (value_relation::apply_transitive): New.
9677 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
9678 (relation_oracle::register_relation): Call register_transitives.
9679 (relation_oracle::register_transitives): New.
9680 * value-relation.h (relation_oracle): Add new temporary bitmap and
9683 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
9686 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
9687 from integer to a pseudo vector register.
9689 2021-08-24 Richard Biener <rguenther@suse.de>
9691 PR tree-optimization/100089
9692 * tree-vectorizer.h (vect_slp_bb): Rename to ...
9693 (vect_slp_if_converted_bb): ... this and get the original
9694 loop as new argument.
9695 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
9696 pass original loop to vect_slp_if_converted_bb.
9697 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9698 If orig_loop was passed scan the not vectorized stmts
9699 for COND_EXPRs and force not profitable if found.
9700 (vect_slp_region): Pass down all SLP instances to costing
9701 if orig_loop was specified.
9702 (vect_slp_bbs): Pass through orig_loop.
9703 (vect_slp_bb): Rename to ...
9704 (vect_slp_if_converted_bb): ... this and get the original
9705 loop as new argument.
9706 (vect_slp_function): Adjust.
9708 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9711 * config/arm/arm.md (attribute arch): Add fix_vlldm.
9712 (arch_enabled): Use it.
9713 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
9714 use when erratum mitigation is needed.
9716 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9719 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
9720 * doc/invoke.texi (Arm Options): Document it.
9721 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
9722 (ALL_QUIRKS): Add quirk_vlldm.
9723 (cortex-m33): Add quirk_vlldm.
9724 (cortex-m35p, cortex-m55): Likewise.
9725 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
9726 targetting an affected CPU and not explicitly controlled on
9729 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9731 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
9732 (lazy_load_multiple_insn): Likewise.
9734 2021-08-24 liuhongt <hongtao.liu@intel.com>
9737 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
9738 Enable avx512 embedded broadcast.
9739 (*<avx512>_vternlog<mode>_all): Ditto.
9740 (<avx512>_vternlog<mode>_mask): Ditto.
9742 2021-08-24 liuhongt <hongtao.liu@intel.com>
9745 * config/i386/i386.c (ix86_rtx_costs): Define cost for
9747 * config/i386/i386.h (STRIP_UNARY): New macro.
9748 * config/i386/predicates.md (reg_or_notreg_operand): New
9750 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
9751 (*<avx512>_vternlog<mode>_1): New pre_reload
9752 define_insn_and_split.
9753 (*<avx512>_vternlog<mode>_2): Ditto.
9754 (*<avx512>_vternlog<mode>_3): Ditto.
9755 (any_logic1,any_logic2): New code iterator.
9756 (logic_op): New code attribute.
9757 (ternlogsuffix): Extend to VNxDF and VNxSF.
9759 2021-08-24 Richard Biener <rguenther@suse.de>
9761 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
9762 * params.opt (--param vect-inner-loop-cost-factor): Adjust
9764 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
9765 inner_loop_cost_factor to the minimum of the estimated number
9766 of iterations of the inner loop and vect-inner-loop-cost-factor.
9768 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9769 Richard Biener <rguenther@suse.de>
9771 * config/i386/i386-features.c (compute_convert_gain): Provide
9772 more accurate values for CONST_INT, when optimizing for size.
9773 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
9774 * config/i386/i386.h (COSTS_N_BYTES): to here.
9776 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9777 Jakub Jelinek <jakub@redhat.com>
9779 PR middle-end/102029
9780 * match.pd (shift transformations): Add an additional check for
9781 !POINTER_TYPE_P in the recently added left shift transformation.
9783 2021-08-24 liuhongt <hongtao.liu@intel.com>
9785 PR tree-optimization/100089
9786 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
9787 loop vectorizer when cost model is very-cheap.
9789 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9791 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
9792 asprintf, which is not available on AIX.
9794 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9796 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
9797 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
9799 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9801 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
9802 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
9804 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
9805 (rbtree.o): Likewise.
9806 (rs6000-gen-builtins): Likewise.
9807 (rs6000-builtins.c): Likewise.
9808 (rs6000-builtins.h): Likewise.
9809 (rs6000.o): Add dependency.
9810 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
9811 (rs6000-vecdefines.h): New target.
9812 (rs6000-builtins.o): Likewise.
9813 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
9814 (rs6000-c.o): Likewise.
9816 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9819 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
9820 Diagnose buffer overrun.
9821 (safe_inc_pos): Fix overrun detection.
9822 (match_identifier): Diagnose buffer overrun.
9823 (match_integer): Likewise.
9824 (match_to_right_bracket): Likewise.
9826 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9828 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
9829 Improve range compare.
9830 (modref_access_node::contains): New member function.
9831 (modref_access_node::search): Remove.
9832 (modref_access_node::insert): Be smarter about subaccesses.
9834 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
9836 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
9837 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
9838 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
9841 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
9843 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
9845 * config/h8300/jumpcall.md (call, call_value): Restrict to
9846 !SIBLING_CALL_P cases.
9847 (subcall, sibcall_value): New patterns & expanders.
9848 * config/h8300/proepi.md (epilogue): Pass new argument to
9849 h8300_expand_epilogue.
9850 (sibcall_epilogue): New expander.
9851 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
9853 (h8300_ok_for_sibcall_p): New function.
9854 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
9856 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9858 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
9859 Handle case where the operand is already the desired mode.
9861 2021-08-23 Richard Biener <rguenther@suse.de>
9864 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
9865 in addition to has_gimple_body.
9867 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9869 PR middle-end/101949
9870 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
9873 2021-08-23 Martin Liska <mliska@suse.cz>
9875 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
9878 2021-08-23 Richard Biener <rguenther@suse.de>
9880 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9881 Properly scale the inner loop cost only once.
9883 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9885 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
9886 Provide bounds for unsigned (and signed with non-negative operands)
9887 division and modulus.
9889 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9891 * simplify-rtx.c (simplify_truncation): Generalize simplification
9892 of (truncate:A (subreg:B X)).
9893 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
9894 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
9895 already has the desired machine mode.
9896 (test_scalar_int_ops): Add tests that useless extensions and
9897 truncations are optimized away.
9898 (test_scalar_int_ext_ops): New self-test function to confirm
9899 that truncations of extensions are correctly simplified.
9900 (test_scalar_int_ext_ops2): New self-test function to check
9901 truncations of truncations, extensions of extensions, and
9902 truncations of extensions.
9903 (test_scalar_ops): Call the above two functions with a
9904 representative sampling of integer machine modes.
9906 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9908 * match.pd (shift transformations): Change the sign of an
9909 LSHIFT_EXPR if it reduces the number of explicit conversions.
9911 2021-08-23 Jakub Jelinek <jakub@redhat.com>
9913 PR tree-optimization/86723
9914 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
9915 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
9916 non-memory permutation of 64-bit src which only has bytes of
9917 0 or [5..8] and n->range is 4.
9918 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
9919 find_bswap_or_nop_finalize caller, support bswap with some bytes
9920 zeroed, as long as at least two bytes are not zeroed.
9921 (bswap_replace): Add mask argument and handle masking of bswap
9923 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
9924 caller, punt if cast64_to_32 or mask is not all ones.
9925 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
9926 caller, for now punt if cast64_to_32.
9928 2021-08-23 Richard Biener <rguenther@suse.de>
9930 PR tree-optimization/79334
9931 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
9932 a type also for COMPONENT_REFs.
9933 (vn_reference_may_trap): Check ARRAY_REF with constant index
9934 against the array domain.
9936 2021-08-23 liuhongt <hongtao.liu@intel.com>
9939 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
9940 TARGET_AVX512BW to condition.
9942 2021-08-23 Jakub Jelinek <jakub@redhat.com>
9945 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
9946 register variables already during early_dwarf if possible.
9948 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
9950 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
9952 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
9955 * config/arm/arm.opt: Fix typo.
9956 * config/arm/t-rmprofile: Fix typo.
9958 2021-08-23 Jakub Jelinek <jakub@redhat.com>
9960 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
9961 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
9962 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
9963 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
9964 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
9965 if either grainsize or num_tasks clause has the strict modifier.
9967 2021-08-23 Martin Liska <mliska@suse.cz>
9969 * dbgcnt.def (DEBUG_COUNTER): New counter.
9970 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
9972 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9974 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
9976 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
9979 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
9981 * config/mips/mips.c (mips_msa_output_shift_immediate): New
9983 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
9984 vlshr<mode>3): Call it.
9986 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
9987 Martin Liska <mliska@suse.cz>
9989 PR middle-end/101949
9990 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
9993 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
9995 * config/mips/mips.c (mips_function_rodata_section,
9996 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
9998 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
10000 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
10002 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10004 * configure.ac (thread-local storage support): Remove tls_first_major
10005 and tls_first_minor. Use "$conftest_s" to check support.
10006 * configure: Regenerate.
10008 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10010 * configure.ac: Fixup formatting.
10012 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10014 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
10015 * configure.ac: ... update all callers.
10017 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10020 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
10021 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
10022 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
10023 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
10024 Remove remaining checks for in-tree assembler.
10025 * configure: Regenerate.
10027 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
10029 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
10030 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
10032 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
10033 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
10034 Improve logical shifts by 27 bits for H8/S.
10035 (get_shift_alg): Corresponding changes.
10036 (h8300_option_override): Revert to loops for -Os when profitable.
10038 2021-08-20 Richard Biener <rguenther@suse.de>
10040 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
10042 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
10043 index as group_id when dataref_groups were not computed.
10044 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
10045 we advace to the next BB.
10047 2021-08-20 Jakub Jelinek <jakub@redhat.com>
10049 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
10052 2021-08-20 Martin Liska <mliska@suse.cz>
10054 PR gcov-profile/89961
10055 * gcov.c (make_gcov_file_name): Rewrite using std::string.
10056 (mangle_name): Simplify, do not used the second argument.
10057 (strip_extention): New function.
10058 (get_md5sum): Likewise.
10059 (get_gcov_intermediate_filename): Handle properly -p and -x
10061 (output_gcov_file): Use string type.
10062 (generate_results): Likewise.
10063 (md5sum_to_hex): Remove.
10065 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
10067 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
10068 (UNSPEC_XXSPLTIW): Move to vsx.md.
10069 (UNSPEC_XXSPLTID): Move to vsx.md.
10070 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
10071 (UNSPEC_XXBLEND): Move to vsx.md.
10072 (UNSPEC_XXPERMX): Move to vsx.md.
10073 (VM3): Move to vsx.md.
10074 (VM3_char): Move to vsx.md.
10075 (xxspltiw_v4si): Move to vsx.md.
10076 (xxspltiw_v4sf): Move to vsx.md.
10077 (xxspltiw_v4sf_inst): Move to vsx.md.
10078 (xxspltidp_v2df): Move to vsx.md.
10079 (xxspltidp_v2df_inst): Move to vsx.md.
10080 (xxsplti32dx_v4si_inst): Move to vsx.md.
10081 (xxsplti32dx_v4sf): Move to vsx.md.
10082 (xxsplti32dx_v4sf_inst): Move to vsx.md.
10083 (xxblend_<mode>): Move to vsx.md.
10084 (xxpermx): Move to vsx.md.
10085 (xxpermx_inst): Move to vsx.md.
10086 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
10087 (UNSPEC_XXSPLTIW): Move from altivec.md.
10088 (UNSPEC_XXSPLTID): Move from altivec.md.
10089 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
10090 (UNSPEC_XXBLEND): Move from altivec.md.
10091 (UNSPEC_XXPERMX): Move from altivec.md.
10092 (VM3): Move from altivec.md.
10093 (VM3_char): Move from altivec.md.
10094 (xxspltiw_v4si): Move from altivec.md.
10095 (xxspltiw_v4sf): Move from altivec.md.
10096 (xxspltiw_v4sf_inst): Move from altivec.md.
10097 (xxspltidp_v2df): Move from altivec.md.
10098 (xxspltidp_v2df_inst): Move from altivec.md.
10099 (xxsplti32dx_v4si_inst): Move from altivec.md.
10100 (xxsplti32dx_v4sf): Move from altivec.md.
10101 (xxsplti32dx_v4sf_inst): Move from altivec.md.
10102 (xxblend_<mode>): Move from altivec.md.
10103 (xxpermx): Move from altivec.md.
10104 (xxpermx_inst): Move from altivec.md.
10106 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
10108 * tree-vect-generic.c (expand_vector_operations_1): Use either
10109 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
10110 when constructing scalar splat expressions.
10112 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
10115 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
10116 pointer to __vector_pair *.
10118 2021-08-19 Martin Sebor <msebor@redhat.com>
10120 * gimple-range.cc: Add comments.
10121 * gimple-range.h: Same.
10123 2021-08-19 Martin Sebor <msebor@redhat.com>
10125 PR middle-end/101984
10126 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
10129 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
10131 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
10132 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
10133 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
10135 2021-08-19 Jakub Jelinek <jakub@redhat.com>
10137 PR middle-end/101950
10138 * optabs.c (expand_clrsb_using_clz): New function.
10139 (expand_unop): Use it as another clrsb expansion fallback.
10141 2021-08-19 liuhongt <hongtao.liu@intel.com>
10144 2021-07-28 liuhongt <hongtao.liu@intel.com>
10147 * config/i386/i386.h (processor_costs): Add new member
10149 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
10150 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
10151 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
10152 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
10153 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
10154 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
10155 generic_cost, core_cost): Initialize integer_to_sse same value
10157 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
10158 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10159 Use integer_to_sse instead of sse_op to calculate the cost of
10162 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
10164 * config.gcc: Include rpath.opt for Darwin.
10165 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
10167 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10169 PR bootstrap/101959
10170 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
10173 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
10175 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
10176 pattern and type cast.
10177 (vld4_lane_f32): Use float RTL pattern.
10178 (vld4q_lane_f64): Use float type cast.
10180 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
10182 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
10185 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10187 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
10188 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
10189 (hash_map_tests_c_tests): Call it.
10191 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10193 * ggc.h (enum ggc_collect): New.
10194 (ggc_collect): Use it.
10195 * ggc-page.c: Adjust.
10196 * ggc-common.c: Likewise.
10197 * ggc-tests.c: Likewise.
10198 * read-rtl-function.c: Likewise.
10199 * selftest-run-tests.c: Likewise.
10200 * doc/gty.texi (Invoking the garbage collector): Likewise.
10202 2021-08-18 liuhongt <hongtao.liu@intel.com>
10205 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
10207 * config/i386/sse.md (*sse3_haddv2df3_low): Add
10208 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
10209 (*sse3_hsubv2df3_low): Ditto.
10210 * config/i386/x86-tune.def
10211 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
10213 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10215 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
10216 (gori_compute::compute_operand_range): Add tracing.
10217 (gori_compute::logical_combine): Ditto.
10218 (gori_compute::compute_logical_operands): Ditto.
10219 (gori_compute::compute_operand1_range): Ditto.
10220 (gori_compute::compute_operand2_range): Ditto.
10221 (gori_compute::outgoing_edge_range_p): Ditto.
10222 * gimple-range-gori.h (class gori_compute): Add range_tracer.
10224 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10226 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
10227 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
10228 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
10229 * params.opt (--param=evrp-mode): Adjust options.
10231 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10233 * Makefile.in (OBJS): Add gimple-range-trace.o.
10234 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
10235 * gimple-range-fold.cc: Adjust headers.
10236 * gimple-range-trace.cc: New.
10237 * gimple-range-trace.h: New.
10238 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
10239 (gimple_ranger::range_of_expr): Add tracing.
10240 (gimple_ranger::range_on_entry): Ditto.
10241 (gimple_ranger::range_on_exit): Ditto.
10242 (gimple_ranger::range_on_edge): Ditto.
10243 (gimple_ranger::fold_range_internal): Ditto.
10244 (gimple_ranger::dump_bb): Do not calculate edge range twice.
10245 (trace_ranger::*): Remove.
10246 (enable_ranger): Never create a trace_ranger.
10247 (debug_seed_ranger): Move to gimple-range-trace.cc.
10248 (dump_ranger): Ditto.
10249 (debug_ranger): Ditto.
10250 * gimple-range.h: Include gimple-range-trace.h.
10251 (range_on_entry, range_on_exit): No longer virtual.
10252 (class trace_ranger): Remove.
10253 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
10255 2021-08-17 Martin Sebor <msebor@redhat.com>
10257 PR middle-end/101854
10258 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
10259 in gimple-ssa-warn-access.cc.
10260 * calls.c (alloc_max_size): Move code to check_alloca.
10261 (get_size_range): Move to pointer-query.cc.
10262 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
10263 (get_attr_nonstring_decl): Move to tree.c.
10264 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
10265 (append_attrname): Same.
10266 (maybe_warn_rdwr_sizes): Same.
10267 (initialize_argument_information): Move code to
10268 gimple-ssa-warn-access.cc.
10269 * calls.h (maybe_warn_alloc_args_overflow): Move to
10270 gimple-ssa-warn-access.h.
10271 (get_attr_nonstring_decl): Move to tree.h.
10272 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
10273 (enum size_range_flags): Move to pointer-query.h.
10274 (get_size_range): Same.
10275 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
10276 to avoid Clang -Wunused-function.
10277 (get_size_range): Declare static.
10278 (maybe_emit_free_warning): Rename...
10279 (maybe_check_dealloc_call): ...to this for consistency.
10280 (class pass_waccess): Add members.
10281 (pass_waccess::~pass_waccess): Defined.
10282 (alloc_max_size): Move here from calls.c.
10283 (maybe_warn_alloc_args_overflow): Same.
10284 (check_alloca): New function.
10285 (check_alloc_size_call): New function.
10286 (check_strncat): Handle another warning flag.
10287 (pass_waccess::check_builtin): Handle alloca.
10288 (fntype_argno_type): Move here from calls.c.
10289 (append_attrname): Same.
10290 (maybe_warn_rdwr_sizes): Same.
10291 (pass_waccess::check_call): Define.
10292 (check_nonstring_args): New function.
10293 (pass_waccess::check): Call new member functions.
10294 (pass_waccess::execute): Enable ranger.
10295 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
10296 (maybe_warn_nonstring_arg): Same.
10297 * gimple-ssa-warn-restrict.c: Remove #include.
10298 * pointer-query.cc (get_size_range): Move here from calls.c.
10299 * pointer-query.h (enum size_range_flags): Same.
10300 (get_size_range): Same.
10301 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
10302 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
10304 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
10306 * ggc.h (ggc_collect): Add 'force_collect' parameter.
10307 * ggc-page.c (ggc_collect): Use that one instead of global
10308 'ggc_force_collect'. Adjust all users.
10309 * doc/gty.texi (Invoking the garbage collector): Update.
10310 * ggc-internal.h (ggc_force_collect): Remove.
10311 * ggc-common.c (ggc_force_collect): Likewise.
10312 * selftest.h (forcibly_ggc_collect): Remove.
10313 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
10314 * read-rtl-function.c (test_loading_labels): Adjust.
10315 * selftest-run-tests.c (run_tests): Likewise.
10317 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10319 * config/darwin.c (darwin_file_end): Reset and reclaim the
10320 section names table at the end of compile.
10322 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10325 * config.in: Regenerate.
10326 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
10327 (ASM_SPEC): Pass options to disable branch shortening where
10329 * configure: Regenerate.
10330 * configure.ac: Detect versions of 'as' that support the
10331 optimisation which has the bug.
10333 2021-08-17 Richard Biener <rguenther@suse.de>
10335 * optabs-query.c (supports_vec_gather_load_p): Also check
10337 (supports_vec_scatter_store_p): Likewise.
10338 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
10339 back to masked variants if non-masked are not supported.
10340 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
10341 When we need to use masked gather/scatter but do not have
10342 a mask set up a constant true one.
10343 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
10344 non-SSA_NAME masks.
10346 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10348 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
10349 algorithm as PLUS_EXPR to improve subtraction bit bounds.
10350 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
10352 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10354 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
10355 calculate the mask-value pair result of a multiplication by an
10357 (bit_value_binop) [MULT_EXPR]: Call it from here for
10358 multiplications by (sparse) non-negative constants.
10360 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
10363 * config.gcc (gcc_cv_initfini_array): Leave undefined for
10364 uclinuxfdpiceabi targets.
10366 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10368 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
10369 reverse debug stmts.
10371 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10373 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
10375 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10377 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
10378 (__LD4Q_LANE_FUNC): Likewise.
10379 (vld4_lane_u8): Define without macro.
10380 (vld4_lane_u16): Likewise.
10381 (vld4_lane_u32): Likewise.
10382 (vld4_lane_u64): Likewise.
10383 (vld4_lane_s8): Likewise.
10384 (vld4_lane_s16): Likewise.
10385 (vld4_lane_s32): Likewise.
10386 (vld4_lane_s64): Likewise.
10387 (vld4_lane_f16): Likewise.
10388 (vld4_lane_f32): Likewise.
10389 (vld4_lane_f64): Likewise.
10390 (vld4_lane_p8): Likewise.
10391 (vld4_lane_p16): Likewise.
10392 (vld4_lane_p64): Likewise.
10393 (vld4q_lane_u8): Likewise.
10394 (vld4q_lane_u16): Likewise.
10395 (vld4q_lane_u32): Likewise.
10396 (vld4q_lane_u64): Likewise.
10397 (vld4q_lane_s8): Likewise.
10398 (vld4q_lane_s16): Likewise.
10399 (vld4q_lane_s32): Likewise.
10400 (vld4q_lane_s64): Likewise.
10401 (vld4q_lane_f16): Likewise.
10402 (vld4q_lane_f32): Likewise.
10403 (vld4q_lane_f64): Likewise.
10404 (vld4q_lane_p8): Likewise.
10405 (vld4q_lane_p16): Likewise.
10406 (vld4q_lane_p64): Likewise.
10407 (vld4_lane_bf16): Likewise.
10408 (vld4q_lane_bf16): Likewise.
10410 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10412 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
10413 (__LD3Q_LANE_FUNC): Delete.
10414 (vld3_lane_u8): Define without macro.
10415 (vld3_lane_u16): Likewise.
10416 (vld3_lane_u32): Likewise.
10417 (vld3_lane_u64): Likewise.
10418 (vld3_lane_s8): Likewise.
10419 (vld3_lane_s16): Likewise.
10420 (vld3_lane_s32): Likewise.
10421 (vld3_lane_s64): Likewise.
10422 (vld3_lane_f16): Likewise.
10423 (vld3_lane_f32): Likewise.
10424 (vld3_lane_f64): Likewise.
10425 (vld3_lane_p8): Likewise.
10426 (vld3_lane_p16): Likewise.
10427 (vld3_lane_p64): Likewise.
10428 (vld3q_lane_u8): Likewise.
10429 (vld3q_lane_u16): Likewise.
10430 (vld3q_lane_u32): Likewise.
10431 (vld3q_lane_u64): Likewise.
10432 (vld3q_lane_s8): Likewise.
10433 (vld3q_lane_s16): Likewise.
10434 (vld3q_lane_s32): Likewise.
10435 (vld3q_lane_s64): Likewise.
10436 (vld3q_lane_f16): Likewise.
10437 (vld3q_lane_f32): Likewise.
10438 (vld3q_lane_f64): Likewise.
10439 (vld3q_lane_p8): Likewise.
10440 (vld3q_lane_p16): Likewise.
10441 (vld3q_lane_p64): Likewise.
10442 (vld3_lane_bf16): Likewise.
10443 (vld3q_lane_bf16): Likewise.
10445 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10447 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
10448 (__LD2Q_LANE_FUNC): Likewise.
10449 (vld2_lane_u8): Define without macro.
10450 (vld2_lane_u16): Likewise.
10451 (vld2_lane_u32): Likewise.
10452 (vld2_lane_u64): Likewise.
10453 (vld2_lane_s8): Likewise.
10454 (vld2_lane_s16): Likewise.
10455 (vld2_lane_s32): Likewise.
10456 (vld2_lane_s64): Likewise.
10457 (vld2_lane_f16): Likewise.
10458 (vld2_lane_f32): Likewise.
10459 (vld2_lane_f64): Likewise.
10460 (vld2_lane_p8): Likewise.
10461 (vld2_lane_p16): Likewise.
10462 (vld2_lane_p64): Likewise.
10463 (vld2q_lane_u8): Likewise.
10464 (vld2q_lane_u16): Likewise.
10465 (vld2q_lane_u32): Likewise.
10466 (vld2q_lane_u64): Likewise.
10467 (vld2q_lane_s8): Likewise.
10468 (vld2q_lane_s16): Likewise.
10469 (vld2q_lane_s32): Likewise.
10470 (vld2q_lane_s64): Likewise.
10471 (vld2q_lane_f16): Likewise.
10472 (vld2q_lane_f32): Likewise.
10473 (vld2q_lane_f64): Likewise.
10474 (vld2q_lane_p8): Likewise.
10475 (vld2q_lane_p16): Likewise.
10476 (vld2q_lane_p64): Likewise.
10477 (vld2_lane_bf16): Likewise.
10478 (vld2q_lane_bf16): Likewise.
10480 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10482 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
10485 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10487 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
10488 (rank_for_schedule): Use it.
10490 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10492 PR rtl-optimization/91598
10493 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
10494 insns after memory reads and before memory writes.
10496 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
10498 * rtl.h (CONST_VECTOR_P): New macro.
10499 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
10500 code testing macros.
10501 (aarch64_ptrue_all_mode): Likewise.
10502 (aarch64_expand_mov_immediate): Likewise.
10503 (aarch64_const_vec_all_in_range_p): Likewise.
10504 (aarch64_rtx_costs): Likewise.
10505 (aarch64_legitimate_constant_p): Likewise.
10506 (aarch64_simd_valid_immediate): Likewise.
10507 (aarch64_simd_make_constant): Likewise.
10508 (aarch64_convert_mult_to_shift): Likewise.
10509 (aarch64_expand_sve_vec_perm): Likewise.
10510 (aarch64_vec_fpconst_pow_of_2): Likewise.
10512 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10514 PR tree-optimization/101938
10515 * range-op.cc (operator_abs::op1_range): Special case
10516 -TYPE_MIN_VALUE for flag_wrapv.
10518 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
10520 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
10523 2021-08-17 Jakub Jelinek <jakub@redhat.com>
10525 * tree.def (OMP_SCOPE): New tree code.
10526 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
10527 * tree-nested.c (convert_nonlocal_reference_stmt,
10528 convert_local_reference_stmt, convert_gimple_call): Handle
10530 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
10531 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
10532 * gimple.c (gimple_build_omp_scope): New function.
10533 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
10534 * gimple.h (gimple_build_omp_scope): Declare.
10535 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
10536 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
10537 gimple_omp_scope_set_clauses): New inline functions.
10538 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
10539 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
10540 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
10541 * gimple-walk.c (walk_gimple_stmt): Likewise.
10542 * gimple-low.c (lower_stmt): Likewise.
10543 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10544 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
10545 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
10546 allowing task reductions. Reject inscan reductions on scope.
10547 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
10548 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
10549 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
10550 (estimate_num_insns): Likewise.
10551 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
10552 contexts if var isn't privatized there.
10553 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
10554 (scan_omp_1_stmt): Likewise.
10555 (maybe_add_implicit_barrier_cancel): Look through outer
10557 (lower_omp_scope): New function.
10558 (lower_omp_task_reductions): Handle OMP_SCOPE.
10559 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
10560 (diagnose_sb_1, diagnose_sb_2): Likewise.
10561 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
10562 (expand_omp): Handle GIMPLE_OMP_SCOPE.
10563 (omp_make_gimple_edges): Likewise.
10564 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
10566 2021-08-17 Richard Biener <rguenther@suse.de>
10568 PR tree-optimization/101925
10569 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
10570 reverse on COMPONENT_REF and ARRAY_REF according to
10571 what reverse_storage_order_for_component_p does.
10572 (vn_reference_eq): Compare reversed on reference ops.
10573 (reverse_storage_order_for_component_p): New overload.
10574 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
10575 on the reference looked up.
10577 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
10579 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
10580 shifts on the H8/S.
10581 (h8300_option_override): Use loops on H8/S more often when optimizing
10583 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
10584 accordingly. Handle various arithmetic right shifts with special
10585 sequences that we couldn't handle before.
10587 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
10589 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
10591 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
10593 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
10594 * config/sparc/sparc.c (sparc_gcov_type_size): New.
10595 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
10596 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
10597 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
10598 * doc/tm.texi.in: Regenerate.
10599 * target.def (gcov_type_size): New target hook.
10600 * targhooks.c (default_gcov_type_size): New.
10601 * targhooks.h (default_gcov_type_size): Declare.
10602 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
10604 (gimple_gen_time_profiler): Likewise.
10606 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
10608 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
10610 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
10612 PR middle-end/101931
10613 * omp-low.c (omp_runtime_api_call): Update for routines
10614 added in the meanwhile.
10616 2021-08-16 Martin Liska <mliska@suse.cz>
10618 PR tree-optimization/100393
10619 * tree-switch-conversion.c (group_cluster::dump): Use
10620 get_comparison_count.
10621 (jump_table_cluster::find_jump_tables): Pre-compute number of
10622 comparisons and then decrement it. Cache also max_ratio.
10623 (jump_table_cluster::can_be_handled): Change signature.
10624 * tree-switch-conversion.h (get_comparison_count): New.
10626 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
10628 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
10629 encodings only when -fgnat-encodings=all is specified.
10630 (add_bound_info): Likewise.
10631 (add_byte_size_attribute): Likewise.
10632 (gen_member_die): Likewise.
10634 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10636 * omp-oacc-neuter-broadcast.cc
10637 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
10639 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10641 * omp-oacc-neuter-broadcast.cc
10642 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
10645 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10647 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
10648 (execute_omp_oacc_neuter_broadcast): ... here.
10649 (install_var_field, build_receiver_ref, build_sender_ref): Take
10650 'field_map_t *' parameter. Adjust all users.
10651 (worker_single_copy, neuter_worker_single): Take a
10652 'record_field_map_t *' parameter. Adjust all users.
10654 2021-08-16 liuhongt <hongtao.liu@intel.com>
10657 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
10660 2021-08-16 Martin Liska <mliska@suse.cz>
10663 * multiple_target.c (create_dispatcher_calls): Make default
10664 function local only if it is a definition.
10666 2021-08-16 Martin Liska <mliska@suse.cz>
10669 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
10670 consider equal SSA_NAMEs when one is a param.
10672 2021-08-16 liuhongt <hongtao.liu@intel.com>
10675 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
10676 Support vpermi2b for V32QI/V16QImode.
10677 (ix86_extract_perm_from_pool_constant): New function.
10678 (ix86_expand_vec_one_operand_perm_avx512): Support
10679 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
10680 (expand_vec_perm_1): Adjust comments for upper.
10681 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
10683 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
10684 (pshufb_truncv4siv4hi_operand): Ditto.
10685 (pshufb_truncv8hiv8qi_operand): Ditto.
10686 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
10687 New pre_reload define_insn_and_split.
10688 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
10689 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
10690 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
10691 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
10692 (*avx512f_permvar_truncv4div4si_1): Ditto.
10693 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
10694 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
10695 (*avx512f_pshufd_truncv2div2si_1): Ditto.
10697 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
10699 * config/riscv/multilib-generator: Support code model option for
10701 * doc/install.texi: Add document of new option for
10702 --with-multilib-generator.
10704 2021-08-15 Clément Chigot <clement.chigot@atos.net>
10706 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
10707 (rs6000_legitimize_tls_address_aix): Use it.
10708 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
10709 xcoff_tls_exec_model_detected is true.
10711 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
10713 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
10714 to allow a bit more code growth, saving many dozens of cycles.
10715 (h8300_option_override): Adjus shift_alg_si if optimizing for
10717 (get_shift_alg): Use special + inline shifts for residuals
10720 2021-08-14 Stafford Horne <shorne@gmail.com>
10723 * config/or1k/or1k-opts.h: New file.
10724 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
10725 Support generating gotha relocations if -mcmodel=large is
10727 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
10729 * config/or1k/or1k.opt (mcmodel=): New option.
10730 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
10732 2021-08-14 Martin Sebor <msebor@redhat.com>
10734 PR middle-end/101791
10735 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
10736 to valid_new_delete_pair_p.
10737 * tree.c (valid_new_delete_pair_p): Add argument.
10738 * tree.h (valid_new_delete_pair_p): Same.
10740 2021-08-14 Jakub Jelinek <jakub@redhat.com>
10743 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
10744 <case E_V64QImode>: For this mode assert
10745 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
10747 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
10750 * config/rs6000/altivec.md (xxeval): Use register_predicate
10751 instead of altivec_register_predicate.
10753 2021-08-13 Martin Sebor <msebor@redhat.com>
10755 PR middle-end/101734
10756 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
10757 (maybe_warn_operand): Call it.
10759 2021-08-13 Martin Liska <mliska@suse.cz>
10762 * attribs.c (decl_attributes): Make naked functions "noipa"
10765 2021-08-13 Martin Liska <mliska@suse.cz>
10768 * symtab.c (symtab_node::noninterposable_alias): Do not create
10769 local aliases for target_clone functions as the clonning pass
10772 2021-08-13 Martin Liska <mliska@suse.cz>
10774 * opts.c (LIVE_PATCHING_OPTION): Define.
10775 (control_options_for_live_patching): Use it in error messages.
10777 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
10779 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
10780 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
10781 ignore_stores_eaf_flags): New constants.
10782 (remove_useless_eaf_flags): New function.
10783 (eaf_flags_useful_p): Use it.
10784 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
10786 (modref_lattice::init): Add EAF_NOREAD.
10787 (modref_lattice::add_escape_point): Do not reacord escape point if
10789 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
10790 use remove_useless_eaf_flags.
10791 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
10792 (modref_lattice::merge_direct_load): Add EAF_NOREAD
10793 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
10794 (analyze_parms): Use remove_useless_eaf_flags.
10795 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
10796 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
10797 use remove_useless_eaf_flags.
10798 (modref_propagate_flags_in_scc): Update.
10799 * ipa-modref.h: Turn eaf_flags_t back to char.
10800 * tree-core.h (EAF_NOT_RETURNED): Fix.
10801 (EAF_NOREAD): New constant
10802 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
10804 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
10805 (handle_pure_call): Likewise.
10807 2021-08-12 Jakub Jelinek <jakub@redhat.com>
10809 * tree.def (OMP_MASKED): New tree code.
10810 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
10811 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
10812 OMP_CLAUSE_FILTER_EXPR): Define.
10813 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
10814 (omp_clause_code_name): Likewise.
10815 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
10816 * tree-nested.c (convert_nonlocal_omp_clauses,
10817 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
10818 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
10819 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
10820 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
10821 (dump_generic_node): Handle OMP_MASTER.
10822 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
10823 * gimple.c (gimple_build_omp_masked): New function.
10824 (gimple_copy): Handle GIMPLE_OMP_MASKED.
10825 * gimple.h (gimple_build_omp_masked): Declare.
10826 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
10827 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
10828 gimple_omp_masked_set_clauses): New inline functions.
10829 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
10830 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
10831 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
10832 * gimple-walk.c (walk_gimple_stmt): Likewise.
10833 * gimple-low.c (lower_stmt): Likewise.
10834 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10835 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
10836 that take one expression rather than decl or constant, force
10837 gimplification of that into a SSA_NAME or temporary unless min
10839 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
10840 (gimplify_expr): Handle OMP_MASKED.
10841 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
10842 (estimate_num_insns): Likewise.
10843 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
10844 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
10845 diagnostics for existence of masked construct.
10846 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
10847 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
10848 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
10851 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
10854 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
10855 (ldexp<mode>3): Use avx512f_scalef<mode>2.
10856 (UNSPEC_SCALEF): Move from ...
10857 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
10859 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
10861 * ipa-split.c (consider_split): Fix condition testing void functions.
10863 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
10865 * doc/invoke.texi: Remove docs for threader-mode param.
10866 * flag-types.h (enum threader_mode): Remove.
10867 * params.opt: Remove threader-mode param.
10868 * tree-ssa-threadbackward.c (class back_threader): Remove
10869 path_is_unreachable_p.
10870 Make find_paths private.
10871 Add maybe_thread and thread_through_all_blocks.
10872 Remove reference marker for m_registry.
10873 Remove reference marker for m_profit.
10874 (back_threader::back_threader): Adjust for registry and profit not
10876 (dump_path): Move down.
10877 (debug): Move down.
10878 (class thread_jumps): Remove.
10879 (class back_threader_registry): Remove m_all_paths.
10881 (thread_jumps::thread_through_all_blocks): Move to back_threader
10883 (fsm_find_thread_path): Remove
10884 (back_threader::maybe_thread): New.
10885 (back_threader::thread_through_all_blocks): Move from
10887 (back_threader_registry::back_threader_registry): Remove
10889 (back_threader_registry::~back_threader_registry): Remove.
10890 (thread_jumps::find_taken_edge): Remove.
10891 (thread_jumps::check_subpath_and_update_thread_path): Remove.
10892 (thread_jumps::maybe_register_path): Remove.
10893 (thread_jumps::handle_phi): Remove.
10894 (handle_assignment_p): Remove.
10895 (thread_jumps::handle_assignment): Remove.
10896 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
10897 (thread_jumps::find_jump_threads_backwards): Remove.
10898 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
10899 (try_thread_blocks): Rename find_jump_threads_backwards to
10901 (pass_early_thread_jumps::execute): Same.
10903 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
10905 * tree-core.h (omp_clause_proc_bind_kind): Add
10906 OMP_CLAUSE_PROC_BIND_PRIMARY.
10907 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
10908 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
10910 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
10912 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
10913 fno-common reference.
10914 * config/arc/arc.c (arc_override_options): Remove overriding of
10917 2021-08-12 Jakub Jelinek <jakub@redhat.com>
10920 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
10921 If d->testing_p, return true after performing checks instead of
10922 actually expanding the insn.
10923 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
10924 !TARGET_AVX512BW and return false.
10926 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
10928 * configure.ac (PE linker --disable-dynamicbase support): New check.
10929 * configure: Regenerate.
10930 * config.in: Likewise.
10931 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
10932 (LINK_SPEC): Use it.
10933 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
10934 (LINK_SPEC): Likewise.
10936 2021-08-12 liuhongt <hongtao.liu@intel.com>
10939 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
10940 post_reload define_insn_and_split.
10941 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
10942 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
10943 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
10944 (*avx2_zero_extendv8hiv8si2_2): Ditto.
10945 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
10946 (*avx512f_zero_extendv8siv8di2_2): Ditto.
10947 (*avx2_zero_extendv4siv4di2_2): Ditto.
10948 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
10949 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
10952 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
10954 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
10957 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
10959 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
10961 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
10963 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
10964 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
10965 initialization of pcvoid_type_node here...
10966 (altivec_init_builtins): ...from here.
10967 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
10968 RS6000_BTI_const_ptr_void.
10969 (pcvoid_type_node): New macro.
10971 2021-08-11 Richard Biener <rguenther@suse.de>
10974 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
10975 hard-register accesses.
10977 2021-08-11 Richard Biener <rguenther@suse.de>
10979 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
10980 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
10981 to determine has_volatile_ops.
10983 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
10985 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
10987 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
10988 Alexandre Oliva <oliva@adacore.com>
10990 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
10992 (modref_lattice::merge_deref): Fix handling of indirect scape points.
10993 (update_escape_summary_1): Likewise.
10994 (update_escape_summary): Likewise.
10995 (ipa_merge_modref_summary_after_inlining): Likewise.
10997 2021-08-11 Richard Biener <rguenther@suse.de>
10999 PR middle-end/101858
11000 * fold-const.c (fold_binary_loc): Guard simplification
11001 of X < (cast) (1 << Y) to integer types.
11003 2021-08-11 Richard Biener <rguenther@suse.de>
11005 PR tree-optimization/101861
11006 * tree-vect-stmts.c (vectorizable_load): Fix error in
11007 previous change with regard to gather vectorization.
11009 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11012 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
11014 (vdup_n_s16): Likewise.
11015 (vdup_n_s32): Likewise.
11016 (vdup_n_s64): Likewise.
11017 (vdup_n_u8): Likewise.
11018 (vdup_n_u16): Likewise.
11019 (vdup_n_u32): Likewise.
11020 (vdup_n_u64): Likewise.
11021 (vdup_n_p8): Likewise.
11022 (vdup_n_p16): Likewise.
11023 (vdup_n_p64): Likewise.
11024 (vdup_n_f16): Likewise.
11025 (vdup_n_f32): Likewise.
11026 (vdupq_n_s8): Likewise.
11027 (vdupq_n_s16): Likewise.
11028 (vdupq_n_s32): Likewise.
11029 (vdupq_n_s64): Likewise.
11030 (vdupq_n_u8): Likewise.
11031 (vdupq_n_u16): Likewise.
11032 (vdupq_n_u32): Likewise.
11033 (vdupq_n_u64): Likewise.
11034 (vdupq_n_p8): Likewise.
11035 (vdupq_n_p16): Likewise.
11036 (vdupq_n_p64): Likewise.
11037 (vdupq_n_f16): Likewise.
11038 (vdupq_n_f32): Likewise.
11039 (vmov_n_s8): Replace call to builtin with call to corresponding
11041 (vmov_n_s16): Likewise.
11042 (vmov_n_s32): Likewise.
11043 (vmov_n_s64): Likewise.
11044 (vmov_n_u8): Likewise.
11045 (vmov_n_u16): Likewise.
11046 (vmov_n_u32): Likewise.
11047 (vmov_n_u64): Likewise.
11048 (vmov_n_p8): Likewise.
11049 (vmov_n_p16): Likewise.
11050 (vmov_n_f16): Likewise.
11051 (vmov_n_f32): Likewise.
11052 (vmovq_n_s8): Likewise.
11053 (vmovq_n_s16): Likewise.
11054 (vmovq_n_s32): Likewise.
11055 (vmovq_n_s64): Likewise.
11056 (vmovq_n_u8): Likewise.
11057 (vmovq_n_u16): Likewise.
11058 (vmovq_n_u32): Likewise.
11059 (vmovq_n_u64): Likewise.
11060 (vmovq_n_p8): Likewise.
11061 (vmovq_n_p16): Likewise.
11062 (vmovq_n_f16): Likewise.
11063 (vmovq_n_f32): Likewise.
11064 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
11066 2021-08-11 liuhongt <hongtao.liu@intel.com>
11069 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
11070 when TARGET_AVX512F and TARGET_SSE_MATH.
11072 2021-08-10 Jakub Jelinek <jakub@redhat.com>
11075 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
11076 for V32HImode if !TARGET_AVX512BW.
11077 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
11078 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
11079 early, but actually check the permutation.
11081 2021-08-10 Richard Biener <rguenther@suse.de>
11083 PR tree-optimization/101809
11084 * tree-vect-stmts.c (get_load_store_type): Allow emulated
11085 gathers with offset vector nunits being a constant multiple
11086 of the data vector nunits.
11087 (vect_get_gather_scatter_ops): Use the appropriate nunits
11088 for the offset vector defs.
11089 (vectorizable_store): Adjust call to
11090 vect_get_gather_scatter_ops.
11091 (vectorizable_load): Likewise. Handle the case of less
11092 offset vectors than data vectors.
11094 2021-08-10 Jakub Jelinek <jakub@redhat.com>
11097 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
11098 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
11101 2021-08-10 Richard Biener <rguenther@suse.de>
11103 PR tree-optimization/101801
11104 PR tree-optimization/101819
11105 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
11106 * tree-vect-loop.c (vect_emulated_vector_p): New function.
11107 (vectorizable_reduction): Re-instantiate a check for emulated
11109 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11110 (vectorizable_operation): Likewise. Cost emulated vector
11111 operations according to the scalar sequence synthesized by
11114 2021-08-10 Richard Biener <rguenther@suse.de>
11116 PR middle-end/101824
11117 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
11118 volatile in case the variable was.
11120 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
11123 * config/i386/constraints.md (BC): Document for integer SSE
11124 constant all bits set operand.
11125 (BF): New constraint for const floating-point all bits set
11127 * config/i386/i386.c (standard_sse_constant_p): Likewise.
11128 (standard_sse_constant_opcode): Likewise.
11129 * config/i386/sse.md (sseconstm1): New mode attribute.
11130 (mov<mode>_internal): Replace BC with <sseconstm1>.
11132 2021-08-10 liuhongt <hongtao.liu@intel.com>
11134 * config/i386/sse.md (cond_<insn><mode>): New expander.
11135 (VI248_AVX512VLBW): New mode iterator.
11136 * config/i386/predicates.md
11137 (nonimmediate_or_const_vec_dup_operand): New predicate.
11139 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
11141 PR tree-optimization/101741
11142 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
11143 type of parameter for toupper/tolower.
11145 2021-08-09 Martin Jambor <mjambor@suse.cz>
11147 PR testsuite/101654
11148 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
11150 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
11152 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
11154 (is_store_insn1): Verify source is a register.
11156 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
11159 * config/i386/mmx.md (<any_logic:code>v2sf3):
11160 Rename from *mmx_<any_logic:code>v2sf3
11162 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11164 * config/nvptx/nvptx.c: Cross-reference parts adapted in
11165 'gcc/omp-oacc-neuter-broadcast.cc'.
11166 * omp-low.c: Likewise.
11167 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
11170 2021-08-09 Julian Brown <julian@codesourcery.com>
11171 Kwok Cheung Yeung <kcy@codesourcery.com>
11172 Thomas Schwinge <thomas@codesourcery.com>
11174 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
11175 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
11176 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
11177 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
11178 (gcn_fork_join): Update comment.
11179 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
11180 (macc_experimental_workers): Remove unused option.
11182 2021-08-09 Julian Brown <julian@codesourcery.com>
11183 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
11184 Kwok Cheung Yeung <kcy@codesourcery.com>
11185 Thomas Schwinge <thomas@codesourcery.com>
11187 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
11188 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
11189 Add documentation hook.
11190 * doc/tm.texi: Regenerate.
11191 * omp-oacc-neuter-broadcast.cc: New file.
11192 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
11193 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
11194 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
11195 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
11196 * target.def (goacc.create_worker_broadcast_record): Add target
11198 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
11200 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
11201 Rename prototype to...
11202 (gcn_goacc_create_worker_broadcast_record): ... this.
11203 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
11205 (gcn_goacc_create_worker_broadcast_record): ... this.
11206 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
11208 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
11210 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
11213 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
11214 the right iterator.
11216 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11218 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
11220 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11222 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
11223 consider '-foffload-abi'.
11224 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
11225 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
11228 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11230 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
11233 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11235 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
11237 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11239 * doc/gty.texi (Files): Update.
11241 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11243 * doc/gty.texi (Files): Fix GTY header file example.
11245 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
11247 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
11248 determine the upper and lower bounds from a mask-value pair.
11249 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
11250 absolute value and unsigned absolute value expressions.
11251 (bit_value_binop): Initialize *VAL's precision.
11252 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
11253 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
11254 support when the operands are unknown but potentially equal.
11255 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
11257 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
11259 * config/aarch64/aarch64.md
11260 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
11262 2021-08-08 Sergei Trofimovich <siarheit@google.com>
11264 * lra-constraints.c: Fix s/otput/output/ typo.
11266 2021-08-06 Martin Sebor <msebor@redhat.com>
11268 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
11269 (expand_builtin_strcat): Same.
11270 (expand_builtin_stpncpy): Same.
11271 (expand_builtin_strncat): Same.
11272 (check_read_access): Same.
11273 (check_memop_access): Same.
11274 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
11275 (expand_builtin_strnlen): Same.
11276 (expand_builtin_memcpy): Same.
11277 (expand_builtin_memmove): Same.
11278 (expand_builtin_mempcpy): Same.
11279 (expand_builtin_strcpy): Same.
11280 (expand_builtin_strcpy_args): Same.
11281 (expand_builtin_stpcpy_1): Same.
11282 (expand_builtin_strncpy): Same.
11283 (expand_builtin_memset): Same.
11284 (expand_builtin_bzero): Same.
11285 (expand_builtin_strcmp): Same.
11286 (expand_builtin_strncmp): Same.
11287 (expand_builtin): Remove handlers.
11288 (fold_builtin_strlen): Add a comment.
11289 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
11290 * calls.c (maybe_warn_nonstring_arg): Same.
11291 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
11292 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
11293 (gimple_fold_builtin_stpcpy): Same.
11294 * gimple-ssa-warn-access.cc (has_location): New function.
11295 (get_location): Same.
11296 (get_callee_fndecl): Same.
11297 (call_nargs): Same.
11299 (warn_string_no_nul): Define.
11300 (unterminated_array): Same.
11301 (check_nul_terminated_array): Same.
11302 (maybe_warn_nonstring_arg): Same.
11303 (maybe_warn_for_bound): Same.
11304 (warn_for_access): Same.
11305 (check_access): Same.
11306 (check_memop_access): Same.
11307 (check_read_access): Same.
11308 (warn_dealloc_offset): Use helper functions.
11309 (maybe_emit_free_warning): Same.
11310 (class pass_waccess): Add members.
11311 (check_strcat): New function.
11312 (check_strncat): New function.
11313 (check_stxcpy): New function.
11314 (check_stxncpy): New function.
11315 (check_strncmp): New function.
11316 (pass_waccess::check_builtin): New function.
11317 (pass_waccess::check): Call it.
11318 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
11320 (maybe_warn_for_bound): Same.
11321 (check_access): Same.
11322 (check_memop_access): Same.
11323 (check_read_access): Same.
11324 * pointer-query.h (struct access_data): Define a ctor overload.
11326 2021-08-06 Richard Biener <rguenther@suse.de>
11328 PR tree-optimization/101801
11329 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
11330 (vect_can_vectorize_without_simd_p): ... to this.
11331 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
11332 (vect_can_vectorize_without_simd_p): ... to this and fold
11333 in vect_min_worthwhile_factor.
11334 (vect_min_worthwhile_factor): Remove.
11335 (vectorizable_reduction): Adjust and remove the cost part.
11336 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11337 (vectorizable_operation): Likewise.
11339 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
11342 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
11343 Add general_gr_operand predicate to operand 3.
11345 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
11347 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
11348 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
11350 2021-08-06 Tamar Christina <tamar.christina@arm.com>
11352 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
11353 register_svprfop): Pass vec<> by pointer.
11354 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
11355 * langhooks.c (lhd_simulate_enum_decl): Likewise.
11356 * langhooks.h (struct lang_hooks_for_types): Likewise.
11358 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11360 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
11361 __builtin_memcpy instead of constructing an additional
11362 __builtin_aarch64_simd_oi one vector at a time.
11363 (vst1q_bf16_x2): Likewise.
11364 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
11365 an additional __builtin_aarch64_simd_ci one vector at a time.
11366 (vst1q_bf16_x3): Likewise.
11367 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
11368 (vst1q_bf16_x4): Likewise.
11369 (vst2_bf16): Use __builtin_memcpy instead of constructing an
11370 additional __builtin_aarch64_simd_oi one vector at a time.
11371 (vst2q_bf16): Likewise.
11372 (vst3_bf16): Use __builtin_memcpy instead of constructing an
11373 additional __builtin_aarch64_simd_ci mode one vector at a
11375 (vst3q_bf16): Likewise.
11376 (vst4_bf16): Use __builtin_memcpy instead of constructing an
11377 additional __builtin_aarch64_simd_xi one vector at a time.
11378 (vst4q_bf16): Likewise.
11380 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11382 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
11383 (__ST2Q_LANE_FUNC): Delete.
11384 (vst2_lane_f16): Use __builtin_memcpy to copy vector
11385 structure instead of constructing __builtin_aarch64_simd_oi
11386 one vector at a time.
11387 (vst2_lane_f32): Likewise.
11388 (vst2_lane_f64): Likewise.
11389 (vst2_lane_p8): Likewise.
11390 (vst2_lane_p16): Likewise.
11391 (vst2_lane_p64): Likewise.
11392 (vst2_lane_s8): Likewise.
11393 (vst2_lane_s16): Likewise.
11394 (vst2_lane_s32): Likewise.
11395 (vst2_lane_s64): Likewise.
11396 (vst2_lane_u8): Likewise.
11397 (vst2_lane_u16): Likewise.
11398 (vst2_lane_u32): Likewise.
11399 (vst2_lane_u64): Likewise.
11400 (vst2_lane_bf16): Likewise.
11401 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
11402 structure instead of using a union.
11403 (vst2q_lane_f32): Likewise.
11404 (vst2q_lane_f64): Likewise.
11405 (vst2q_lane_p8): Likewise.
11406 (vst2q_lane_p16): Likewise.
11407 (vst2q_lane_p64): Likewise.
11408 (vst2q_lane_s8): Likewise.
11409 (vst2q_lane_s16): Likewise.
11410 (vst2q_lane_s32): Likewise.
11411 (vst2q_lane_s64): Likewise.
11412 (vst2q_lane_u8): Likewise.
11413 (vst2q_lane_u16): Likewise.
11414 (vst2q_lane_u32): Likewise.
11415 (vst2q_lane_u64): Likewise.
11416 (vst2q_lane_bf16): Likewise.
11418 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11420 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
11421 (__ST3Q_LANE_FUNC): Delete.
11422 (vst3_lane_f16): Use __builtin_memcpy to copy vector
11423 structure instead of constructing __builtin_aarch64_simd_ci
11424 one vector at a time.
11425 (vst3_lane_f32): Likewise.
11426 (vst3_lane_f64): Likewise.
11427 (vst3_lane_p8): Likewise.
11428 (vst3_lane_p16): Likewise.
11429 (vst3_lane_p64): Likewise.
11430 (vst3_lane_s8): Likewise.
11431 (vst3_lane_s16): Likewise.
11432 (vst3_lane_s32): Likewise.
11433 (vst3_lane_s64): Likewise.
11434 (vst3_lane_u8): Likewise.
11435 (vst3_lane_u16): Likewise.
11436 (vst3_lane_u32): Likewise.
11437 (vst3_lane_u64): Likewise.
11438 (vst3_lane_bf16): Likewise.
11439 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
11440 structure instead of using a union.
11441 (vst3q_lane_f32): Likewise.
11442 (vst3q_lane_f64): Likewise.
11443 (vst3q_lane_p8): Likewise.
11444 (vst3q_lane_p16): Likewise.
11445 (vst3q_lane_p64): Likewise.
11446 (vst3q_lane_s8): Likewise.
11447 (vst3q_lane_s16): Likewise.
11448 (vst3q_lane_s32): Likewise.
11449 (vst3q_lane_s64): Likewise.
11450 (vst3q_lane_u8): Likewise.
11451 (vst3q_lane_u16): Likewise.
11452 (vst3q_lane_u32): Likewise.
11453 (vst3q_lane_u64): Likewise.
11454 (vst3q_lane_bf16): Likewise.
11456 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11458 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
11459 (__ST4Q_LANE_FUNC): Delete.
11460 (vst4_lane_f16): Use __builtin_memcpy to copy vector
11461 structure instead of constructing __builtin_aarch64_simd_xi
11462 one vector at a time.
11463 (vst4_lane_f32): Likewise.
11464 (vst4_lane_f64): Likewise.
11465 (vst4_lane_p8): Likewise.
11466 (vst4_lane_p16): Likewise.
11467 (vst4_lane_p64): Likewise.
11468 (vst4_lane_s8): Likewise.
11469 (vst4_lane_s16): Likewise.
11470 (vst4_lane_s32): Likewise.
11471 (vst4_lane_s64): Likewise.
11472 (vst4_lane_u8): Likewise.
11473 (vst4_lane_u16): Likewise.
11474 (vst4_lane_u32): Likewise.
11475 (vst4_lane_u64): Likewise.
11476 (vst4_lane_bf16): Likewise.
11477 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
11478 structure instead of using a union.
11479 (vst4q_lane_f32): Likewise.
11480 (vst4q_lane_f64): Likewise.
11481 (vst4q_lane_p8): Likewise.
11482 (vst4q_lane_p16): Likewise.
11483 (vst4q_lane_p64): Likewise.
11484 (vst4q_lane_s8): Likewise.
11485 (vst4q_lane_s16): Likewise.
11486 (vst4q_lane_s32): Likewise.
11487 (vst4q_lane_s64): Likewise.
11488 (vst4q_lane_u8): Likewise.
11489 (vst4q_lane_u16): Likewise.
11490 (vst4q_lane_u32): Likewise.
11491 (vst4q_lane_u64): Likewise.
11492 (vst4q_lane_bf16): Likewise.
11494 2021-08-06 Martin Liska <mliska@suse.cz>
11496 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
11497 a target option is restored, it can have
11498 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
11499 and error should not be emitted.
11501 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
11503 * gcov-io.h (gcov_write): Declare.
11504 * gcov-io.c (gcov_write): New.
11505 (gcov_write_counter): Remove.
11506 (gcov_write_tag_length): Likewise.
11507 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
11508 gcov_write_unsigned().
11509 * doc/invoke.texi (fprofile-info-section): Mention
11510 __gcov_info_to_gdca().
11512 2021-08-06 Martin Sebor <msebor@redhat.com>
11514 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
11515 arguments to by-reference.
11516 (iterate_fix_dominators): Same.
11517 * dominance.h (iterate_fix_dominators): Same.
11518 * ipa-prop.h: Call auto_vec::to_vec_legacy.
11519 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
11520 arguments to by-reference.
11521 (debug_data_dependence_relation): Same.
11522 (dump_data_dependence_relations): Same.
11523 * tree-data-ref.h (debug_data_dependence_relation): Same.
11524 (dump_data_dependence_relations): Same.
11525 * tree-predcom.c (dump_chains): Same.
11526 (initialize_root_vars_lm): Same.
11527 (determine_unroll_factor): Same.
11528 (replace_phis_by_defined_names): Same.
11529 (insert_init_seqs): Same.
11530 (pcom_worker::tree_predictive_commoning_loop): Call
11531 auto_vec::to_vec_legacy.
11532 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
11533 arguments to by-reference.
11534 * tree-ssa-threadbackward.c (populate_worklist): Same.
11535 (back_threader::resolve_def): Same.
11536 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
11537 (vect_enhance_data_refs_alignment): Same.
11538 (vect_check_lower_bound): Same.
11539 (vect_prune_runtime_alias_test_list): Same.
11540 (vect_permute_store_chain): Same.
11541 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
11542 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
11543 * tree-vectorizer.h (vect_permute_store_chain): Same.
11544 * vec.c (test_init): New function.
11545 (vec_c_tests): Call new function.
11546 * vec.h (vec): Declare ctors, dtor, and assignment.
11547 (auto_vec::vec_to_legacy): New function.
11548 (vec::copy): Adjust initialization.
11550 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
11553 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
11554 callee only uses GPRs.
11555 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
11556 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
11557 * config/i386/x86gprintrin.h: Add
11558 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
11559 to disable non-GPR ISAs.
11561 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11563 PR middle-end/101787
11564 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
11566 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11568 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
11569 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
11570 (vect_is_extending_load, vect_is_integer_truncation): New functions,
11571 moved from aarch64.c but given different names.
11572 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
11573 (aarch64_is_reduction, aarch64_reduc_type)
11574 (aarch64_embedded_comparison_type, aarch64_comparison_type)
11575 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
11576 in favor of the above. Update callers accordingly.
11578 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11581 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
11582 writing .cpu directive in asm output.
11583 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
11584 (arm_last_printed_arch_string): Delete.
11585 (arm_last-printed_fpu_string): Delete.
11586 (arm_configure_build_target): If use of floating-point/SIMD is
11587 disabled, remove all fp/simd related features from the target ISA.
11588 (last_arm_targ_options): New variable.
11589 (arm_print_asm_arch_directives): Add new parameters. Change order
11590 of emitted directives and handle all cases here.
11591 (arm_file_start): Always call arm_print_asm_arch_directives, move
11592 all generation of .arch/.arch_extension here.
11593 (arm_file_end): Call arm_print_asm_arch.
11594 (arm_declare_function_name): Call arm_print_asm_arch_directives
11595 instead of printing .arch/.fpu directives directly.
11597 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11599 * config/arm/arm.c (arm_configure_build_target): Don't call
11600 arm_option_reconfigure_globals.
11601 (arm_option_restore): Call arm_option_reconfigure_globals after
11602 reconfiguring the target.
11603 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
11605 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11607 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
11608 arch_name is always set.
11610 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
11612 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
11613 of vec_select high-half from being added into Neon subtract
11616 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
11618 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
11619 of vec_select high-half from being added into Neon add cost.
11621 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
11623 * cfgloop.h (loops_list::loops_list): Add one optional argument
11624 root and adjust accordingly, update loop tree walking and factor
11626 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
11628 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
11630 PR tree-optimization/101626
11631 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
11632 reverse scalar storage order on a pointer or vector component.
11634 2021-08-05 liuhongt <hongtao.liu@intel.com>
11636 * config/i386/sse.md (cond_<code><mode>): New expander.
11638 2021-08-05 liuhongt <hongtao.liu@intel.com>
11640 * config/i386/sse.md (cond_<code><mode>): New expander.
11642 2021-08-05 liuhongt <hongtao.liu@intel.com>
11644 * config/i386/sse.md (cond_<code><mode>): New expander.
11646 2021-08-04 David Malcolm <dmalcolm@redhat.com>
11649 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
11651 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11654 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
11655 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
11657 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11660 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
11661 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
11662 data with SSE register from one memory location to another.
11664 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11666 * config/s390/s390.c (expand_perm_with_vpdi): New function.
11667 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
11668 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
11669 parameterized expander.
11670 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
11672 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11674 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
11675 (struct expand_vec_perm_d): Define struct.
11676 (expand_perm_with_merge): New function.
11677 (vectorize_vec_perm_const_1): New function.
11678 (s390_vectorize_vec_perm_const): New function.
11679 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
11681 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11683 * config/s390/vector.md (V_HW_64): Remove mode iterator.
11684 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
11685 * config/s390/vx-builtins.md
11686 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
11689 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11691 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
11693 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
11695 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
11696 instead of an unspec.
11698 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11700 * config/s390/s390-modes.def: Add more vector modes to support
11701 concatenation of two vectors.
11702 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
11704 (s390_expand_merge): Likewise.
11705 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
11706 (s390_expand_merge): New function.
11707 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
11708 Remove constant definitions.
11709 * config/s390/vector.md (V_HW_2): Add mode iterators.
11710 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
11711 (vec_2x_nelts, vec_2x_wide): New mode attributes.
11712 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
11713 New pattern definitions.
11714 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
11715 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
11716 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
11717 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
11719 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
11721 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
11722 emit vec merge pattern.
11724 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
11726 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
11728 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
11729 vec_select high-half from being added into Neon multiply
11731 * rtlanal.c (vec_series_highpart_p): Define.
11732 * rtlanal.h (vec_series_highpart_p): Declare.
11734 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
11736 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
11738 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
11739 vec_select cost from being added into Neon multiply cost.
11741 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11743 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
11744 which old_loop_vinfo is an epilogue loop that handles a constant
11745 number of iterations.
11747 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11749 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
11750 when a reanalyzed loop fails to be cheaper than the current
11753 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11755 * config/aarch64/aarch64.c: Fix a typo.
11757 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
11759 PR gcov-profile/101773
11760 * gcov-io.c (gcov_close): Check return code of a fclose.
11762 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
11765 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
11766 .file statement when needed.
11768 2021-08-04 Richard Biener <rguenther@suse.de>
11770 * tree-vect-data-refs.c (vect_check_gather_scatter):
11771 Include widening conversions only when the result is
11772 still handed by native gather or the current offset
11773 size not already matches the data size.
11774 Also succeed analysis in case there's no native support,
11775 noted by a IFN_LAST ifn and a NULL decl.
11776 (vect_analyze_data_refs): Always consider gathers.
11777 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
11778 Test for no IFN gather rather than decl gather.
11779 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
11780 gather-scatter info and cost emulated gathers accordingly.
11781 (vect_truncate_gather_scatter_offset): Properly test for
11783 (vect_use_strided_gather_scatters_p): Likewise.
11784 (get_load_store_type): Handle emulated gathers and its
11786 (vectorizable_load): Likewise. Emulate them by extracting
11787 scalar offsets, doing scalar loads and a vector construct.
11789 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11792 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
11793 argument to set m_max_size.
11794 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
11795 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
11796 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
11798 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
11799 Marc Glisse <marc.glisse@inria.fr>
11801 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
11802 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
11803 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
11805 2021-08-04 Richard Biener <rguenther@suse.de>
11807 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
11808 out code to decompose vector loads ...
11809 (optimize_vector_load): ... here. Generalize it to
11810 handle intermediate widening and TARGET_MEM_REF loads
11811 and apply it to loads with a supported vector mode as well.
11813 2021-08-04 Richard Biener <rguenther@suse.de>
11815 PR tree-optimization/101756
11816 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
11817 the result of the reduction epilogue is compatible to the original
11820 2021-08-04 liuhongt <hongtao.liu@intel.com>
11823 * config/i386/i386.md (peephole2): Refine predicate from
11824 register_operand to general_reg_operand.
11826 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
11828 * gimple-range-path.h (path_range_query::dump): Mark override.
11830 2021-08-04 Richard Biener <rguenther@suse.de>
11832 PR tree-optimization/101769
11833 * tree-tailcall.c (eliminate_tail_call): Add the created loop
11834 for the first recursion and return it via the new output parameter.
11835 (optimize_tail_call): Pass through new output param.
11836 (tree_optimize_tail_calls_1): After creating all latches,
11837 add the created loop to the loop tree. Do not mark loops for fixup.
11839 2021-08-04 Martin Liska <mliska@suse.cz>
11841 * doc/invoke.texi: Document threader-mode param.
11843 2021-08-04 liuhongt <hongtao.liu@intel.com>
11845 * config/i386/sse.md (cond_fma<mode>): New expander.
11846 (cond_fms<mode>): Ditto.
11847 (cond_fnma<mode>): Ditto.
11848 (cond_fnms<mode>): Ditto.
11850 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
11852 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
11854 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
11856 * config/rs6000/constraints.md: Remove "e" from the list of available
11857 constraint characters.
11859 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
11861 PR gcov-profile/71672
11862 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
11864 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
11866 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
11868 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
11870 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
11871 try XMM31 to avoid vzeroupper.
11873 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11875 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
11876 -mcpu=neoverse-512tvb.
11877 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
11878 * config/aarch64/aarch64-tune.md: Regenerate.
11879 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
11880 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
11881 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
11882 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
11883 (aarch64_adjust_body_cost): Likewise.
11885 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11887 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
11888 record issue information for operations that occur in the
11891 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11893 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
11894 parameter. Detect cases in which an Advanced SIMD MLA would almost
11895 certainly require a MOV.
11896 (aarch64_count_ops): Update accordingly.
11898 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11900 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
11901 function, split out from...
11902 (aarch64_detect_vector_stmt_subtype): ...here.
11903 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
11905 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11907 * config/aarch64/aarch64-protos.h (sve_vec_cost):
11908 Add gather_load_x32_cost and gather_load_x64_cost.
11909 * config/aarch64/aarch64.c (generic_sve_vector_cost)
11910 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
11911 accordingly, using the values given by the scalar_load * number
11912 of elements calculation that we used previously.
11913 (aarch64_detect_vector_stmt_subtype): Use the new fields.
11915 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11917 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
11918 function, split out from...
11919 (aarch64_adjust_body_cost): ...here.
11921 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11923 * config/aarch64/fractional-cost.h: New file.
11924 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
11925 and cost_fraction.h.
11926 (vec_cost_fraction): New typedef.
11927 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
11928 (aarch64_detect_vector_stmt_subtype): Likewise.
11929 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
11930 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
11932 (aarch64_adjust_body_cost): Likewise.
11933 (aarch64_test_cost_fraction): New function.
11934 (aarch64_run_selftests): Call it.
11936 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11938 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
11940 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
11942 (aarch64_estimated_poly_value): Likewise. Use the least significant
11943 set bit for the minimum and likely values. Use the most significant
11944 set bit for the maximum value.
11946 2021-08-03 liuhongt <hongtao.liu@intel.com>
11948 * config/i386/sse.md (cond_<insn><mode>): New expander.
11949 (cond_mul<mode>): Ditto.
11951 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
11953 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
11955 2021-08-03 liuhongt <hongtao.liu@intel.com>
11957 * config/i386/sse.md (cond_<insn><mode>):New expander.
11958 (cond_mul<mode>): Ditto.
11959 (cond_div<mode>): Ditto.
11961 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
11963 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
11964 check stack_realign_needed for stack realignment.
11965 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
11966 than the largest integer supported by vector register.
11967 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
11968 (MOVE_MAX): Set to bytes of the largest integer supported by
11970 (STORE_MAX_PIECES): New.
11972 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
11974 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
11975 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
11976 data from one memory location to another.
11978 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
11980 PR middle-end/90773
11981 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
11983 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
11985 PR tree-optimization/101724
11986 * params.opt: Remove --param=threader-iterative.
11987 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
11990 2021-08-02 Tom de Vries <tdevries@suse.de>
11992 PR middle-end/101665
11993 * doc/extend.texi (nonnull attribute): Improve documentation.
11995 2021-08-02 Andrew Pinski <apinski@marvell.com>
11997 PR rtl-optimization/101683
11998 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
12000 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
12002 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
12003 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
12004 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
12005 (tree_ssa_phiop_worker): Update call to function above.
12007 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
12010 * config/i386/i386.md (bsr_rex64_1_zext): New.
12011 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
12012 with gen_bsr_rex64_1_zext.
12014 2021-07-31 Jakub Jelinek <jakub@redhat.com>
12017 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
12018 define_insn patterns.
12019 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
12020 Add combine splitters for constant - clz.
12021 (clz<mode>2): Use a temporary pseudo for bsr result.
12023 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12025 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
12026 _mm_floor_sd, _mm_floor_ss): New.
12028 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12030 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
12031 _mm_ceil_sd, _mm_ceil_ss): New.
12033 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12035 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
12036 _mm_blend_ps, _mm_blendv_ps): New.
12038 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
12039 Uroš Bizjak <ubizjak@gmail.com>
12041 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
12042 to generate a conditional move using the carry flag after sub $1.
12043 (peephole2): Eliminate a register-to-register move by inverting
12044 the condition of a conditional move.
12046 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
12048 * config/mmix/mmix.md ("call", "call_value", "*call_real")
12049 ("*call_value_real"): Don't generate rtx mentioning the generic
12050 operands 1 and 2 to "call", and similarly for "call_value".
12051 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
12052 (mmix_print_operand): Use '!' instead of 'p'.
12054 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
12056 * doc/md.texi (call): Correct information about operand 2.
12057 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
12059 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12061 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
12063 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12065 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
12067 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12069 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
12070 const_basic_block..
12071 (*::get_bb_range): Ditto.
12072 (*::bb_range_p): Ditto.
12073 * gimple-range-cache.h: Change prototypes.
12075 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12077 PR middle-end/90773
12078 * builtins.c (builtin_memcpy_read_str): Change the mode argument
12079 from scalar_int_mode to fixed_size_mode.
12080 (builtin_strncpy_read_str): Likewise.
12081 (gen_memset_value_from_prev): New function.
12082 (builtin_memset_read_str): Change the mode argument from
12083 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
12084 and support CONST_VECTOR.
12085 (builtin_memset_gen_str): Likewise.
12086 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
12088 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
12089 with fixed_size_mode.
12090 (builtin_memset_read_str): Likewise.
12091 * expr.c (widest_int_mode_for_size): Renamed to ...
12092 (widest_fixed_size_mode_for_size): Add a bool argument to
12093 indicate if QI vector mode can be used.
12094 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
12095 instead of widest_int_mode_for_size.
12096 (pieces_addr::adjust): Change the mode argument from
12097 scalar_int_mode to fixed_size_mode.
12098 (op_by_pieces_d): Make m_len read-only. Add a bool member,
12099 m_qi_vector_mode, to indicate that QI vector mode can be used.
12100 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
12101 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
12102 instead of widest_int_mode_for_size.
12103 (op_by_pieces_d::get_usable_mode): Change the mode argument from
12104 scalar_int_mode to fixed_size_mode. Call
12105 widest_fixed_size_mode_for_size instead of
12106 widest_int_mode_for_size.
12107 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
12108 function to return the smallest integer or QI vector mode.
12109 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
12110 instead of widest_int_mode_for_size. Call
12111 smallest_fixed_size_mode_for_size instead of
12112 smallest_int_mode_for_size.
12113 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
12114 indicate that QI vector mode can be used and pass it to
12115 op_by_pieces_d::op_by_pieces_d.
12116 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
12117 instead of widest_int_mode_for_size. Pass memsetp to
12118 widest_fixed_size_mode_for_size to support QI vector mode.
12119 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
12120 (store_by_pieces): Pass memsetp to
12121 store_by_pieces_d::store_by_pieces_d.
12122 (clear_by_pieces_1): Removed.
12123 (clear_by_pieces): Replace clear_by_pieces_1 with
12124 builtin_memset_read_str and pass true to store_by_pieces_d to
12125 support vector mode broadcast.
12126 (string_cst_read_str): Change the mode argument from
12127 scalar_int_mode to fixed_size_mode.
12128 * expr.h (by_pieces_constfn): Change scalar_int_mode to
12130 (by_pieces_prev): Likewise.
12131 * rtl.h (lowpart_subreg_regno): New.
12132 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
12133 simplify_subreg_regno.
12134 * target.def (gen_memset_scratch_rtx): New hook.
12135 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
12136 * doc/tm.texi: Regenerated.
12138 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12141 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
12142 TARGET_EXPR instead of MODIFY_EXPR.
12144 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12147 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
12148 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
12149 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
12151 (vec_cmpu<IMSA:mode><mode_i>): New expander.
12153 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12156 * config/i386/i386-options.c (ix86_option_override_internal):
12157 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
12159 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12162 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
12163 explicitly dereferencing __a.
12164 (vld1_s64): Likewise.
12165 (vld1_u64): Likewise.
12166 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
12167 and change to VAR13.
12169 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
12171 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
12172 use of m_range_analyzer.
12173 (loop_versioning::lv_dom_walker::before_dom_children): Same.
12174 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
12175 (loop_versioning::prune_loop_conditions): Replace vr_values use
12176 with range_query interface.
12177 (pass_loop_versioning::execute): Use ranger.
12179 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12182 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
12183 enum values, and emit a warning if they mismatch.
12185 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
12187 * cfgloop.h (as_const): New function.
12188 (class loop_iterator): Rename to ...
12189 (class loops_list): ... this.
12190 (loop_iterator::next): Rename to ...
12191 (loops_list::Iter::fill_curr_loop): ... this and adjust.
12192 (loop_iterator::loop_iterator): Rename to ...
12193 (loops_list::loops_list): ... this and adjust.
12194 (loops_list::Iter): New class.
12195 (loops_list::iterator): New type.
12196 (loops_list::const_iterator): New type.
12197 (loops_list::begin): New function.
12198 (loops_list::end): Likewise.
12199 (loops_list::begin const): Likewise.
12200 (loops_list::end const): Likewise.
12201 (FOR_EACH_LOOP): Remove.
12202 (FOR_EACH_LOOP_FN): Remove.
12203 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
12204 for loop with loops_list instance.
12205 (sort_sibling_loops): Likewise.
12206 (disambiguate_loops_with_multiple_latches): Likewise.
12207 (verify_loop_structure): Likewise.
12208 * cfgloopmanip.c (create_preheaders): Likewise.
12209 (force_single_succ_latches): Likewise.
12210 * config/aarch64/falkor-tag-collision-avoidance.c
12211 (execute_tag_collision_avoidance): Likewise.
12212 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
12213 * config/s390/s390.c (s390_adjust_loops): Likewise.
12214 * doc/loop.texi: Likewise.
12215 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
12216 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
12217 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
12218 (loop_versioning::make_versioning_decisions): Likewise.
12219 * gimple-ssa-split-paths.c (split_paths): Likewise.
12220 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
12221 * graphite.c (canonicalize_loop_form): Likewise.
12222 (graphite_transform_loops): Likewise.
12223 * ipa-fnsummary.c (analyze_function_body): Likewise.
12224 * ipa-pure-const.c (analyze_function): Likewise.
12225 * loop-doloop.c (doloop_optimize_loops): Likewise.
12226 * loop-init.c (loop_optimizer_finalize): Likewise.
12227 (fix_loop_structure): Likewise.
12228 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
12229 (move_loop_invariants): Likewise.
12230 * loop-unroll.c (decide_unrolling): Likewise.
12231 (unroll_loops): Likewise.
12232 * modulo-sched.c (sms_schedule): Likewise.
12233 * predict.c (predict_loops): Likewise.
12234 (pass_profile::execute): Likewise.
12235 * profile.c (branch_prob): Likewise.
12236 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
12237 (sel_find_rgns): Likewise.
12238 * tree-cfg.c (replace_loop_annotate): Likewise.
12239 (replace_uses_by): Likewise.
12240 (move_sese_region_to_fn): Likewise.
12241 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
12242 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
12243 * tree-parloops.c (parallelize_loops): Likewise.
12244 * tree-predcom.c (tree_predictive_commoning): Likewise.
12245 * tree-scalar-evolution.c (scev_initialize): Likewise.
12246 (scev_reset): Likewise.
12247 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
12248 * tree-ssa-live.c (remove_unused_locals): Likewise.
12249 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
12250 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
12251 (tree_ssa_lim_initialize): Likewise.
12252 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
12253 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
12254 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
12255 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
12256 (free_numbers_of_iterations_estimates): Likewise.
12257 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
12258 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
12259 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
12260 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
12261 (pass_scev_cprop::execute): Likewise.
12262 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12263 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
12264 * tree-ssa-threadupdate.c
12265 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
12266 * tree-vectorizer.c (vectorize_loops): Likewise.
12267 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
12269 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
12271 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
12272 generating a VOIDmode register for e.g the
12273 function_arg_info::end_marker.
12275 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
12277 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
12278 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
12279 (cc mode_attr): Similarly.
12280 (ccz subst_attr): Similarly.
12281 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
12282 * config/h8300/testcompare.md: Remove various cc0 based patterns
12283 that had been commented out. Add pattern to set CCZ from a bit
12286 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
12287 Julian Brown <julian@codesourcery.com>
12288 Kwok Cheung Yeung <kcy@codesourcery.com>
12290 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
12291 'update_stmt' after modification.
12292 (pass_oacc_loop_designation): New function, extracted out of...
12293 (pass_oacc_device_lower): ... this.
12294 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
12295 (make_pass_oacc_loop_designation): New
12296 * passes.def: Add it.
12297 * tree-parloops.c (create_parallel_loop): Adjust.
12298 * tree-pass.h (make_pass_oacc_loop_designation): New.
12300 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
12302 * flag-types.h (enum threader_mode): New.
12303 * params.opt: Add entry for --param=threader-mode.
12304 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
12305 (class back_threader): New.
12306 (back_threader::back_threader): New.
12307 (back_threader::~back_threader): New.
12308 (back_threader::maybe_register_path): New.
12309 (back_threader::find_taken_edge): New.
12310 (back_threader::find_taken_edge_switch): New.
12311 (back_threader::find_taken_edge_cond): New.
12312 (back_threader::resolve_def): New.
12313 (back_threader::resolve_phi): New.
12314 (back_threader::find_paths_to_names): New.
12315 (back_threader::find_paths): New.
12318 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
12319 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
12320 (pass_thread_jumps::execute): Abstract out code...
12321 (try_thread_blocks): ...here.
12322 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
12323 Abstract out threading candidate code to...
12324 (single_succ_to_potentially_threadable_block): ...here.
12325 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
12327 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
12328 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
12329 Return bool from register_jump_thread.
12331 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
12333 * target.def: in0 and in1 do not need to be registers.
12334 * doc/tm.texi: Regenerate.
12336 2021-07-29 liuhongt <hongtao.liu@intel.com>
12339 * config/i386/i386.c (ix86_widen_mult_cost): New function.
12340 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
12343 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
12346 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
12347 (rs6000_preferred_doloop_mode): New hook.
12348 * doc/tm.texi: Regenerate.
12349 * doc/tm.texi.in: Add hook preferred_doloop_mode.
12350 * target.def (preferred_doloop_mode): New hook.
12351 * targhooks.c (default_preferred_doloop_mode): New hook.
12352 * targhooks.h (default_preferred_doloop_mode): New hook.
12353 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
12354 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
12355 and compute_doloop_base_on_mode.
12357 2021-07-28 Martin Sebor <msebor@redhat.com>
12359 PR middle-end/101494
12360 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
12361 and size computation.
12363 2021-07-28 Martin Sebor <msebor@redhat.com>
12365 PR middle-end/101601
12366 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
12368 Handle pointers to functions.
12370 2021-07-28 Martin Sebor <msebor@redhat.com>
12372 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
12373 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
12374 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
12375 (access_ref::access_ref): Same.
12376 (access_ref::phi): Same.
12377 (access_ref::get_ref): Same.
12378 (access_ref::size_remaining): Same.
12379 (access_ref::offset_in_range): Same.
12380 (access_ref::add_offset): Same.
12381 (access_ref::inform_access): Same.
12382 (ssa_name_limit_t::visit_phi): Same.
12383 (ssa_name_limit_t::leave_phi): Same.
12384 (ssa_name_limit_t::next): Same.
12385 (ssa_name_limit_t::next_phi): Same.
12386 (ssa_name_limit_t::~ssa_name_limit_t): Same.
12387 (pointer_query::pointer_query): Same.
12388 (pointer_query::get_ref): Same.
12389 (pointer_query::put_ref): Same.
12390 (pointer_query::flush_cache): Same.
12391 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
12392 (check_nul_terminated_array): Same.
12393 (unterminated_array): Same.
12394 (maybe_warn_for_bound): Same.
12395 (check_read_access): Same.
12396 (warn_for_access): Same.
12397 (get_size_range): Same.
12398 (check_access): Same.
12399 (gimple_call_alloc_size): Move to tree.c.
12400 (gimple_parm_array_size): Move to pointer-query.cc.
12401 (get_offset_range): Same.
12402 (gimple_call_return_array): Same.
12403 (handle_min_max_size): Same.
12404 (handle_array_ref): Same.
12405 (handle_mem_ref): Same.
12406 (compute_objsize): Same.
12407 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
12408 (call_dealloc_argno): Same.
12409 (fndecl_dealloc_argno): Same.
12410 (new_delete_mismatch_p): Same.
12411 (matching_alloc_calls_p): Same.
12412 (warn_dealloc_offset): Same.
12413 (maybe_emit_free_warning): Same.
12414 * builtins.h (check_nul_terminated_array): Move to
12415 gimple-ssa-warn-access.h.
12416 (check_nul_terminated_array): Same.
12417 (warn_string_no_nul): Same.
12418 (unterminated_array): Same.
12419 (class ssa_name_limit_t): Same.
12420 (class pointer_query): Same.
12421 (struct access_ref): Same.
12422 (class range_query): Same.
12423 (struct access_data): Same.
12424 (gimple_call_alloc_size): Same.
12425 (gimple_parm_array_size): Same.
12426 (compute_objsize): Same.
12427 (class access_data): Same.
12428 (maybe_emit_free_warning): Same.
12429 * calls.c (initialize_argument_information): Remove call to
12430 maybe_emit_free_warning.
12431 * gimple-array-bounds.cc: Include new header..
12432 * gimple-fold.c: Same.
12433 * gimple-ssa-sprintf.c: Same.
12434 * gimple-ssa-warn-restrict.c: Same.
12435 * passes.def: Add pass_warn_access.
12436 * tree-pass.h (make_pass_warn_access): Declare.
12437 * tree-ssa-strlen.c: Include new headers.
12438 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
12439 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
12440 * gimple-ssa-warn-access.cc: New file.
12441 * gimple-ssa-warn-access.h: New file.
12442 * pointer-query.cc: New file.
12443 * pointer-query.h: New file.
12445 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12447 PR middle-end/101624
12448 * ubsan.c (maybe_instrument_pointer_overflow,
12449 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
12450 PARM_DECLs or RESULT_DECLs.
12451 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
12453 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12455 PR middle-end/101642
12456 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
12457 to type of bswap16 for comparison.
12458 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
12460 2021-07-28 Richard Biener <rguenther@suse.de>
12462 PR tree-optimization/101615
12463 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
12464 at CTOR SLP graph entries.
12466 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12468 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
12469 Add "r,w" alternative.
12471 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
12474 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
12475 AVX_U128_DIRTY when all bits are zero.
12477 2021-07-28 Richard Biener <rguenther@suse.de>
12479 PR tree-optimization/101615
12480 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
12481 external nodes cannot be permuted so make them perm_out 0.
12483 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
12486 * config.in: Regenerate.
12487 * config/gcn/gcn-hsa.h (A_FIJI): New define.
12488 (A_900): New define.
12489 (A_906): New define.
12490 (A_908): New define.
12491 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
12492 * config/gcn/gcn.c (output_file_start): Adjust attributes according
12493 to the assembler capabilities.
12494 * config/gcn/mkoffload.c (main): Likewise.
12495 * configure: Regenerate.
12496 * configure.ac: Add tests for LLVM assembler attribute features.
12498 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
12500 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
12501 cond_false and cond_true on branches.
12503 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12505 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
12508 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12510 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
12512 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12515 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
12516 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
12517 and has special !TARGET_AVX512VL expansion.
12518 (vashrv2di3<mask_name>): Rename to ...
12519 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
12520 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
12522 2021-07-28 Martin Uecker <muecker@gwdg.de>
12524 * calls.c (maybe_warn_rdwr_sizes): Correct argument
12525 numbers in warning that were switched.
12527 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
12529 PR tree-optimization/101596
12530 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
12531 by using new_type's precision instead.
12533 2021-07-28 liuhongt <hongtao.liu@intel.com>
12536 * config/i386/i386.h (processor_costs): Add new member
12538 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
12539 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
12540 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
12541 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
12542 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
12543 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
12544 generic_cost, core_cost): Initialize integer_to_sse same value
12546 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
12547 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12548 Use integer_to_sse instead of sse_op to calculate the cost of
12551 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12553 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
12555 (write_init_file): Call write_ovld_static_init.
12557 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12559 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
12561 (write_init_file): Call write_bif_static_init.
12563 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12565 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
12566 (TYPE_MAP_SIZE): New macro.
12567 (type_map): New initialized variable.
12568 (typemap_cmp): New function.
12569 (write_type_node): Likewise.
12570 (write_fntype_init): Implement.
12572 2021-07-27 Martin Sebor <msebor@redhat.com>
12574 PR tree-optimization/101584
12575 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
12576 (check_defs): Call it.
12578 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12580 * tree-ssa-dom.c (dom_jump_threader_simplifier):
12581 Put avail_exprs_stack in the class, instead of passing it to
12582 jump_threader_simplifier.
12583 (dom_jump_threader_simplifier::simplify): Add state argument.
12584 (dom_opt_dom_walker): Add state.
12585 (pass_dominator::execute): Pass state to threader.
12586 (dom_opt_dom_walker::before_dom_children): Use state.
12587 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
12588 arguments by state.
12589 (jump_threader::record_temporary_equivalences_from_phis):
12590 Register equivalences through the state variable.
12591 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
12592 Record ranges in a statement through the state variable.
12593 (jump_threader::simplify_control_stmt_condition): Pass state to
12595 (jump_threader::simplify_control_stmt_condition_1): Same.
12596 (jump_threader::thread_around_empty_blocks): Remove obsolete
12598 (jump_threader::thread_through_normal_block): Record equivalences
12599 on edge through the state variable.
12600 (jump_threader::thread_across_edge): Abstract state pushing.
12601 (jt_state::jt_state): New.
12602 (jt_state::push): New.
12603 (jt_state::pop): New.
12604 (jt_state::register_equiv): New.
12605 (jt_state::record_ranges_from_stmt): New.
12606 (jt_state::register_equivs_on_edge): New.
12607 (jump_threader_simplifier::jump_threader_simplifier): Move from
12609 (jump_threader_simplifier::simplify): Add state argument.
12610 * tree-ssa-threadedge.h (class jt_state): New.
12611 (class jump_threader): Add state to constructor.
12612 (class jump_threader_simplifier): Add state to simplify. Remove
12613 avail_exprs_stack from class.
12614 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
12616 (vrp_jump_threader::vrp_jump_threader): Add state.
12617 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
12619 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12621 * Makefile.in (OBJS): Add gimple-range-path.o.
12622 * gimple-range-path.cc: New file.
12623 * gimple-range-path.h: New file.
12625 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
12627 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
12628 inside vec_duplicate for all patterns.
12629 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
12630 Push sign/zero-extension inside vec_duplicate.
12632 2021-07-27 Richard Biener <rguenther@suse.de>
12634 PR tree-optimization/101573
12635 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
12636 looking at uninitialized PHI arg defs in some constrained cases.
12637 (warn_uninitialized_vars): Call it.
12638 (execute_early_warn_uninitialized): Calculate dominators.
12640 2021-07-27 Richard Biener <rguenther@suse.de>
12642 PR tree-optimization/39821
12643 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
12644 vector_stmt for widening arithmetic.
12645 (vectorizable_conversion): Adjust.
12647 2021-07-27 Martin Jambor <mjambor@suse.cz>
12649 * cgraph.h (ipa_replace_map): New field force_load_ref.
12650 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
12651 aded new flag load_dereferenced, adjusted comments.
12652 (ipa_get_param_dereferenced): New function.
12653 (ipa_set_param_dereferenced): Likewise.
12654 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
12655 * ipa-cp.c: Include gimple.h.
12656 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
12657 (get_replacement_map): New parameter force_load_ref, set the
12658 appropriate flag in ipa_replace_map if set.
12659 (struct symbol_and_index_together): New type.
12660 (adjust_refs_in_act_callers): New function.
12661 (adjust_references_in_caller): Likewise.
12662 (create_specialized_node): When appropriate, call
12663 adjust_references_in_caller and force only load references.
12664 * ipa-prop.c (load_from_dereferenced_name): New function.
12665 (ipa_analyze_controlled_uses): Also detect loads from a
12666 dereference, harden testing of call statements.
12667 (ipa_write_node_info): Stream the dereferenced flag.
12668 (ipa_read_node_info): Likewise.
12669 (ipa_set_jf_constant): Also create refdesc when jump function
12670 references a variable.
12671 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
12672 also on references of variables and return a symtab_node. Adjust
12674 (propagate_controlled_uses): Also remove references to VAR_DECLs.
12676 2021-07-27 Jakub Jelinek <jakub@redhat.com>
12678 PR middle-end/101586
12679 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
12680 positions above or equal to sz except for diagnostics of flexible
12683 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
12685 PR tree-optimization/78888
12686 * gimple-range-fold.cc (get_letter_range): New.
12687 (fold_using_range::range_of_builtin_call): Call get_letter_range.
12689 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
12691 PR tree-optimization/78888
12692 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
12693 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
12695 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
12696 Marc Glisse <marc.glisse@inria.fr>
12698 * match.pd (rotate): Simplify equality/inequality of rotations.
12699 (bswap): Simplify equality/inequality tests of byte swapping.
12701 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12703 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
12706 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12708 * range-op.cc (operator_lshift::fold_range): Pass rel to
12709 base class fold_range.
12710 (operator_rshift::fold_range): Same.
12712 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
12715 * toplev.h (min_align_loops_log): Remove declaration.
12716 (min_align_jumps_log, min_align_labels_log): Likewise.
12717 (min_align_functions_log): Likewise.
12719 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12721 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
12723 (execute_vrp): Abstract out simplification of conditionals...
12724 (simplify_casted_conds): ...here.
12726 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12728 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
12729 Add gimple argument.
12730 (array_bounds_checker::check_array_ref): Same.
12731 (array_bounds_checker::check_addr_expr): Same.
12732 (array_bounds_checker::check_array_bounds): Pass statement to
12733 check_array_bounds and check_addr_expr.
12734 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
12735 (check_addr_expr): Same.
12736 (get_value_range): Same.
12738 2021-07-26 Tamar Christina <tamar.christina@arm.com>
12740 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
12741 (sdot_prod, udot_prod): ... This.
12742 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
12744 (<sur>dot_prod<vsi2qi>): ... this.
12745 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
12746 Change operands order.
12747 (<sur>sadv16qi): Use new operands order.
12748 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
12749 vdotq_s32): Use new RTL ordering.
12751 2021-07-26 Tamar Christina <tamar.christina@arm.com>
12753 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
12754 aarch64_types_ternop_suss_qualifiers): New.
12755 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
12756 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
12757 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
12759 2021-07-23 Jakub Jelinek <jakub@redhat.com>
12761 PR rtl-optimization/101562
12762 * expmed.c (store_integral_bit_field): Only use movstrict_optab
12763 if the operand isn't paradoxical.
12765 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
12767 * gimple-array-bounds.h (class array_bounds_checker): Change
12768 ranges type to range_query.
12770 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12772 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
12773 __builtin_memcpy instead of constructing
12774 __builtin_aarch64_simd_oi one vector at a time.
12775 (vst1_u64_x2): Likewise.
12776 (vst1_f64_x2): Likewise.
12777 (vst1_s8_x2): Likewise.
12778 (vst1_p8_x2): Likewise.
12779 (vst1_s16_x2): Likewise.
12780 (vst1_p16_x2): Likewise.
12781 (vst1_s32_x2): Likewise.
12782 (vst1_u8_x2): Likewise.
12783 (vst1_u16_x2): Likewise.
12784 (vst1_u32_x2): Likewise.
12785 (vst1_f16_x2): Likewise.
12786 (vst1_f32_x2): Likewise.
12787 (vst1_p64_x2): Likewise.
12788 (vst1q_s8_x2): Likewise.
12789 (vst1q_p8_x2): Likewise.
12790 (vst1q_s16_x2): Likewise.
12791 (vst1q_p16_x2): Likewise.
12792 (vst1q_s32_x2): Likewise.
12793 (vst1q_s64_x2): Likewise.
12794 (vst1q_u8_x2): Likewise.
12795 (vst1q_u16_x2): Likewise.
12796 (vst1q_u32_x2): Likewise.
12797 (vst1q_u64_x2): Likewise.
12798 (vst1q_f16_x2): Likewise.
12799 (vst1q_f32_x2): Likewise.
12800 (vst1q_f64_x2): Likewise.
12801 (vst1q_p64_x2): Likewise.
12803 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12805 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
12806 __builtin_memcpy instead of constructing
12807 __builtin_aarch64_simd_ci one vector at a time.
12808 (vst1_u64_x3): Likewise.
12809 (vst1_f64_x3): Likewise.
12810 (vst1_s8_x3): Likewise.
12811 (vst1_p8_x3): Likewise.
12812 (vst1_s16_x3): Likewise.
12813 (vst1_p16_x3): Likewise.
12814 (vst1_s32_x3): Likewise.
12815 (vst1_u8_x3): Likewise.
12816 (vst1_u16_x3): Likewise.
12817 (vst1_u32_x3): Likewise.
12818 (vst1_f16_x3): Likewise.
12819 (vst1_f32_x3): Likewise.
12820 (vst1_p64_x3): Likewise.
12821 (vst1q_s8_x3): Likewise.
12822 (vst1q_p8_x3): Likewise.
12823 (vst1q_s16_x3): Likewise.
12824 (vst1q_p16_x3): Likewise.
12825 (vst1q_s32_x3): Likewise.
12826 (vst1q_s64_x3): Likewise.
12827 (vst1q_u8_x3): Likewise.
12828 (vst1q_u16_x3): Likewise.
12829 (vst1q_u32_x3): Likewise.
12830 (vst1q_u64_x3): Likewise.
12831 (vst1q_f16_x3): Likewise.
12832 (vst1q_f32_x3): Likewise.
12833 (vst1q_f64_x3): Likewise.
12834 (vst1q_p64_x3): Likewise.
12836 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
12839 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
12840 hard register when LRA is in progress.
12842 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12844 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
12845 __builtin_memcpy instead of using a union.
12846 (vst1q_s8_x4): Likewise.
12847 (vst1_s16_x4): Likewise.
12848 (vst1q_s16_x4): Likewise.
12849 (vst1_s32_x4): Likewise.
12850 (vst1q_s32_x4): Likewise.
12851 (vst1_u8_x4): Likewise.
12852 (vst1q_u8_x4): Likewise.
12853 (vst1_u16_x4): Likewise.
12854 (vst1q_u16_x4): Likewise.
12855 (vst1_u32_x4): Likewise.
12856 (vst1q_u32_x4): Likewise.
12857 (vst1_f16_x4): Likewise.
12858 (vst1q_f16_x4): Likewise.
12859 (vst1_f32_x4): Likewise.
12860 (vst1q_f32_x4): Likewise.
12861 (vst1_p8_x4): Likewise.
12862 (vst1q_p8_x4): Likewise.
12863 (vst1_p16_x4): Likewise.
12864 (vst1q_p16_x4): Likewise.
12865 (vst1_s64_x4): Likewise.
12866 (vst1_u64_x4): Likewise.
12867 (vst1_p64_x4): Likewise.
12868 (vst1q_s64_x4): Likewise.
12869 (vst1q_u64_x4): Likewise.
12870 (vst1q_p64_x4): Likewise.
12871 (vst1_f64_x4): Likewise.
12872 (vst1q_f64_x4): Likewise.
12874 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
12876 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
12877 instead of constructing __builtin_aarch64_simd_oi one vector
12879 (vst2_u64): Likewise.
12880 (vst2_f64): Likewise.
12881 (vst2_s8): Likewise.
12882 (vst2_p8): Likewise.
12883 (vst2_s16): Likewise.
12884 (vst2_p16): Likewise.
12885 (vst2_s32): Likewise.
12886 (vst2_u8): Likewise.
12887 (vst2_u16): Likewise.
12888 (vst2_u32): Likewise.
12889 (vst2_f16): Likewise.
12890 (vst2_f32): Likewise.
12891 (vst2_p64): Likewise.
12892 (vst2q_s8): Likewise.
12893 (vst2q_p8): Likewise.
12894 (vst2q_s16): Likewise.
12895 (vst2q_p16): Likewise.
12896 (vst2q_s32): Likewise.
12897 (vst2q_s64): Likewise.
12898 (vst2q_u8): Likewise.
12899 (vst2q_u16): Likewise.
12900 (vst2q_u32): Likewise.
12901 (vst2q_u64): Likewise.
12902 (vst2q_f16): Likewise.
12903 (vst2q_f32): Likewise.
12904 (vst2q_f64): Likewise.
12905 (vst2q_p64): Likewise.
12907 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12909 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
12910 instead of constructing __builtin_aarch64_simd_ci one vector
12912 (vst3_u64): Likewise.
12913 (vst3_f64): Likewise.
12914 (vst3_s8): Likewise.
12915 (vst3_p8): Likewise.
12916 (vst3_s16): Likewise.
12917 (vst3_p16): Likewise.
12918 (vst3_s32): Likewise.
12919 (vst3_u8): Likewise.
12920 (vst3_u16): Likewise.
12921 (vst3_u32): Likewise.
12922 (vst3_f16): Likewise.
12923 (vst3_f32): Likewise.
12924 (vst3_p64): Likewise.
12925 (vst3q_s8): Likewise.
12926 (vst3q_p8): Likewise.
12927 (vst3q_s16): Likewise.
12928 (vst3q_p16): Likewise.
12929 (vst3q_s32): Likewise.
12930 (vst3q_s64): Likewise.
12931 (vst3q_u8): Likewise.
12932 (vst3q_u16): Likewise.
12933 (vst3q_u32): Likewise.
12934 (vst3q_u64): Likewise.
12935 (vst3q_f16): Likewise.
12936 (vst3q_f32): Likewise.
12937 (vst3q_f64): Likewise.
12938 (vst3q_p64): Likewise.
12940 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12942 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
12943 instead of constructing __builtin_aarch64_simd_xi one vector
12945 (vst4_u64): Likewise.
12946 (vst4_f64): Likewise.
12947 (vst4_s8): Likewise.
12948 (vst4_p8): Likewise.
12949 (vst4_s16): Likewise.
12950 (vst4_p16): Likewise.
12951 (vst4_s32): Likewise.
12952 (vst4_u8): Likewise.
12953 (vst4_u16): Likewise.
12954 (vst4_u32): Likewise.
12955 (vst4_f16): Likewise.
12956 (vst4_f32): Likewise.
12957 (vst4_p64): Likewise.
12958 (vst4q_s8): Likewise.
12959 (vst4q_p8): Likewise.
12960 (vst4q_s16): Likewise.
12961 (vst4q_p16): Likewise.
12962 (vst4q_s32): Likewise.
12963 (vst4q_s64): Likewise.
12964 (vst4q_u8): Likewise.
12965 (vst4q_u16): Likewise.
12966 (vst4q_u32): Likewise.
12967 (vst4q_u64): Likewise.
12968 (vst4q_f16): Likewise.
12969 (vst4q_f32): Likewise.
12970 (vst4q_f64): Likewise.
12971 (vst4q_p64): Likewise.
12973 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12975 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
12976 instead of constructing __builtin_aarch64_simd_oi one vector
12978 (vtbx4_u8): Likewise.
12979 (vtbx4_p8): Likewise.
12981 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12983 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
12984 instead of constructing __builtin_aarch64_simd_oi one vector
12986 (vtbl3_u8): Likewise.
12987 (vtbl3_p8): Likewise.
12988 (vtbl4_s8): Likewise.
12989 (vtbl4_u8): Likewise.
12990 (vtbl4_p8): Likewise.
12992 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12994 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
12995 instead of constructing __builtin_aarch64_simd_oi one vector
12997 (vqtbx2_u8): Likewise.
12998 (vqtbx2_p8): Likewise.
12999 (vqtbx2q_s8): Likewise.
13000 (vqtbx2q_u8): Likewise.
13001 (vqtbx2q_p8): Likewise.
13002 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
13003 __builtin_aarch64_simd_ci one vector at a time.
13004 (vqtbx3_u8): Likewise.
13005 (vqtbx3_p8): Likewise.
13006 (vqtbx3q_s8): Likewise.
13007 (vqtbx3q_u8): Likewise.
13008 (vqtbx3q_p8): Likewise.
13009 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
13010 __builtin_aarch64_simd_xi one vector at a time.
13011 (vqtbx4_u8): Likewise.
13012 (vqtbx4_p8): Likewise.
13013 (vqtbx4q_s8): Likewise.
13014 (vqtbx4q_u8): Likewise.
13015 (vqtbx4q_p8): Likewise.
13017 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13019 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
13020 instead of constructing __builtin_aarch64_simd_oi one vector
13022 (vqtbl2_u8): Likewise.
13023 (vqtbl2_p8): Likewise.
13024 (vqtbl2q_s8): Likewise.
13025 (vqtbl2q_u8): Likewise.
13026 (vqtbl2q_p8): Likewise.
13027 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
13028 __builtin_aarch64_simd_ci one vector at a time.
13029 (vqtbl3_u8): Likewise.
13030 (vqtbl3_p8): Likewise.
13031 (vqtbl3q_s8): Likewise.
13032 (vqtbl3q_u8): Likewise.
13033 (vqtbl3q_p8): Likewise.
13034 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
13035 __builtin_aarch64_simd_xi one vector at a time.
13036 (vqtbl4_u8): Likewise.
13037 (vqtbl4_p8): Likewise.
13038 (vqtbl4q_s8): Likewise.
13039 (vqtbl4q_u8): Likewise.
13040 (vqtbl4q_p8): Likewise.
13042 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
13045 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
13047 2021-07-22 Andrew Pinski <apinski@marvell.com>
13049 PR tree-optimization/10153
13050 * tree-tailcall.c (create_tailcall_accumulator):
13051 Don't call fold_convert as the type should be correct already.
13052 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
13053 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
13055 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
13057 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
13058 varying_p check for null/non-null check.
13060 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13062 PR tree-optimization/101511
13063 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
13064 is in ssa2's equiv set, and don't trap if so.
13066 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13068 PR tree-optimization/101497
13069 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
13072 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13074 PR tree-optimization/101496
13075 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
13076 first, then vrp_visit_cond_Stmt.
13078 2021-07-22 liuhongt <hongtao.liu@intel.com>
13080 * config/i386/i386-expand.c
13081 (ix86_broadcast_from_integer_constant): Rename to ..
13082 (ix86_broadcast_from_constant): .. this, and extend it to
13084 (ix86_expand_vector_move): Extend to float mode.
13085 * config/i386/i386-features.c
13086 (replace_constant_pool_with_broadcast): Remove.
13087 (remove_partial_avx_dependency_gate): Ditto.
13088 (constant_pool_broadcast): Ditto.
13089 (class pass_constant_pool_broadcast): Ditto.
13090 (make_pass_constant_pool_broadcast): Ditto.
13091 (remove_partial_avx_dependency): Adjust gate.
13092 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
13093 * config/i386/i386-protos.h
13094 (make_pass_constant_pool_broadcast): Remove.
13096 2021-07-22 liuhongt <hongtao.liu@intel.com>
13098 * config/i386/constraints.md (Wb): New constraint.
13100 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
13102 (*ashlqi3_1): Ditto.
13103 (*<insn><mode>3_1): Split to ..
13104 (*ashr<mode>3_1): this, ...
13105 (*lshr<mode>3_1): and this, also extend this pattern to avx512
13107 (*<insn><mode>3_1): Split to ..
13108 (*ashr<mode>3_1): this, ...
13109 (*lshrqi3_1): and this, also extend this pattern to avx512
13111 (*lshrhi3_1): And this, also extend this pattern to avx512
13113 * config/i386/sse.md (k<code><mode>): New define_split after
13114 it to convert generic shift pattern to mask shift ones.
13116 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
13117 Joseph Myers <joseph@codesourcery.com>
13118 Cesar Philippidis <cesar@codesourcery.com>
13120 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
13121 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
13123 * tree-pretty-print.c (dump_omp_clause): Likewise.
13124 * omp-general.c (oacc_verify_routine_clauses): Likewise.
13125 * gimplify.c (gimplify_scan_omp_clauses)
13126 (gimplify_adjust_omp_clauses): Likewise.
13127 * tree-nested.c (convert_nonlocal_omp_clauses)
13128 (convert_local_omp_clauses): Likewise.
13129 * omp-low.c (scan_sharing_clauses): Likewise.
13130 * omp-offload.c (execute_oacc_device_lower): Update.
13132 2021-07-21 Martin Sebor <msebor@redhat.com>
13134 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
13136 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13138 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
13141 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13143 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
13145 (write_fntype_init): New stub function.
13146 (write_init_bif_table): Likewise.
13147 (write_init_ovld_table): New function.
13148 (write_init_file): Implement.
13150 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13152 * config/rs6000/rs6000-gen-builtins.c
13153 (write_autogenerated_header): New function.
13154 (write_decls): Likewise.
13155 (write_extern_fntype): New callback function.
13156 (write_header_file): Implement.
13158 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13160 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
13163 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13165 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
13167 (complete_base_type): Likewise.
13168 (construct_fntype_id): Likewise.
13169 (parse_bif_entry): Call contruct_fntype_id.
13170 (parse_ovld_entry): Likewise.
13172 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13174 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
13175 (MAXOVLDSTANZAS): New macro.
13176 (ovld_stanzas): New variable.
13177 (curr_ovld_stanza): Likewise.
13178 (MAXOVLDS): New macro.
13179 (ovlddata): New struct.
13180 (ovlds): New variable.
13181 (curr_ovld): Likewise.
13182 (max_ovld_args): Likewise.
13183 (parse_ovld_entry): New function.
13184 (parse_ovld_stanza): Likewise.
13185 (parse_ovld): Implement.
13187 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13189 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
13192 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13194 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
13195 (parse_prototype): Implement.
13197 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13199 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
13200 (curr_bif_stanza): New variable.
13201 (stanza_entry): New struct.
13202 (stanza_map): New initialized variable.
13203 (enable_string): Likewise.
13204 (fnkinds): New enum.
13205 (typelist): New struct.
13206 (attrinfo): Likewise.
13207 (MAXRESTROPNDS): New macro.
13208 (prototype): New struct.
13209 (MAXBIFS): New macro.
13210 (bifdata): New struct.
13211 (bifs): New variable.
13212 (curr_bif): Likewise.
13213 (bif_order): Likewise.
13214 (bif_index): Likewise.
13215 (fatal): New function.
13216 (stanza_name_to_stanza): Likewise.
13217 (parse_bif_attrs): New stub function.
13218 (parse_prototype): Likewise.
13219 (parse_bif_entry): New function.
13220 (parse_bif_stanza): Likewise.
13221 (parse_bif): Implement.
13222 (set_bif_order): New function.
13223 (create_bif_order): Implement.
13225 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13227 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
13228 (num_bifs): New variable.
13229 (num_ovld_stanzas): Likewise.
13230 (num_ovlds): Likewise.
13231 (parse_codes): New enum.
13232 (bif_rbt): New variable.
13233 (ovld_rbt): Likewise.
13234 (fntype_rbt): Likewise.
13235 (bifo_rbt): Likewise.
13236 (parse_bif): New stub function.
13237 (create_bif_order): Likewise.
13238 (parse_ovld): Likewise.
13239 (write_header_file): Likewise.
13240 (write_init_file): Likewise.
13241 (write_defines_file): Likewise.
13242 (delete_output_files): New function.
13245 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
13248 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
13249 from CRC32 _builtin functions.
13251 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
13253 * coverage.c (build_gcov_info_var_registration): Mark the object placed
13254 in the linker set as referenced so that it does not get optimized away.
13256 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
13259 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13261 * config.gcc (riscv*-*-*): Detect which python is available.
13263 2021-07-21 Jakub Jelinek <jakub@redhat.com>
13265 PR middle-end/101535
13266 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
13267 contexts in which decl isn't privatized and for ORT_TARGET return
13268 false if decl is mapped.
13270 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13272 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
13273 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
13274 has been successfully unrolled.
13276 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13278 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
13279 exit condition if no tail loop is needed, and if the original exit
13280 condition should therefore be kept as-is.
13281 (tree_transform_and_unroll_loop): Handle that case here too.
13283 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
13285 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
13287 (free_data_refs): Likewise.
13288 * tree-data-ref.h (free_dependence_relations): Likewise.
13289 (free_data_refs): Likewise.
13290 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
13292 (struct component): Likewise.
13293 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
13294 (pcom_worker::~pcom_worker): Likewise.
13295 (pcom_worker::release_chain): Adjust as auto_vec changes.
13296 (pcom_worker::loop): Rename to ...
13297 (pcom_worker::m_loop): ... this.
13298 (pcom_worker::datarefs): Rename to ...
13299 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
13300 (pcom_worker::dependences): Rename to ...
13301 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
13302 (pcom_worker::chains): Rename to ...
13303 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
13304 (pcom_worker::looparound_phis): Rename to ...
13305 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
13307 (pcom_worker::cache): Rename to ...
13308 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
13309 (pcom_worker::release_chain): Adjust for auto_vec changes.
13310 (pcom_worker::release_chains): Adjust for auto_vec and renaming
13312 (release_component): Remove.
13313 (release_components): Adjust for release_component removal.
13314 (component_of): Adjust to use vec.
13315 (merge_comps): Likewise.
13316 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
13317 (pcom_worker::determine_offset): Likewise.
13318 (class comp_ptrs): Remove.
13319 (pcom_worker::split_data_refs_to_components): Adjust for renaming
13320 changes, for comp_ptrs removal with auto_vec.
13321 (pcom_worker::suitable_component_p): Adjust for renaming changes.
13322 (pcom_worker::filter_suitable_components): Adjust for release_component
13324 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
13325 (pcom_worker::find_looparound_phi): Likewise.
13326 (pcom_worker::add_looparound_copies): Likewise.
13327 (pcom_worker::determine_roots_comp): Likewise.
13328 (pcom_worker::single_nonlooparound_use): Likewise.
13329 (pcom_worker::execute_pred_commoning_chain): Likewise.
13330 (pcom_worker::execute_pred_commoning): Likewise.
13331 (pcom_worker::try_combine_chains): Likewise.
13332 (pcom_worker::prepare_initializers_chain): Likewise.
13333 (pcom_worker::prepare_initializers): Likewise.
13334 (pcom_worker::prepare_finalizers_chain): Likewise.
13335 (pcom_worker::prepare_finalizers): Likewise.
13336 (pcom_worker::tree_predictive_commoning_loop): Likewise.
13338 2021-07-20 Martin Sebor <msebor@redhat.com>
13340 PR middle-end/101397
13341 * builtins.c (gimple_call_return_array): Add argument. Correct
13342 offsets for memchr, mempcpy, stpcpy, and stpncpy.
13343 (compute_objsize_r): Adjust offset computation for argument returning
13346 2021-07-20 Martin Sebor <msebor@redhat.com>
13348 PR middle-end/101300
13349 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
13351 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
13353 * function.c (assign_parm_setup_block): Use adjust_address instead
13354 of change_address to preserve MEM_EXPR and friends.
13356 2021-07-20 Martin Sebor <msebor@redhat.com>
13358 * cfgloop.h (single_likely_exit): Adjust by-value argument to
13359 by-const-reference.
13360 * cfgloopanal.c (single_likely_exit): Same.
13361 * cgraph.h (struct cgraph_node): Same.
13362 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
13363 * genautomata.c (merge_states): Same.
13364 * genextract.c (VEC_char_to_string): Same.
13365 * genmatch.c (dt_node::gen_kids_1): Same.
13366 (walk_captures): Adjust by-value argument to by-reference.
13367 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
13368 to by-const-reference.
13369 * gimple.c (gimple_build_call_vec): Same.
13370 (gimple_build_call_internal_vec): Same.
13371 (gimple_build_switch): Same.
13372 (sort_case_labels): Same.
13373 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13375 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
13376 by-const-reference.
13377 (gimple_build_call_internal_vec): Same.
13378 (gimple_build_switch): Same.
13379 (sort_case_labels): Same.
13380 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13382 * haifa-sched.c (calc_priorities): Adjust by-value argument to
13383 by-const-reference.
13384 (sched_init_luids): Same.
13385 (haifa_init_h_i_d): Same.
13386 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
13387 (adjust_callers_for_value_intersection): Adjust by-value argument to
13389 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
13390 by-const-reference.
13391 (find_more_contexts_for_caller_subset): Same.
13392 (find_aggregate_values_for_callers_subset): Same.
13393 (copy_useful_known_contexts): Same.
13394 * ipa-fnsummary.c (remap_edge_summaries): Same.
13395 (remap_freqcounting_predicate): Same.
13396 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
13398 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
13399 to by-const-reference.
13400 * ipa-predicate.h (predicate::remap_after_inlining): Same.
13401 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
13402 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
13403 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
13404 * read-rtl.c (add_overload_instance): Same.
13405 * rtl.h (native_decode_rtx): Same.
13406 (native_decode_vector_rtx): Same.
13407 * sched-int.h (sched_init_luids): Same.
13408 (haifa_init_h_i_d): Same.
13409 * simplify-rtx.c (native_decode_vector_rtx): Same.
13410 (native_decode_rtx): Same.
13411 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
13412 (shrink_wrap_one_built_in_call_with_conds): Same.
13413 (shrink_wrap_conditional_dead_built_in_calls): Same.
13414 * tree-data-ref.c (create_runtime_alias_checks): Same.
13415 (compute_all_dependences): Same.
13416 * tree-data-ref.h (compute_all_dependences): Same.
13417 (create_runtime_alias_checks): Same.
13418 (index_in_loop_nest): Same.
13419 * tree-if-conv.c (mask_exists): Same.
13420 * tree-loop-distribution.c (class loop_distribution): Same.
13421 (loop_distribution::create_rdg_vertices): Same.
13422 (dump_rdg_partitions): Same.
13423 (debug_rdg_partitions): Same.
13424 (partition_contains_all_rw): Same.
13425 (loop_distribution::distribute_loop): Same.
13426 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
13427 (oacc_entry_exit_single_gang): Same.
13428 * tree-ssa-loop-im.c (hoist_memory_references): Same.
13429 (loop_suitable_for_sm): Same.
13430 * tree-ssa-loop-niter.c (bound_index): Same.
13431 * tree-ssa-reassoc.c (update_ops): Same.
13432 (swap_ops_for_binary_stmt): Same.
13433 (rewrite_expr_tree): Same.
13434 (rewrite_expr_tree_parallel): Same.
13435 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
13436 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
13437 * tree-ssa-structalias.c (process_all_all_constraints): Same.
13438 (make_constraints_to): Same.
13439 (handle_lhs_call): Same.
13440 (find_func_aliases_for_builtin_call): Same.
13441 (sort_fieldstack): Same.
13442 (check_for_overlaps): Same.
13443 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
13444 (vect_create_cond_for_unequal_addrs): Same.
13445 (vect_create_cond_for_lower_bounds): Same.
13446 (vect_create_cond_for_alias_checks): Same.
13447 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
13448 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
13449 (vect_make_slp_decision): Same.
13450 (vect_slp_bbs): Same.
13451 (duplicate_and_interleave): Same.
13452 (vect_transform_slp_perm_load): Same.
13453 (vect_schedule_slp): Same.
13454 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
13455 (vect_schedule_slp): Same.
13456 (duplicate_and_interleave): Same.
13457 * tree.c (build_vector_from_ctor): Same.
13458 (build_vector): Same.
13459 (check_vector_cst): Same.
13460 (check_vector_cst_duplicate): Same.
13461 (check_vector_cst_fill): Same.
13462 (check_vector_cst_stepped): Same.
13463 * tree.h (build_vector_from_ctor): Same.
13465 2021-07-20 Jakub Jelinek <jakub@redhat.com>
13468 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
13469 type from bool to int.
13470 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
13471 EASY_VECTOR_MSB case if either step or copies is not 1.
13472 (vspltis_shifted): Fix comment typo.
13473 (easy_altivec_constant): Change return type from bool to int, instead
13474 of returning true return byte size of the element mode that should be
13475 used to synthetize the constant.
13476 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
13477 that vspltis_shifted is 0, handle the case where easy_altivec_constant
13478 assumes using different vector mode from CONST_VECTOR's mode.
13479 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
13480 easy_altivec_constant to determine mode in which -1 >> -1 should be
13481 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
13483 2021-07-20 Richard Biener <rguenther@suse.de>
13486 * dwarf2out.h (dwarf_file_data): Add key member.
13487 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
13488 (dwarf_file_hasher::hash): Hash key.
13489 (lookup_filename): Remap the filename and store it in the
13490 filename member of dwarf_file_data when creating a new
13492 (file_name_acquire): Do not remap the filename again.
13493 (maybe_emit_file): Likewise.
13495 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
13497 * config/aarch64/aarch64-simd-builtins.def: Use two variant
13498 generators for all TBL/TBX intrinsics and rename to
13499 consistent forms: qtbl[1234] or qtbx[1234].
13500 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
13502 (aarch64_qtbl1<mode>): This.
13503 (aarch64_tbx1<mode>): Rename to...
13504 (aarch64_qtbx1<mode>): This.
13505 (aarch64_tbl2v16qi): Delete.
13506 (aarch64_tbl3<mode>): Rename to...
13507 (aarch64_qtbl2<mode>): This.
13508 (aarch64_tbx4<mode>): Rename to...
13509 (aarch64_qtbx2<mode>): This.
13510 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
13511 renamed qtbl1 and qtbl2 RTL patterns.
13512 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
13514 (vqtbl1_s8): Likewise.
13515 (vqtbl1_u8): Likewise.
13516 (vqtbl1q_p8): Likewise.
13517 (vqtbl1q_s8): Likewise.
13518 (vqtbl1q_u8): Likewise.
13519 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
13520 (vqtbx1_u8): Likewise.
13521 (vqtbx1_p8): Likewise.
13522 (vqtbx1q_s8): Likewise.
13523 (vqtbx1q_u8): Likewise.
13524 (vqtbx1q_p8): Likewise.
13525 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
13526 (vtbl1_u8): Likewise.
13527 (vtbl1_p8): Likewise.
13528 (vtbl2_s8): Likewise
13529 (vtbl2_u8): Likewise.
13530 (vtbl2_p8): Likewise.
13531 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
13532 (vtbl3_u8): Likewise.
13533 (vtbl3_p8): Likewise.
13534 (vtbl4_s8): Likewise.
13535 (vtbl4_u8): Likewise.
13536 (vtbl4_p8): Likewise.
13537 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
13538 (vtbx2_u8): Likewise.
13539 (vtbx2_p8): Likewise.
13540 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
13541 (vqtbl2_u8): Likewise.
13542 (vqtbl2_p8): Likewise.
13543 (vqtbl2q_s8): Likewise.
13544 (vqtbl2q_u8): Likewise.
13545 (vqtbl2q_p8): Likewise.
13546 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
13547 (vqtbx2_u8): Likewise.
13548 (vqtbx2_p8): Likewise.
13549 (vqtbx2q_s8): Likewise.
13550 (vqtbx2q_u8): Likewise.
13551 (vqtbx2q_p8): Likewise.
13552 (vtbx4_s8): Likewise.
13553 (vtbx4_u8): Likewise.
13554 (vtbx4_p8): Likewise.
13556 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
13559 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
13561 (define_peephole2 atomic_loaddi_fpu): Ditto.
13563 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13565 * config.gcc (riscv*-*-*): Detect which python is available.
13567 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13569 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
13570 (smul<mode>3_highpart): ... this.
13571 (mulhu_<mode>): Rename to...
13572 (umul<mode>3_highpart): ... this.
13573 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
13574 MULHU_V2DI, MULHU_V4SI): Adjust.
13576 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13578 PR tree-optimization/100696
13579 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
13580 * internal-fn.def (IFN_MULH): New internal function.
13581 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
13582 recog normal multiply highpart as IFN_MULH.
13583 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
13586 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13588 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
13589 (BTF_DEBUGGING_INFO): Likewise.
13590 * doc/tm.texi.in: Document the new macros.
13591 * doc/tm.texi: Regenerated.
13592 * toplev.c: Guard initialization of debug hooks.
13594 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13596 * flags.h (ctf_debuginfo_p): New function declaration.
13597 * opts.c (ctf_debuginfo_p): New function definition.
13599 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
13602 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
13603 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
13604 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
13605 * config/gcn/gcn-valu.md: Add a warning comment.
13606 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
13607 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
13608 * config/gcn/gcn.md: Add a warning comment.
13609 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
13610 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
13612 (EF_AMDGPU_XNACK): New.
13613 (EF_AMDGPU_SRAM_ECC): New.
13615 (copy_early_debug_info): Use elf_flags.
13616 (main): Handle -mxnack and -msram-ecc options.
13617 * doc/invoke.texi: Document -mxnack and -msram-ecc.
13619 2021-07-19 Andrew Pinski <apinski@marvell.com>
13622 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
13623 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
13625 2021-07-19 Richard Biener <rguenther@suse.de>
13627 PR tree-optimization/101505
13628 * tree-vect-patterns.c (vect_determine_precisions): Walk
13629 PHIs also for loop vectorization.
13631 2021-07-19 Richard Biener <rguenther@suse.de>
13633 * gimple.h (gimple_expr_type): Remove.
13634 * doc/gimple.texi: Remove gimple_expr_type documentation.
13636 2021-07-19 Richard Biener <rguenther@suse.de>
13638 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
13639 (ao_ref_init_from_vn_reference): Likewise.
13640 (fully_constant_reference): Likewise.
13641 (vn_reference_lookup_call): Do not set vr->type to random
13643 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
13645 * tree-vect-generic.c (expand_vector_piecewise): Pass in
13646 whether we expanded parallel.
13647 (expand_vector_parallel): Adjust.
13648 (expand_vector_addition): Likewise.
13649 (expand_vector_comparison): Likewise.
13650 (expand_vector_operation): Likewise.
13651 (expand_vector_scalar_condition): Likewise.
13652 (expand_vector_conversion): Likewise.
13654 2021-07-19 Richard Biener <rguenther@suse.de>
13656 * tree-vrp.c (register_edge_assert_for_2): Use the
13658 (vrp_folder::fold_predicate_in): Likewise.
13659 * vr-values.c (gimple_assign_nonzero_p): Likewise.
13660 (vr_values::extract_range_from_comparison): Likewise.
13661 (vr_values::extract_range_from_ubsan_builtin): Use the
13662 type of the first operand.
13663 (vr_values::extract_range_basic): Push down type
13664 computation, use the appropriate LHS.
13665 (vr_values::extract_range_from_assignment): Use the
13668 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
13671 * common/config/i386/i386-common.c (ix86_handle_option): For
13672 -mgeneral-regs-only, enable the GPR only instructions which are
13673 enabled implicitly by SSE ISAs unless they have been disabled
13676 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
13679 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
13680 ix86_avx_u128_mode_needed.
13681 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
13682 returns AVX register.
13684 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
13686 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
13687 (handle_const_call): Liekise
13688 (handle_pure_call): Liekise
13690 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
13692 PR tree-optimization/96542
13693 * range-op.cc (range_operator::wi_fold_in_parts): New.
13694 (range_operator::fold_range): Call wi_fold_in_parts.
13695 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
13696 * range-op.h (wi_fold_in_parts): Add prototype.
13698 2021-07-16 David Malcolm <dmalcolm@redhat.com>
13700 * doc/analyzer.texi: Add __analyzer_dump_state.
13702 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13704 * config/rs6000/rbtree.c: New file.
13705 * config/rs6000/rbtree.h: New file.
13707 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13709 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
13710 (typeinfo): Add restr field.
13711 (match_bracketed_pair): New function.
13712 (match_const_restriction): Implement.
13714 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13716 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
13718 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13720 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
13721 (basetype): Likewise.
13722 (typeinfo): Likewise.
13723 (handle_pointer): New function.
13724 (match_basetype): New stub function.
13725 (match_const_restriction): Likewise.
13726 (match_type): New function.
13728 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13730 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
13732 (advance_line): Likewise.
13733 (safe_inc_pos): Likewise.
13734 (match_identifier): Likewise.
13735 (match_integer): Likewise.
13736 (match_to_right_bracket): Likewise.
13738 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13740 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
13741 (ovld_file): Likewise.
13742 (header_file): Likewise.
13743 (init_file): Likewise.
13744 (defines_file): Likewise.
13745 (pgm_path): Likewise.
13746 (bif_path): Likewise.
13747 (ovld_path): Likewise.
13748 (header_path): Likewise.
13749 (init_path): Likewise.
13750 (defines_path): Likewise.
13751 (LINELEN): New macro.
13752 (linebuf): New variable.
13756 (bif_diag): New function.
13757 (ovld_diag): Likewise.
13759 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13761 * config/rs6000/rs6000-builtin-new.def: New.
13762 * config/rs6000/rs6000-overload.def: New.
13764 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13766 * config/rs6000/rs6000-gen-builtins.c: New.
13768 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13770 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
13771 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
13772 * gengtype-state.c (state_writer::write_state_file_list): Add a
13773 parameter to the fileslist expression for the number of build
13775 (read_state_files_list): Detect build headers and strip the
13776 initial "./" or ".\" from their names.
13777 * gengtype.c (build_headers): New global variable.
13778 (num_build_headers): Likewise.
13779 (open_base_files): Emit #include for each build header.
13780 (main): Detect and count build headers.
13781 * gengtype.h (build_headers): New extern variable.
13782 (num_build_headers): Likewise.
13784 2021-07-16 Richard Biener <rguenther@suse.de>
13786 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
13787 the type of the LHS.
13788 (find_bswap_or_nop_1): Likewise.
13789 (find_bswap_or_nop): Likewise.
13790 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
13792 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13793 Remove unused parameters, pass in the scalar type. Fix
13794 internal store function handling.
13795 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
13796 (vect_get_vector_types_for_stmt): Move down check for
13797 existing vector stmt after we've determined a scalar type.
13798 Pass down the used scalar type to vect_get_smallest_scalar_type.
13799 * tree-vect-generic.c (expand_vector_condition): Use
13800 the type of the LHS.
13801 (expand_vector_scalar_condition): Likewise.
13802 (expand_vector_operations_1): Likewise.
13803 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
13804 (vect_recog_dot_prod_pattern): Likewise.
13805 (vect_recog_sad_pattern): Likewise.
13806 (vect_recog_widen_op_pattern): Likewise.
13807 (vect_recog_widen_sum_pattern): Likewise.
13808 (vect_recog_mixed_size_cond_pattern): Likewise.
13810 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
13812 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
13813 (dump_eaf_flags): Dump EAF_NOT_RETURNED
13814 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
13815 and EAF_NOT_RETURNED.
13816 (modref_summary::useful_p): Likewise.
13817 (modref_summary_lto::useful_p): Likewise.
13818 (struct) modref_summary_lto: Use eaf_fleags_t.
13819 (deref_flags): Handle EAF_NOT_RETURNED.
13820 (struct escape_point): Use min_flags.
13821 (modref_lattice::init): Add EAF_NOT_RETURNED.
13822 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
13823 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
13825 (analyze_parms): Also analyze const functions; update conition on
13827 (modref_write): Update streaming.
13828 (read_section): Update streaming.
13829 (remap_arg_flags): Use eaf_flags_t.
13830 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
13831 * ipa-modref.h: (eaf_flags_t): New typedef.
13832 (struct modref_summary): Use eaf_flags_t.
13833 * tree-core.h (EAF_NOT_RETURNED): New constant.
13835 2021-07-16 Richard Biener <rguenther@suse.de>
13837 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
13838 the type of the LHS.
13839 (gimple_assign_nonnegative_warnv_p): Likewise.
13840 (gimple_call_nonnegative_warnv_p): Likewise. Return false
13841 if the call has no LHS.
13842 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
13843 * tree-eh.c (stmt_could_throw_1_p): Likewise.
13844 * tree-inline.c (insert_init_stmt): Likewise.
13845 * tree-ssa-loop-niter.c (get_val_for): Likewise.
13846 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
13848 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
13849 gassign *. Use the type of the lhs.
13850 (vn_nary_op_lookup_stmt): Adjust.
13851 (vn_nary_op_insert_stmt): Likewise.
13853 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
13855 * config/s390/predicates.md (bras_sym_operand): Accept all
13856 functions in 64-bit mode, use UNSPEC_PLT31.
13857 (larl_operand): Use UNSPEC_PLT31.
13858 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
13859 (legitimize_pic_address): Likewise.
13860 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
13862 (s390_delegitimize_address): Use UNSPEC_PLT31.
13863 (s390_output_addr_const_extra): Likewise.
13864 (print_operand): Add @PLT to TLS calls, handle %K.
13865 (s390_function_profiler): Mark __fentry__/_mcount as function,
13866 use %K, use UNSPEC_PLT31.
13867 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
13868 (s390_emit_call): Use UNSPEC_PLT31.
13869 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
13870 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
13871 (*movdi_64): Use %K.
13872 (reload_base_64): Likewise.
13873 (*sibcall_brc): Likewise.
13874 (*sibcall_brcl): Likewise.
13875 (*sibcall_value_brc): Likewise.
13876 (*sibcall_value_brcl): Likewise.
13878 (*brasl): Likewise.
13879 (*bras_r): Likewise.
13880 (*brasl_r): Likewise.
13881 (*bras_tls): Likewise.
13882 (*brasl_tls): Likewise.
13883 (main_base_64): Likewise.
13884 (reload_base_64): Likewise.
13885 (@split_stack_call<mode>): Likewise.
13887 2021-07-16 Richard Biener <rguenther@suse.de>
13889 PR tree-optimization/101467
13890 * tree-vect-stmts.c (vect_gen_while): Properly guard
13891 make_temp_ssa_name usage.
13893 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
13895 * config.gcc: Don't use forked print-sysroot-suffix.sh and
13896 t-sysroot-suffix for C-SKY.
13897 * config/csky/print-sysroot-suffix.sh: Delete.
13898 * config/csky/t-csky-linux: Delete.
13899 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
13900 instead of CSKY_MULTILIB_DIRNAMES.
13902 2021-07-16 Richard Biener <rguenther@suse.de>
13904 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
13905 conversion issues with the partial reduction of the reused
13906 vector accumulator.
13908 2021-07-16 Richard Biener <rguenther@suse.de>
13910 * config/i386/i386-options.c (ix86_option_override_internal): Set
13911 param_vect_partial_vector_usage to zero if not set.
13913 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
13916 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
13917 (VALID_INT_MODE_P): Add SDmode and DDmode.
13918 Add TDmode for TARGET_64BIT.
13919 (VALID_DFP_MODE_P): Remove.
13920 * config/i386/i386.c (ix86_hard_regno_mode_ok):
13921 Do not use VALID_DFP_MODE_P.
13923 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
13925 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
13927 (fold_using_range::fold_stmt): Ditto.
13928 (fold_using_range::range_of_range_op): Ditto.
13929 (fold_using_range::range_of_phi): Ditto.
13930 (fold_using_range::range_of_call): Ditto.
13931 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
13932 (fold_using_range::range_of_builtin_call): Ditto.
13933 (fold_using_range::range_of_cond_expr): Ditto.
13934 * gimple-range-fold.h (gimple_range_type): New.
13936 2021-07-15 Martin Sebor <msebor@redhat.com>
13938 PR middle-end/97027
13939 * tree-ssa-strlen.c (handle_assign): New function.
13940 (maybe_warn_overflow): Add argument.
13941 (nonzero_bytes_for_type): New function.
13942 (count_nonzero_bytes): Handle more tree types. Call
13943 nonzero_bytes_for_tye.
13944 (count_nonzero_bytes): Handle types.
13945 (handle_store): Handle stores from function calls.
13946 (strlen_check_and_optimize_call): Move code to handle_assign. Call
13947 it for assignments from function calls.
13949 2021-07-15 David Malcolm <dmalcolm@redhat.com>
13954 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
13956 2021-07-15 David Malcolm <dmalcolm@redhat.com>
13958 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
13960 2021-07-15 Martin Sebor <msebor@redhat.com>
13964 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
13965 (operand_compare::verify_hash_value): Same.
13966 * tree-core.h (OEP_DECL_NAME): New.
13968 2021-07-15 Martin Jambor <mjambor@suse.cz>
13970 * profile-count.h (profile_count::value): Change the return type to
13972 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
13974 * tree-cfg.c (dump_function_to_file): Likewise.
13976 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
13979 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
13980 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
13983 2021-07-15 Richard Biener <rguenther@suse.de>
13985 * tree-vectorizer.h (vect_gen_while): Match up with
13986 vect_gen_while_not.
13987 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
13988 of vect_gen_while_not.
13989 (vect_gen_while_not): Adjust.
13990 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
13992 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
13994 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
13995 (ranger_cache::range_of_def): Call adjust_range.
13996 (ranger_cache::entry_range): Same.
13997 * gimple-range-cache.h (non_null_ref::adjust_range): New.
13998 * gimple-range.cc (gimple_ranger::range_of_expr): Call
14000 (gimple_ranger::range_on_entry): Same.
14002 2021-07-15 Tamar Christina <tamar.christina@arm.com>
14005 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14007 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
14009 2021-07-15 Tamar Christina <tamar.christina@arm.com>
14012 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14014 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
14015 (sdot_prod, udot_prod): ...These.
14016 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
14017 (aarch64_<sur>dot<vsi2qi>): Rename to...
14018 (<sur>dot_prod<vsi2qi>): ...This.
14019 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
14022 2021-07-15 Jakub Jelinek <jakub@redhat.com>
14024 PR middle-end/101437
14025 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
14026 types even if they have non-BLKmode TYPE_MODE.
14028 2021-07-15 Richard Biener <rguenther@suse.de>
14031 * gcc.c (process_command): Process -gtoggle like process_options
14032 would after parsing options.
14034 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14036 * cfgexpand.c (expand_asm_loc): Adjust.
14037 (expand_asm_stmt): Likewise.
14038 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
14039 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
14040 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
14041 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
14042 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
14043 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
14044 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
14045 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
14046 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
14047 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
14048 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
14049 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
14050 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
14051 * doc/tm.texi: Regenerate.
14052 * target.def: Add location argument to md_asm_adjust.
14054 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14056 * tree-diagnostic.c (diagnostic_report_current_function): Use the
14057 diagnostic's location, not input_location.
14059 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14061 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
14063 (expand_asm_stmt): Likewise.
14065 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
14067 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
14068 addressed memory rtx, if any.
14069 (rs6000_split_multireg_move): Fix code formatting.
14070 Handle MMA build built-ins with operands in adjacent memory locations.
14072 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
14074 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
14077 2021-07-14 Jason Merrill <jason@redhat.com>
14079 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
14081 2021-07-14 Jason Merrill <jason@redhat.com>
14083 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
14086 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14088 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
14089 Remove erroneous line.
14091 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
14093 * params.opt (param_evrp_mode): Change default.
14095 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14097 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
14098 (sdot_prod, udot_prod): ...These.
14099 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
14100 (aarch64_<sur>dot<vsi2qi>): Rename to...
14101 (<sur>dot_prod<vsi2qi>): ...This.
14102 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
14105 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14107 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
14109 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14111 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
14113 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14115 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
14117 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14119 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
14120 (usdot_prod<vsi2qi>): ... This.
14121 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
14122 (usdot_prod): ...This.
14123 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
14124 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
14126 (@<sur>dot_prod<vsi2qi>): ...This.
14127 * config/aarch64/aarch64-sve-builtins-base.cc
14128 (svusdot_impl::expand): Use it.
14130 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14132 * optabs.def (usdot_prod_optab): New.
14133 * doc/md.texi: Document it and clarify other dot prod optabs.
14134 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
14135 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
14136 * optabs.c (expand_widen_pattern_expr): Likewise.
14137 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14138 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
14139 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
14141 (vect_widened_op_tree): Optionally ignore
14143 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
14145 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
14148 * config/i386/driver-i386.c (host_detect_local_cpu): Check
14149 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
14150 Enable UINTR only for 64-bit codegen.
14151 * config/i386/i386-options.c
14152 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
14154 * config/i386/i386.h (ARCH_ARG): New.
14155 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
14156 "[arch|tune] 64" for 64-bit codegen.
14158 2021-07-14 Richard Biener <rguenther@suse.de>
14160 PR tree-optimization/101445
14161 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
14162 of the IV in the correct direction for negative stride
14165 2021-07-14 Jakub Jelinek <jakub@redhat.com>
14168 * godump.c (godump_str_hash): New type.
14169 (godump_container::pot_dummy_types): Use string_hash instead of
14170 ptr_hash in the hash_set.
14172 2021-07-14 Richard Biener <rguenther@suse.de>
14174 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
14175 vector types where the old vector type has a multiple of
14176 the new vector type elements.
14177 (vect_create_partial_epilog): New function, split out from...
14178 (vect_create_epilog_for_reduction): ... here.
14179 (vect_transform_cycle_phi): Reduce the re-used accumulator
14180 to the new vector type.
14182 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14184 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
14185 non-'t'-sized arg check.
14187 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14189 * tree-cfg.c (cleanup_dead_labels_eh): Update
14190 post_landing_pad label upon change of landing pad block's
14192 (cleanup_dead_labels): Check that a removed label is not that
14195 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
14197 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
14199 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
14200 Add Neon to general purpose register case for zero-extend
14202 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
14203 case to prevent some cases opting to go through memory.
14204 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
14205 * rtl.c (rtvec_series_p): Define predicate to determine
14206 whether a vector contains a linear series of integers.
14207 * rtl.h (rtvec_series_p): Define.
14208 * rtlanal.c (vec_series_lowpart_p): Define predicate to
14209 determine if a vector selection is equivalent to the low part
14211 * rtlanal.h (vec_series_lowpart_p): Define.
14212 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
14213 Add vec_select -> subreg simplification.
14215 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
14217 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
14218 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
14219 _mm_test_mix_ones_zeros): New.
14221 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
14222 Richard Biener <rguenther@suse.de>
14224 * gimple.c (gimple_could_trap_p_1): Make S argument a
14225 "const gimple*". Preserve constness in call to
14226 gimple_asm_volatile_p.
14227 (gimple_could_trap_p): Make S argument a "const gimple*".
14228 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
14229 Update function prototypes.
14231 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14233 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
14234 (_loop_vec_info::main_loop_edge): New field.
14235 (_loop_vec_info::skip_main_loop_edge): Likewise.
14236 (_loop_vec_info::skip_this_loop_edge): Likewise.
14237 (_loop_vec_info::reusable_accumulators): Likewise.
14238 (_stmt_vec_info::reduc_scalar_results): Likewise.
14239 (_stmt_vec_info::reused_accumulator): Likewise.
14240 (vect_get_main_loop_result): Declare.
14241 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
14242 reduc_scalar_inputs.
14243 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
14244 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
14245 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
14246 skip_main_loop_edge and skip_this_loop_edge fields.
14247 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
14248 (vect_emit_reduction_init_stmts): New function.
14249 (get_initial_def_for_reduction): Use it.
14250 (get_initial_defs_for_reduction): Likewise. Change the vinfo
14251 parameter to a loop_vec_info.
14252 (vect_create_epilog_for_reduction): Store the scalar results
14253 in the reduc_info. If an epilogue loop is reusing an accumulator
14254 from the main loop, and if the epilogue loop can also be skipped,
14255 try to place the reduction code in the join block. Record
14256 accumulators that could potentially be reused by epilogue loops.
14257 (vect_transform_cycle_phi): When vectorizing epilogue loops,
14258 try to reuse accumulators from the main loop. Record the initial
14259 value in reduc_info for non-SLP reductions too.
14261 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14263 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
14264 adjustment handling. Take the neutral value as an argument,
14265 in place of the code argument.
14266 (vect_transform_cycle_phi): Update accordingly. Handle the
14267 initial values of cond reductions separately from code reductions.
14268 Choose the adjustment here rather than in
14269 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
14271 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14273 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
14274 (neutral_op_for_reduction): ...this, providing a more general
14276 (vect_create_epilog_for_reduction): Update accordingly.
14277 (vectorizable_reduction): Likewise.
14278 (vect_transform_cycle_phi): Likewise.
14280 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14282 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
14283 reduc_info instead of the original stmt_vec_info.
14284 (vect_transform_cycle_phi): Update accordingly.
14286 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14288 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
14289 reduc_info as an additional parameter.
14290 (vect_transform_cycle_phi): Update accordingly.
14292 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14294 * tree-vectorizer.h: Include tree-ssa-operands.h.
14295 (vect_phi_initial_value): New function.
14296 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
14297 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
14298 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14299 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
14301 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14303 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
14304 the phi results to vectype after creating them. Remove later
14305 conversion code that thus becomes redundant.
14307 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14309 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
14310 the new_phis vector with a reduc_inputs vector. Combine handling
14311 of reduction chains and ncopies > 1.
14313 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14315 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
14316 scalar_results to group_size elements after reducing down from
14317 N*group_size elements. Construct an array_slice of the live-out
14318 stmts and assert that there is one stmt per scalar result.
14320 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14322 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
14323 nested_in_vect_loop and use double_reduc everywhere. Remove dead
14324 assignment to "loop".
14326 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14328 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
14329 vector types first. For scalar types, consider both the preferred
14330 vector mode and the alternative vector modes.
14331 * optabs-query.c (can_vec_mask_load_store_p): Use the same
14332 structure as above, in particular using related_vector_mode
14333 for modes provided by autovectorize_vector_modes.
14335 2021-07-13 Jakub Jelinek <jakub@redhat.com>
14336 Richard Biener <rguenther@suse.de>
14338 PR tree-optimization/101419
14339 * tree-pass.h (PROP_objsz): Define.
14340 (make_pass_early_object_sizes): Declare.
14341 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
14342 there to pass_early_object_sizes, drop parameter.
14343 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
14344 drop parameter, move pass_post_ipa_warn right after that.
14345 * tree-object-size.c (pass_object_sizes::execute): Rename to...
14346 (object_sizes_execute): ... this. Add insert_min_max_p argument.
14347 (pass_data_object_sizes): Move after object_sizes_execute.
14348 (pass_object_sizes): Likewise. In execute method call
14349 object_sizes_execute, drop set_pass_param method and insert_min_max_p
14350 non-static data member and its initializer in the ctor.
14351 (pass_data_early_object_sizes, pass_early_object_sizes,
14352 make_pass_early_object_sizes): New.
14353 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
14354 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
14356 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
14359 * config/riscv/constraints.md ("S"): Update description and remove
14361 * doc/md.texi (Machine Constraints): Document the 'S' constraints
14364 2021-07-13 Richard Biener <rguenther@suse.de>
14367 2021-07-12 Richard Biener <rguenther@suse.de>
14369 * tree-vect-slp.c (vect_slp_region): Show the number of
14370 SLP graph entries in the optimization message.
14372 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
14374 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
14376 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
14377 return type to long.
14378 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
14381 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14383 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
14384 Query relation between the 2 operands and use it.
14386 2021-07-12 Sergei Trofimovich <siarheit@google.com>
14388 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
14390 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14393 * config/i386/predicates.md (vec_setm_sse41_operand):
14394 Rename from vec_setm_operand.
14395 (vec_setm_avx2_operand): New predicate.
14396 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
14397 Use vec_setm_sse41_operand as operand 2 predicate.
14398 (vec_set<V_256_512:mode): New expander.
14399 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
14400 as operand 2 predicate.
14402 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14404 PR tree-optimization/101335
14405 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
14407 2021-07-12 Andrew Pinski <apinski@marvell.com>
14409 * tree-ssa-phiopt.c (match_simplify_replacement): Move
14410 insert of the sequence before the movement of the
14411 statement. Check if to see if the statement is used
14412 outside of the original phi to see if we should move it.
14414 2021-07-12 Richard Biener <rguenther@suse.de>
14416 * dump-context.h (debug_dump_context::debug_dump_context):
14417 Add FILE * parameter defaulted to stderr.
14418 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
14419 * tree-vect-slp.c (dot_slp_tree): New functions.
14421 2021-07-12 Richard Biener <rguenther@suse.de>
14423 PR tree-optimization/101373
14424 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
14425 references when the BB may not return.
14426 (compute_avail): Pass in the function we're working on and
14427 replace cfun references with it. Externally throwing
14428 const calls also possibly terminate the function.
14429 (pass_pre::execute): Pass down the function we're working on.
14430 * gcse.c (compute_hash_table_work): Externally throwing
14431 const/pure calls also need record_last_mem_set_info.
14432 * postreload-gcse.c (record_opr_changes): Looping or externally
14433 throwing const/pure calls also need record_last_mem_set_info.
14435 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14437 * recog.c (memory_address_addr_space_p): Change the type to bool.
14438 Return true/false instead of 1/0.
14439 (offsettable_memref_p): Ditto.
14440 (offsettable_nonstrict_memref_p): Ditto.
14441 (offsettable_address_addr_space_p): Ditto.
14442 Change the type of addressp indirect function to bool.
14443 * recog.h (memory_address_addr_space_p): Change the type to bool.
14444 (strict_memory_address_addr_space_p): Ditto.
14445 (offsettable_memref_p): Ditto.
14446 (offsettable_nonstrict_memref_p): Ditto.
14447 (offsettable_address_addr_space_p): Ditto.
14448 * reload.c (maybe_memory_address_addr_space_p): Ditto.
14449 (strict_memory_address_addr_space_p): Change the type to bool.
14450 Return true/false instead of 1/0.
14451 (maybe_memory_address_addr_space_p): Change the type to bool.
14453 2021-07-12 Richard Biener <rguenther@suse.de>
14455 * tree-vect-slp.c (vect_slp_region): Show the number of
14456 SLP graph entries in the optimization message.
14458 2021-07-12 Richard Biener <rguenther@suse.de>
14460 PR tree-optimization/101394
14461 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
14462 copies from abnormals for a full redundancy.
14464 2021-07-12 Richard Biener <rguenther@suse.de>
14466 PR middle-end/101423
14467 * gimple.c (gimple_could_trap_p_1): Internal function calls
14469 * tree-eh.c (tree_could_trap_p): Likewise.
14471 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14474 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
14476 (vmulq_n_u32): Likewise.
14477 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
14478 (vmulq_n_f32): Likewise.
14479 (vmul_n_f16): Likewise.
14480 (vmulq_n_f16): Likewise.
14482 2021-07-12 Martin Liska <mliska@suse.cz>
14484 PR sanitizer/101425
14485 * gcc.c (check_offload_target_name): Call
14486 candidates_list_and_hint only if we have a candidate.
14488 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14491 * config/arm/neon.md (vec_init): Move to ...
14492 * config/arm/vec-common.md (vec_init): ... here.
14493 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
14495 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
14497 PR tree-optimization/101403
14498 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
14499 signedness of the shift is not the same as the signedness of
14500 the type extension.
14502 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
14503 Uroš Bizjak <ubizjak@gmail.com>
14505 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
14506 divmod of a constant numerator with new define_insn_and_split.
14508 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
14511 * config/i386/i386-expand.c (ix86_expand_call): If a call is
14512 to a non-local-binding, or local but to a public symbol, then
14513 assume that it might be indirected via the lazy symbol binder.
14514 Mark R10 and R10 as clobbered in that case.
14516 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
14519 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
14520 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
14521 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
14523 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
14525 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
14526 log2 range of operands[3] to [1,31].
14527 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
14529 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14531 * doc/md.texi: Don't split @smallexample in multiple @groups.
14533 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14535 * doc/md.texi: Add missing 'see' word.
14537 2021-07-09 Andrew Pinski <apinski@marvell.com>
14539 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
14540 to take sequence and gimple_match_op. Accept the case where
14541 op is a SSA_NAME and one statement in the sequence.
14542 Also allow constants.
14543 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
14544 Update call to phiopt_early_allow. Discard the sequence if not
14547 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
14552 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
14553 instead of ISA_HAS_MSA.
14554 (mips_expand_vec_unpack): Likewise.
14555 (mips_expand_vector_init): Likewise.
14557 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
14559 * config/rs6000/vsx.md (mods_<mode>): Rename to...
14560 (mod<mode>3): ... this.
14561 (modu_<mode>): Rename to...
14562 (umod<mode>3): ... this.
14563 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
14564 MODU_V4SI): Adjust.
14566 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
14568 * config/h8300/shiftrotate.md (variable shifts): Expose condition
14569 code handling for the test before the loop.
14571 2021-07-08 Martin Jambor <mjambor@suse.cz>
14574 * ipa-sra.c (class isra_call_summary): New member
14575 m_before_any_store, initialize it in the constructor.
14576 (isra_call_summary::dump): Dump the new field.
14577 (ipa_sra_call_summaries::duplicate): Copy it.
14578 (process_scan_results): Set it.
14579 (isra_write_edge_summary): Stream it.
14580 (isra_read_edge_summary): Likewise.
14581 (param_splitting_across_edge): Only override
14582 safe_to_import_accesses if m_before_any_store is set.
14584 2021-07-08 Martin Sebor <msebor@redhat.com>
14586 PR bootstrap/101374
14587 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
14588 Use Object Size Type 0 instead of 1.
14590 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
14592 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
14595 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
14597 * match.pd: Simplify an extend-operate-truncate sequence involving
14600 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
14601 Richard Biener <rguenther@suse.de>
14603 PR tree-optimization/40210
14604 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
14605 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
14606 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
14608 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
14611 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
14613 * config/i386/mmx.md (V_32): New mode iterator.
14614 (mov<V_32:mode>): Use V_32 mode iterator.
14615 (*mov<V_32:mode>_internal): Ditto.
14616 (*push<V_32:mode>2_rex64): Ditto.
14617 (*push<V_32:mode>2): Ditto.
14618 (movmisalign<V_32:mode>): Ditto.
14619 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
14620 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
14621 (vec_unpacks_lo_v4qi): New expander.
14622 (vec_unpacks_hi_v4qi): Ditto.
14623 (vec_unpacku_lo_v4qi): Ditto.
14624 (vec_unpacku_hi_v4qi): Ditto.
14625 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
14626 (VALID_INT_MODE_P): Ditto.
14628 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
14631 * config/rs6000/rs6000.md (udivti3): New insn.
14632 (divti3): New insn.
14633 (umodti3): New insn.
14634 (modti3): New insn.
14636 2021-07-07 Martin Sebor <msebor@redhat.com>
14638 PR tree-optimization/100137
14639 PR tree-optimization/99121
14640 PR tree-optimization/97027
14641 * builtins.c (access_ref::access_ref): Also set offmax.
14642 (access_ref::offset_in_range): Define new function.
14643 (access_ref::add_offset): Set offmax.
14644 (access_ref::inform_access): Handle access_none.
14645 (handle_mem_ref): Clear ostype.
14646 (compute_objsize_r): Handle ASSERT_EXPR.
14647 * builtins.h (struct access_ref): Add offmax member.
14648 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
14649 compute_objsize() and simplify.
14651 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
14653 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
14654 and VSX_BUILTIN_STXVP.
14656 2021-07-07 Martin Sebor <msebor@redhat.com>
14659 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
14660 a stray %K from error_at() missed in r12-2088.
14662 2021-07-07 Richard Biener <rguenther@suse.de>
14664 PR tree-optimization/99728
14665 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
14667 (mem_refs_may_alias_p): Add assert we handled aggregate
14669 (sm_seq_valid_bb): Give up when running into aggregate copies.
14670 (ref_indep_loop_p): Handle aggregate copies as never
14671 being invariant themselves but allow other refs to be
14672 disambiguated against them.
14673 (can_sm_ref_p): Do not try to apply store-motion to aggregate
14676 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
14679 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
14682 2021-07-06 Martin Sebor <msebor@redhat.com>
14684 * gimple-pretty-print.c (percent_G_format): Remove.
14685 * tree-diagnostic.c (default_tree_printer): Remove calls.
14686 * tree-pretty-print.c (percent_K_format): Remove.
14687 * tree-pretty-print.h (percent_K_format): Remove.
14689 2021-07-06 Martin Sebor <msebor@redhat.com>
14691 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
14692 Remove %K and use error_at.
14693 (aarch64_expand_fcmla_builtin): Same.
14694 (aarch64_expand_builtin_tme): Same.
14695 (aarch64_expand_builtin_memtag): Same.
14696 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
14697 (arm_expand_builtin): Same.
14698 * config/arm/arm.c (bounds_check): Same.
14700 2021-07-06 Martin Sebor <msebor@redhat.com>
14702 * builtins.c (warn_string_no_nul): Remove %G.
14703 (maybe_warn_for_bound): Same.
14704 (warn_for_access): Same.
14705 (check_access): Same.
14706 (check_strncat_sizes): Same.
14707 (expand_builtin_strncat): Same.
14708 (expand_builtin_strncmp): Same.
14709 (expand_builtin): Same.
14710 (expand_builtin_object_size): Same.
14711 (warn_dealloc_offset): Same.
14712 (maybe_emit_free_warning): Same.
14713 * calls.c (maybe_warn_alloc_args_overflow): Same.
14714 (maybe_warn_nonstring_arg): Same.
14715 (maybe_warn_rdwr_sizes): Same.
14716 * expr.c (expand_expr_real_1): Remove %K.
14717 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
14718 (gimple_fold_builtin_strncat): Same.
14719 * gimple-ssa-sprintf.c (format_directive): Same.
14720 (handle_printf_call): Same.
14721 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
14722 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
14723 (maybe_diag_access_bounds): Same. Call gimple_location.
14724 (check_bounds_or_overlap): Same.
14725 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
14726 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
14727 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
14728 (maybe_diag_stxncpy_trunc): Same.
14729 (handle_builtin_stxncpy_strncat): Same.
14730 (maybe_warn_pointless_strcmp): Same.
14731 * tree-ssa-uninit.c (maybe_warn_operand): Same.
14733 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
14736 * config/i386/predicates.md (vec_setm_operand): Enable
14737 register_operand for TARGET_SSE4_1.
14738 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
14739 as operand 2 predicate. Call ix86_expand_vector_set_var
14740 for non-constant index operand.
14741 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
14742 Call ix86_expand_vector_set_var for non-constant index operand.
14744 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
14746 * config/h8300/jumpcall.md (*branch): When possible, generate
14747 the comparison in CCZN mode.
14748 * config/h8300/predicates.md (simple_memory_operand): Reject all
14749 auto-increment addressing modes.
14751 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
14753 PR bootstrap/100246
14754 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
14757 2021-07-06 Richard Biener <rguenther@suse.de>
14759 * doc/md.texi (vec_fmaddsub<mode>4): Document.
14760 (vec_fmsubadd<mode>4): Likewise.
14761 * optabs.def (vec_fmaddsub$a4): Add.
14762 (vec_fmsubadd$a4): Likewise.
14763 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
14764 (IFN_VEC_FMSUBADD): Likewise.
14765 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
14766 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
14767 (addsub_pattern::build): Likewise.
14768 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
14769 and CFN_VEC_FMSUBADD are not transparent for permutes.
14770 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
14771 (vec_fmsubadd<mode>4): Likewise.
14773 2021-07-06 Richard Biener <rguenther@suse.de>
14775 * doc/invoke.texi (fmove-loop-stores): Document.
14776 * common.opt (fmove-loop-stores): New option.
14777 * opts.c (default_options_table): Enable -fmove-loop-stores
14778 at -O1 but not -Og.
14779 * tree-ssa-loop-im.c (pass_lim::execute): Pass
14780 flag_move_loop_stores instead of true to
14781 loop_invariant_motion_in_fun.
14783 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
14785 * doc/install.texi: Document --with-dsymutil.
14787 2021-07-06 Andrew Pinski <apinski@marvell.com>
14789 PR tree-optimization/101256
14790 * dbgcnt.def (phiopt_edge_range): New counter.
14791 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
14792 Check to make sure the new name is defined in the same
14793 bb as the conditional before duplicating range info.
14794 Also add debug counter.
14796 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
14798 PR rtl-optimization/100328
14799 * config/i386/i386-options.c (ix86_option_override_internal):
14800 Set param_ira_consider_dup_in_all_alts to 0.
14802 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
14804 PR rtl-optimization/100328
14805 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
14807 * ira.c (ira_get_dup_out_num): Adjust as parameter
14808 param_ira_consider_dup_in_all_alts.
14809 * params.opt (ira-consider-dup-in-all-alts): New.
14810 * ira-conflicts.c (process_regs_for_copy): Add one parameter
14811 single_input_op_has_cstr_p.
14812 (get_freq_for_shuffle_copy): New function.
14813 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
14814 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
14816 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
14818 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
14819 generate condition code aware RTL directly.
14821 2021-07-05 Andrew Pinski <apinski@marvell.com>
14823 PR tree-optimization/101039
14824 * match.pd (A CMP 0 ? A : -A): New patterns.
14825 * tree-ssa-phiopt.c (abs_replacement): Delete function.
14826 (tree_ssa_phiopt_worker): Don't call abs_replacement.
14827 Update comment about abs_replacement.
14829 2021-07-05 Andrew Pinski <apinski@marvell.com>
14831 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
14832 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
14834 2021-07-05 Andrew Pinski <apinski@marvell.com>
14836 * tree-ssa-phiopt.c (match_simplify_replacement):
14837 Add early_p argument. Call gimple_simplify_phiopt
14838 instead of gimple_simplify.
14839 (tree_ssa_phiopt_worker): Update call to
14840 match_simplify_replacement and allow unconditionally.
14841 (phiopt_early_allow): New function.
14842 (gimple_simplify_phiopt): New function.
14844 2021-07-05 Andrew Pinski <apinski@marvell.com>
14846 PR middle-end/101237
14847 * fold-const.c (negate_expr_p): Remove call to element_mode
14848 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
14849 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
14850 (fold_negate_expr_1): Likewise.
14851 (const_unop): Likewise.
14852 (fold_cond_expr_with_comparison): Likewise.
14853 (fold_binary_loc): Likewise.
14854 (fold_ternary_loc): Likewise.
14855 (tree_call_nonnegative_warnv_p): Likewise.
14856 * match.pd (-(A + B) -> (-B) - A): Likewise.
14858 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
14860 * configure.ac: Handle --with-dsymutil in the same way as we
14861 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
14862 Extract the type and version for the dsymutil configured or
14863 found by the default searches.
14864 * config.in: Regenerated.
14865 * configure: Regenerated.
14866 * collect2.c (do_dsymutil): Handle locating dsymutil in the
14867 same way as for the assembler and linker.
14868 * config/darwin.h (DSYMUTIL): Delete.
14869 * gcc.c: Report a configured dsymutil correctly.
14870 * exec-tool.in: Allow for dsymutil.
14872 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
14874 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
14875 Handle V4QI and V2HI modes.
14876 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
14877 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
14878 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
14879 Handle 4-byte dual operands with TARGET_XOP and single operands
14880 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
14881 mmx_pshufbv4qi3 for TARGET_SSSE3.
14882 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
14883 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
14884 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
14885 (expand_vec_perm_even_odd_1): Handle V4QI mode.
14886 (expand_vec_perm_broadcast_1): Handle V4QI mode.
14887 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
14888 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
14889 (mmx_pshufbv4qi3): Ditto.
14890 (*mmx_pblendw32): Ditto.
14891 (*mmx_pblendw64): Rename from *mmx_pblendw.
14892 (mmx_punpckhbw_low): New insn_and_split pattern.
14893 (mmx_punpcklbw_low): Ditto.
14895 2021-07-05 Richard Biener <rguenther@suse.de>
14897 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
14898 set LOOP_C_INFINITE on the vectorized loop.
14900 2021-07-05 Richard Biener <rguenther@suse.de>
14902 PR middle-end/101291
14903 * cfgloopmanip.c (loop_version): Set the loop copy of the
14904 versioned loop to the new loop.
14906 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
14909 * config.gcc: Ensure that Darwin biarch definitions are
14910 added before i386.h.
14911 * config/i386/darwin.h (TARGET_64BIT): Remove.
14912 (PR80556_WORKAROUND): New.
14913 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
14914 (DARWIN_SUBARCH_SPEC): New.
14915 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
14916 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
14917 (REAL_LIBGCC_SPEC): Remove.
14918 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
14919 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
14920 (REAL_LIBGCC_SPEC): Remove.
14922 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
14924 PR middle-end/101294
14925 * expr.c (store_constructor): Don't use vec_duplicate on vector.
14927 2021-07-02 Martin Sebor <msebor@redhat.com>
14929 PR middle-end/98871
14930 PR middle-end/98512
14931 * diagnostic.c (get_any_inlining_info): New.
14932 (update_effective_level_from_pragmas): Handle inlining context.
14933 (diagnostic_enabled): Same.
14934 (diagnostic_report_diagnostic): Same.
14935 * diagnostic.h (struct diagnostic_info): Add ctor.
14936 (struct diagnostic_context): Add new member.
14937 * tree-diagnostic.c (set_inlining_locations): New.
14938 (tree_diagnostics_defaults): Set new callback pointer.
14940 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
14942 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
14944 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
14945 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
14946 lxvp and stxvp built-ins.
14947 (mma_init_builtins): Handle lxvp and stxvp built-ins.
14948 (builtin_function_type): Likewise.
14949 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
14951 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14953 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
14954 additional argument for the code.
14955 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
14956 argument for the code. Just return if the ZN bits are useful or
14957 not rather than the old style CC_* enums.
14958 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
14959 more generic shiftqi patterns.
14960 (shifthi_noscratch, shiftsi_noscratch): Similarly.
14961 (shiftqi_noscratch_set_flags): New pattern.
14962 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
14964 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
14966 PR tree-optimization/101223
14967 * range-op.cc (build_lt): Add -1 for signed values.
14968 (built_gt): Subtract -1 for signed values.
14970 2021-07-02 David Faust <david.faust@oracle.com>
14972 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
14973 (btf_asm_type): Likewise.
14975 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14977 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
14978 an array of rtx rather than a pointer to rtx. Add code argument.
14979 (compute_a_shift_length): Similarly.
14980 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
14981 shift itself isn't an operand. Create dummy operand[0] to carry
14982 a mode and pass a suitable rtx code to compute_a_shift_length.
14983 (get_shift_alg): Adjust operand number of clobber in output templates.
14984 (output_a_shift): Make first argument an array of rtx rather than
14985 a pointer to rtx. Add code argument for the type of shift.
14986 Adjust now that the shift itself is no longer an operand.
14987 (compute_a_shift_length): Similarly.
14988 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
14989 iterator rather than nshift_operator.
14990 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
14991 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
14992 and compute_a_shift_length.
14993 (shiftqi_noscratch_clobber_flags): Likewise.
14994 (shifthi_noscratch_clobber_flags): Likewise.
14995 (shiftsi_noscratch_clobber_flags): Likewise.
14997 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
15000 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
15001 dsymutil for BTF/CTF.
15003 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
15006 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
15007 segment to include BTF.
15008 (BTF_INFO_SECTION_NAME): New.
15010 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15012 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
15013 (small_data_operand, memreg_operand, small_insn_p): Likewise.
15014 * config/m32r/m32r.c (call_operand): Adjust return type.
15015 (small_data_operand, memreg_operand): Likewise.
15017 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15019 * config/frv/frv-protos.h (integer_register_operand): Adjust return
15021 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
15022 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
15023 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
15024 (move_source_operand, move_destination_operand): Likewise.
15025 (condexec_source_operand, condexec_dest_operand): Likewise.
15026 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
15027 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
15028 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
15029 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
15030 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
15031 (even_fpr_operand, odd_fpr_operand): Likewise.
15032 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
15033 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
15034 (uint4_operand, uint1_operand, int_2word_operand): Likewise
15035 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
15036 (relational_operator, float_relational_operator): Likewise.
15037 (ccr_eqne_operator, minmax_operator): Likewise.
15038 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
15039 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
15040 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
15041 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
15042 (quad_acc_operand, accg_operand): Likewise.
15044 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15046 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
15047 return type to a bool.
15048 (nonimmediate_nonstack_operand): Likewise.
15049 (xstormy16_splittable_below100_operand): Likewise.
15050 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
15052 (xstormy16_splittable_below100_operand): Likewise.
15054 2021-07-02 Richard Biener <rguenther@suse.de>
15056 PR tree-optimization/101293
15057 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
15058 with combined offsets.
15059 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
15060 combined with the rest of the offset.
15062 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
15064 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
15065 PIC encodings for PE-COFF targets.
15067 2021-07-02 Jakub Jelinek <jakub@redhat.com>
15070 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
15071 Return nullptr for TImode inner mode.
15073 2021-07-02 Richard Biener <rguenther@suse.de>
15075 PR tree-optimization/101280
15076 PR tree-optimization/101173
15077 * gimple-loop-interchange.cc
15078 (tree_loop_interchange::valid_data_dependences): Properly
15079 guard all dependence checks with DDR_REVERSED_P or its
15082 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
15084 * config/i386/i386-expand.c (ix86_expand_builtin):
15085 Add branch to clear odata when ZF is set for asedecenc_expand
15086 and wideaesdecenc_expand.
15088 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
15090 * config/i386/gcc-auto-profile: regenerate
15092 2021-07-02 liuhongt <hongtao.liu@intel.com>
15094 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
15095 (trunc<mode><pmov_dst_4_lower>2): this.
15097 2021-07-01 David Malcolm <dmalcolm@redhat.com>
15099 * diagnostic.h (diagnostic_context::m_file_cache): New field.
15100 * input.c (class fcache): Rename to...
15101 (class file_cache_slot): ...this, making most members private and
15102 prefixing fields with "m_".
15103 (file_cache_slot::get_file_path): New accessor.
15104 (file_cache_slot::get_use_count): New accessor.
15105 (file_cache_slot::missing_trailing_newline_p): New accessor.
15106 (file_cache_slot::inc_use_count): New.
15107 (fcache_buffer_size): Move to...
15108 (file_cache_slot::buffer_size): ...here.
15109 (fcache_line_record_size): Move to...
15110 (file_cache_slot::line_record_size): ...here.
15111 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
15112 (fcache_tab_size): Move to file_cache::num_file_slots.
15113 (diagnostic_file_cache_init): Update for move of fcache_tab
15114 to global_dc->m_file_cache.
15115 (diagnostic_file_cache_fini): Likewise.
15116 (lookup_file_in_cache_tab): Convert to...
15117 (file_cache::lookup_file): ...this.
15118 (diagnostics_file_cache_forcibly_evict_file): Update for move of
15119 fcache_tab to global_dc->m_file_cache, moving most of
15120 implementation to...
15121 (file_cache::forcibly_evict_file): ...this new function and...
15122 (file_cache_slot::evict): ...this new function.
15123 (evicted_cache_tab_entry): Convert to...
15124 (file_cache::evicted_cache_tab_entry): ...this.
15125 (add_file_to_cache_tab): Convert to...
15126 (file_cache::add_file): ...this, moving bulk of implementation
15128 (file_cache_slot::create): ..this new function.
15129 (file_cache::file_cache): New.
15130 (file_cache::~file_cache): New.
15131 (lookup_or_add_file_to_cache_tab): Convert to...
15132 (file_cache::lookup_or_add_file): ..this new function.
15133 (fcache::fcache): Rename to...
15134 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
15136 (fcache::~fcache): Rename to...
15137 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
15139 (needs_read): Convert to...
15140 (file_cache_slot::needs_read_p): ...this.
15141 (needs_grow): Convert to...
15142 (file_cache_slot::needs_grow_p): ...this.
15143 (maybe_grow): Convert to...
15144 (file_cache_slot::maybe_grow): ...this.
15145 (read_data): Convert to...
15146 (file_cache_slot::read_data): ...this.
15147 (maybe_read_data): Convert to...
15148 (file_cache_slot::maybe_read_data): ...this.
15149 (get_next_line): Convert to...
15150 (file_cache_slot::get_next_line): ...this.
15151 (goto_next_line): Convert to...
15152 (file_cache_slot::goto_next_line): ...this.
15153 (read_line_num): Convert to...
15154 (file_cache_slot::read_line_num): ...this.
15155 (location_get_source_line): Update for moving of globals to
15156 global_dc->m_file_cache.
15157 (location_missing_trailing_newline): Likewise.
15158 * input.h (class file_cache_slot): New forward decl.
15159 (class file_cache): New.
15161 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
15163 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
15164 128-bit floating point conditional move support.
15165 (have_compare_and_set_mask): Add IEEE 128-bit floating point
15167 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
15168 (mov<mode>cc_p10, IEEE128 iterator): New insn.
15169 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
15170 (fpmask<mode>, IEEE128 iterator): New insn.
15171 (xxsel<mode>, IEEE128 iterator): New insn.
15173 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
15176 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
15178 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15180 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15182 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
15184 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
15185 (vec_duplicate<mode>): New expander.
15187 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15190 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15192 (ix86_byte_broadcast): New function.
15193 (ix86_convert_const_wide_int_to_broadcast): Likewise.
15194 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
15195 size is 16 bytes or bigger.
15196 (ix86_broadcast_from_integer_constant): New function.
15197 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
15198 to broadcast if mode size is 16 bytes or bigger.
15199 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
15201 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
15203 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15205 * config/i386/predicates.md (ix86_endbr_immediate_operand):
15206 Return true/false instead of 1/0.
15207 (movq_parallel): Ditto.
15209 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15211 * recog.c (general_operand): Return true/false instead of 1/0.
15212 (register_operand): Ditto.
15213 (immediate_operand): Ditto.
15214 (const_int_operand): Ditto.
15215 (const_scalar_int_operand): Ditto.
15216 (const_double_operand): Ditto.
15217 (push_operand): Ditto.
15218 (pop_operand): Ditto.
15219 (memory_operand): Ditto.
15220 (indirect_operand): Ditto.
15222 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15224 * genpreds.c (write_predicate_subfunction):
15225 Change the type of written subfunction to bool.
15226 (write_one_predicate_function):
15227 Change the type of written function to bool.
15228 (write_tm_preds_h): Ditto.
15229 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
15230 * recog.c (general_operand): Change the type to bool.
15231 (address_operand): Ditto.
15232 (register_operand): Ditto.
15233 (pmode_register_operand): Ditto.
15234 (scratch_operand): Ditto.
15235 (immediate_operand): Ditto.
15236 (const_int_operand): Ditto.
15237 (const_scalar_int_operand): Ditto.
15238 (const_double_operand): Ditto.
15239 (nonimmediate_operand): Ditto.
15240 (nonmemory_operand): Ditto.
15241 (push_operand): Ditto.
15242 (pop_operand): Ditto.
15243 (memory_operand): Ditto.
15244 (indirect_operand): Ditto.
15245 (ordered_comparison_operator): Ditto.
15246 (comparison_operator): Ditto.
15247 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
15248 Change the type of indirect predicate function to bool.
15249 * config/rs6000/rs6000.c (easy_vector_constant):
15250 Change the type to bool.
15251 * config/mips/mips-protos.h (m16_based_address_p):
15252 Change the type of operand 3 to bool.
15254 2021-07-01 Richard Biener <rguenther@suse.de>
15256 PR tree-optimization/101280
15257 PR tree-optimization/101173
15258 * gimple-loop-interchange.cc
15259 (tree_loop_interchange::valid_data_dependences): Revert
15260 previous change and instead correctly handle DDR_REVERSED_P
15263 2021-07-01 Richard Biener <rguenther@suse.de>
15265 PR tree-optimization/101278
15266 * tree-ssa-dse.c (dse_classify_store): First check for
15267 uses, then ignore stmt for chaining purposes.
15269 2021-07-01 Richard Biener <rguenther@suse.de>
15271 PR tree-optimization/100778
15272 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
15273 vectorized ops ahead of their scalar BB.
15275 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15278 * config/i386/i386.md (*nabs<dwi>2_doubleword):
15279 New insn_and_split pattern.
15280 (*nabs<dwi>2_1): Ditto.
15281 * config/i386/i386-features.c
15282 (general_scalar_chain::compute_convert_gain):
15283 Handle (NEG (ABS (...))) RTX. Rewrite src code
15284 scanner as switch statement.
15285 (general_scalar_chain::convert_insn):
15286 Handle (NEG (ABS (...))) RTX.
15287 (general_scalar_to_vector_candidate_p):
15288 Detect (NEG (ABS (...))) RTX. Reorder case statements
15289 for (AND (NOT (...) ...)) fallthrough.
15291 2021-07-01 Richard Biener <rguenther@suse.de>
15293 PR tree-optimization/101178
15294 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
15295 (slpg::perm_in): Add.
15296 (slpg::get_perm_in): Remove.
15297 (slpg::get_perm_materialized): Add.
15298 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
15299 during permute propagation and materialization.
15301 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15304 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
15306 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15308 PR middle-end/94366
15309 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
15310 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
15311 use boolean_type_node instead of integer_type_node as NE_EXPR type.
15312 (lower_reduction_clauses): Likewise.
15314 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15316 * config/gcn/gcn.c: Include dwarf2.h.
15317 (gcn_addr_space_debug): New function.
15318 (TARGET_ADDR_SPACE_DEBUG): New hook.
15320 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15322 * common/config/gcn/gcn-common.c
15323 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
15324 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
15326 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
15327 (gcn_frame_pointer_rqd): New function.
15328 (TARGET_FRAME_POINTER_REQUIRED): New hook.
15330 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15332 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
15333 prologue register saves.
15334 (gcn_debug_unwind_info): Use UI_DWARF2.
15335 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
15336 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
15337 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
15338 (DWARF_LINK_REGISTER): New define.
15339 (FIRST_PSEUDO_REGISTER): Increment.
15340 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
15341 (CALL_USED_REGISTERS): Likewise.
15342 (REGISTER_NAMES): Likewise.
15344 2021-06-30 Richard Biener <rguenther@suse.de>
15346 PR tree-optimization/101267
15347 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
15348 API and use SLP compatible interface of vect_is_simple_use.
15349 Reject not vectorized SLP defs for callers that do not support
15351 (vect_check_store_rhs): Handle masked stores and pass down
15352 the appropriate operator index.
15353 (vectorizable_call): Adjust.
15354 (vectorizable_store): Likewise.
15355 (vectorizable_load): Likewise. Handle SLP pecularity of
15357 (vect_is_simple_use): Remove special-casing of masked stores.
15359 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15361 * common.opt (foffload): Remove help as Driver only.
15362 * gcc.c (display_help): Add -foffload.
15364 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15366 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
15367 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
15368 a candidate list; better inform no offload target is configured
15369 and fix hint extraction when passed target is not '\0' at [len].
15370 * common.opt (foffload): Add tailing '.'.
15371 (foffload-options): Likewise; fix flag name in the help string.
15373 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
15376 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
15377 (vcage_f32): Gate comparison on __FAST_MATH__.
15378 (vcageq_f32): Likewise.
15379 (vcale_f32): Likewise.
15380 (vcaleq_f32): Likewise.
15381 (vcagt_f32): Likewise.
15382 (vcagtq_f32): Likewise.
15383 (vcalt_f32): Likewise.
15384 (vcaltq_f32): Likewise.
15385 (vcage_f16): Likewise.
15386 (vcageq_f16): Likewise.
15387 (vcale_f16): Likewise.
15388 (vcaleq_f16): Likewise.
15389 (vcagt_f16): Likewise.
15390 (vcagtq_f16): Likewise.
15391 (vcalt_f16): Likewise.
15392 (vcaltq_f16): Likewise.
15394 2021-06-30 Richard Biener <rguenther@suse.de>
15396 PR tree-optimization/101264
15397 * tree-vect-slp.c (vect_optimize_slp): Propagate the
15398 computed perm_in to all "any" permute successors
15399 we cannot de-duplicate immediately.
15401 2021-06-30 liuhongt <hongtao.liu@intel.com>
15404 * config/i386/sse.md
15405 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
15407 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
15409 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
15410 * config/i386/subst.md (maskz_scalar): New define_subst.
15411 (maskz_scalar_name): New subst_attr.
15412 (maskz_scalar_op5): Ditto.
15413 (round_saeonly_maskz_scalar_op5): Ditto.
15414 (round_saeonly_maskz_scalar_operand5): Ditto.
15416 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
15418 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
15419 Increase code CSECT alignment to at least 32 bytes.
15420 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
15421 alignment designation.
15423 2021-06-29 Sergei Trofimovich <siarheit@google.com>
15425 * doc/generic.texi: Fix s/net yet/not yet/ typo.
15427 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15429 PR tree-optimization/101254
15430 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
15431 wrapping/non-wrapping when setting the result range.
15433 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15435 * value-query.cc (gimple_range_global): Allow phis.
15437 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15439 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
15440 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
15441 statement for location context.
15442 (check_for_binary_op_overflow): Ditto.
15443 (simplify_using_ranges::get_vr_for_comparison): Ditto.
15444 (simplify_using_ranges::compare_name_with_value): Ditto.
15445 (simplify_using_ranges::compare_names): Ditto.
15446 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
15447 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
15448 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
15449 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
15450 (simplify_using_ranges::two_valued_val_range_p): Ditto.
15451 (simplify_using_ranges::simplify): Ditto.
15452 * vr-values.h: Adjust prototypes.
15454 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
15457 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
15459 2021-06-29 Julian Brown <julian@codesourcery.com>
15461 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
15462 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
15463 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
15465 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
15466 (MAX_FIXED_MODE_SIZE): Change to 128.
15468 2021-06-29 Julian Brown <julian@codesourcery.com>
15470 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
15471 (s_mnemonic): Add clrsb.
15472 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
15473 (clrsb<mode>2): Add expander for SImode/DImode.
15475 2021-06-29 Julian Brown <julian@codesourcery.com>
15477 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
15478 muldi3): Add patterns.
15480 2021-06-29 Julian Brown <julian@codesourcery.com>
15482 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
15483 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
15485 2021-06-29 Julian Brown <julian@codesourcery.com>
15487 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
15489 2021-06-29 Joseph Myers <joseph@codesourcery.com>
15491 * btfout.c, ctfout.c: Include "memmodel.h".
15493 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15495 * gcc.c (check_offload_target_name): Cast len argument to
15496 %q.*s to 'int'; avoid -Wstringop-truncation warning.
15498 2021-06-29 Richard Biener <rguenther@suse.de>
15500 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
15501 to "any" permute nodes and relax "any" permute proapgation
15502 during iterative backward propagation.
15504 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15507 * common.opt (-foffload=): Update description.
15508 (-foffload-options=): New.
15509 * doc/invoke.texi (C Language Options): Document
15510 -foffload and -foffload-options.
15511 * gcc.c (check_offload_target_name): New, split off from
15512 handle_foffload_option.
15513 (check_foffload_target_names): New.
15514 (handle_foffload_option): Handle -foffload=default.
15515 (driver_handle_option): Update for -foffload-options.
15516 * lto-opts.c (lto_write_options): Use -foffload-options
15517 instead of -foffload.
15518 * lto-wrapper.c (merge_and_complain, append_offload_options):
15520 * opts.c (common_handle_option): Likewise.
15522 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15524 * doc/invoke.texi (C Language Options): Sort options
15525 alphabetically in optlist and also the description itself.
15526 Remove leftover -fallow-single-precision from and add missing
15527 -fgnu-tm to the optlist.
15529 2021-06-29 Richard Biener <rguenther@suse.de>
15531 * tree-vect-slp.c (slpg_vertex::visited): Remove.
15532 (vect_slp_perms_eq): Handle -1 permutes.
15533 (vect_optimize_slp): Rewrite permute propagation.
15535 2021-06-29 Jakub Jelinek <jakub@redhat.com>
15538 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
15539 perform the optimization in GENERIC when sanitizing and x has a
15542 2021-06-29 Richard Biener <rguenther@suse.de>
15544 PR tree-optimization/101242
15545 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
15546 PHIs with not represented initial values as leafs.
15548 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
15550 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
15551 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
15552 "register" keyword.
15553 (pdp11_initial_elimination_offset) Remove unused variable.
15554 (pdp11_cmp_length) Ditto.
15555 (pdp11_insn_cost): Ditto, and fix signedness warning.
15557 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
15559 * btfout.c: Include tm_p.h.
15562 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15564 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
15566 (bpf_expand_epilogue): Likewise.
15567 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
15568 Do not define DBX_DEBUGGING_INFO.
15570 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15572 * doc/invoke.texi: Document the CTF and BTF debug info options.
15574 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15575 David Faust <david.faust@oracle.com>
15576 Jose E. Marchesi <jose.marchesi@oracle.com>
15577 Weimin Pan <weimin.pan@oracle.com>
15579 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
15580 GTFILES. Add new object files.
15581 * common.opt: Add CTF and BTF debug info options.
15582 * btfout.c: New file.
15583 * ctfc.c: Likewise.
15584 * ctfc.h: Likewise.
15585 * ctfout.c: Likewise.
15586 * dwarf2ctf.c: Likewise.
15587 * dwarf2ctf.h: Likewise.
15588 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
15590 * dwarf2out.c (dwarf2out_source_line): Likewise.
15591 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
15593 (debug_format_do_cu): New function.
15594 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
15596 Include dwarf2ctf.c.
15597 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
15599 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
15600 (CTF_DEBUG): New bitmask.
15601 (BTF_DEBUG): Likewise.
15602 (enum ctf_debug_info_levels): New enum.
15603 * gengtype.c (open_base_files): Handle ctfc.h.
15604 (main): Handle uint32_t type.
15605 * flags.h (btf_debuginfo_p): New definition.
15606 (dwarf_based_debuginfo_p): Likewise.
15607 * opts.c (debug_type_names): Add entries for CTF and BTF.
15608 (btf_debuginfo_p): New function.
15609 (dwarf_based_debuginfo_p): Likewise.
15610 (common_handle_option): Handle -gctfN and -gbtf options.
15611 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
15612 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
15615 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
15617 * dwarf2out.c (AT_class): Function is no longer static.
15618 (AT_int): Likewise.
15619 (AT_unsigned): Likewise.
15620 (AT_loc): Likewise.
15621 (get_AT): Likewise.
15622 (get_AT_string): Likewise.
15623 (get_AT_flag): Likewise.
15624 (get_AT_unsigned): Likewise.
15625 (get_AT_ref): Likewise.
15626 (new_die_raw): Likewise.
15627 (lookup_decl_die): Likewise.
15628 (base_type_die): Likewise.
15629 (add_name_attribute): Likewise.
15630 (add_AT_int): Likewise.
15631 (add_AT_unsigned): Likewise.
15632 (add_AT_loc): Likewise.
15633 (dw_get_die_tag): New function.
15634 (dw_get_die_child): Likewise.
15635 (dw_get_die_sib): Likewise.
15636 (struct dwarf_file_data): Move from here to dwarf2out.h
15637 (struct dw_attr_struct): Likewise.
15638 * dwarf2out.h: Analogous changes.
15640 2021-06-28 Martin Jambor <mjambor@suse.cz>
15643 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
15644 members m_dead_stmts and m_dead_ssas.
15645 * ipa-param-manipulation.c
15646 (ipa_param_body_adjustments::mark_dead_statements): New function.
15647 (ipa_param_body_adjustments::common_initialization): Call it on
15648 all removed but not split parameters.
15649 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
15651 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
15653 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
15654 dead debug statements.
15655 (copy_phis_for_bb): Do not copy dead PHI nodes.
15657 2021-06-28 Martin Jambor <mjambor@suse.cz>
15660 * symtab-clones.h (clone_info): Removed member param_adjustments.
15661 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
15662 deal with pass-through splits now.
15663 (ipa_param_performed_split): Removed.
15664 (ipa_param_adjustments::modify_call): Adjusted parameters.
15665 (class ipa_param_body_adjustments): Adjusted parameters of
15666 register_replacement, modify_gimple_stmt and modify_call_stmt.
15667 (ipa_verify_edge_has_no_modifications): Declare.
15668 (ipa_edge_modifications_finalize): Declare.
15669 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
15670 performed_splits processing, pas only edge to padjs->modify_call,
15671 check that call arguments were not modified if they should not have
15673 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
15675 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
15676 (ipa_edge_modification_info): Likewise.
15677 (ipa_edge_modification_sum): Likewise.
15678 (ipa_edge_modifications): New edge summary.
15679 (ipa_verify_edge_has_no_modifications): New function.
15680 (transitive_split_p): Removed.
15681 (transitive_split_map): Likewise.
15682 (init_transitive_splits): Likewise.
15683 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
15684 summary instead of performed_splits.
15685 (ipa_param_body_adjustments::register_replacement): Drop dummy
15686 parameter, set base_index of the created ipa_param_body_replacement.
15687 (phi_arg_will_live_p): New function.
15688 (ipa_param_body_adjustments::common_initialization): Do not create
15689 IPA_SRA dummy decls.
15690 (simple_tree_swap_info): Removed.
15691 (remap_split_decl_to_dummy): Likewise.
15692 (record_argument_state_1): New function.
15693 (record_argument_state): Likewise.
15694 (ipa_param_body_adjustments::modify_call_stmt): New parameter
15695 orig_stmt. Do not work with dummy decls, save necessary info about
15696 changes to ipa_edge_modifications.
15697 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
15698 orig_stmt, pass it to modify_call_stmt.
15699 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
15700 modify_gimple_stmt.
15701 (ipa_edge_modifications_finalize): New function.
15702 * tree-inline.c (remap_gimple_stmt): Pass original statement to
15703 modify_gimple_stmt.
15704 (copy_phis_for_bb): Do not copy dead PHI nodes.
15705 (expand_call_inline): Do not remap performed_splits.
15706 (update_clone_info): Likewise.
15707 * toplev.c: Include ipa-param-manipulation.h.
15708 (toplev::finalize): Call ipa_edge_modifications_finalize.
15710 2021-06-28 Andrew Pinski <apinski@marvell.com>
15712 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
15713 info if we're the only things setting the target PHI.
15714 (value_replacement): Don't duplicate range here.
15715 (minmax_replacement): Likewise.
15717 2021-06-28 Richard Biener <rguenther@suse.de>
15719 PR tree-optimization/101229
15720 * gimple-walk.c (gimple_walk_op): Handle PHIs.
15722 2021-06-28 Martin Liska <mliska@suse.cz>
15724 * config/v850/v850.c (construct_dispose_instruction): Allocate
15726 (construct_prepare_instruction): Likewise.
15728 2021-06-28 Martin Liska <mliska@suse.cz>
15730 * config/v850/v850.c (v850_option_override): Build default
15732 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
15733 ignored for inlining.
15734 (TARGET_CAN_INLINE_P): New.
15736 2021-06-28 Richard Biener <rguenther@suse.de>
15738 PR tree-optimization/101207
15739 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
15740 permute eliding for load permutations properly.
15742 2021-06-28 Richard Biener <rguenther@suse.de>
15744 PR tree-optimization/101173
15745 * gimple-loop-interchange.cc
15746 (tree_loop_interchange::valid_data_dependences): Disallow outer
15747 loop dependence distance of zero.
15749 2021-06-28 liuhongt <hongtao.liu@intel.com>
15752 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
15753 define_insn_and_split.
15754 (*avx_cmp<mode>3_ltint): Ditto.
15755 (*avx2_pcmp<mode>3_3): Ditto.
15756 (*avx2_pcmp<mode>3_4): Ditto.
15757 (*avx2_pcmp<mode>3_5): Ditto.
15759 2021-06-28 liuhongt <hongtao.liu@intel.com>
15761 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
15762 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
15763 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
15764 IX86_BUILTIN_PBLENDVB128): Replace icode with
15766 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
15768 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
15769 New pre_reload splitter.
15771 2021-06-27 Andrew Pinski <apinski@marvell.com>
15773 PR middle-end/101230
15774 * fold-const.c (fold_ternary_loc): Check
15775 the return value of invert_tree_comparison.
15777 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
15779 * config.gcc: Add SPDX License Identifier.
15780 (powerpc-ibm-aix789): Default to aix73.h.
15781 (powerpc-ibm-aix7.2.*.*): New stanza.
15782 * config/rs6000/aix72.h: Add SPDX License Identifier.
15783 * config/rs6000/aix73.h: New file.
15785 2021-06-26 Jason Merrill <jason@redhat.com>
15787 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
15789 2021-06-26 Andrew Pinski <apinski@marvell.com>
15791 * genmatch.c (lower_cond): Copy for_subst_vec
15792 for the simplify also.
15793 (lower): Swap the order for lower_for and lower_cond.
15795 2021-06-26 Andrew Pinski <apinski@marvell.com>
15797 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
15798 flow senatitive info on the moved ssa set.
15800 2021-06-26 Andrew Pinski <apinski@marvell.com>
15802 * fold-const.c (fold_cond_expr_with_comparison):
15803 Exand arg0 into comp_code, arg00, and arg01.
15804 (fold_ternary_loc): Use invert_tree_comparison
15805 instead of fold_invert_truthvalue for the case
15806 where we have A CMP B ? C : A.
15808 2021-06-25 Martin Sebor <msebor@redhat.com>
15810 PR middle-end/101216
15811 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
15813 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
15815 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
15817 2021-06-25 Richard Biener <rguenther@suse.de>
15819 PR tree-optimization/101202
15820 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
15823 2021-06-25 Richard Biener <rguenther@suse.de>
15825 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
15826 STMT_VINFO_REDUC_DEF from the original representative.
15828 2021-06-25 Martin Sebor <msebor@redhat.com>
15830 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
15831 gimple_no_warning_p and gimple_set_no_warning with
15832 warning_suppressed_p, and suppress_warning.
15834 (maybe_warn_for_bound): Same.
15835 (warn_for_access): Same.
15836 (check_access): Same.
15837 (expand_builtin_strncmp): Same.
15838 (fold_builtin_varargs): Same.
15839 * calls.c (maybe_warn_nonstring_arg): Same.
15840 (maybe_warn_rdwr_sizes): Same.
15841 * cfgexpand.c (expand_call_stmt): Same.
15842 * cgraphunit.c (check_global_declaration): Same.
15843 * fold-const.c (fold_undefer_overflow_warnings): Same.
15844 (fold_truth_not_expr): Same.
15845 (fold_unary_loc): Same.
15846 (fold_checksum_tree): Same.
15847 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
15848 (array_bounds_checker::check_mem_ref): Same.
15849 (array_bounds_checker::check_addr_expr): Same.
15850 (array_bounds_checker::check_array_bounds): Same.
15851 * gimple-expr.c (copy_var_decl): Same.
15852 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
15853 (gimple_fold_builtin_strncat): Same.
15854 (gimple_fold_builtin_stxcpy_chk): Same.
15855 (gimple_fold_builtin_stpcpy): Same.
15856 (gimple_fold_builtin_sprintf): Same.
15857 (fold_stmt_1): Same.
15858 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
15859 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
15860 * gimple-ssa-sprintf.c (handle_printf_call): Same.
15861 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
15862 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
15863 * gimple-ssa-warn-restrict.h: Adjust declarations.
15864 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
15865 gimple_no_warning_p and gimple_set_no_warning with
15866 warning_suppressed_p, and suppress_warning.
15867 (check_call): Same.
15868 (check_bounds_or_overlap): Same.
15869 * gimple.c (gimple_build_call_from_tree): Same.
15870 * gimplify.c (gimplify_return_expr): Same.
15871 (gimplify_cond_expr): Same.
15872 (gimplify_modify_expr_complex_part): Same.
15873 (gimplify_modify_expr): Same.
15874 (gimple_push_cleanup): Same.
15875 (gimplify_expr): Same.
15876 * omp-expand.c (expand_omp_for_generic): Same.
15877 (expand_omp_taskloop_for_outer): Same.
15878 * omp-low.c (lower_rec_input_clauses): Same.
15879 (lower_lastprivate_clauses): Same.
15880 (lower_send_clauses): Same.
15881 (lower_omp_target): Same.
15882 * tree-cfg.c (pass_warn_function_return::execute): Same.
15883 * tree-complex.c (create_one_component_var): Same.
15884 * tree-inline.c (remap_gimple_op_r): Same.
15885 (copy_tree_body_r): Same.
15886 (declare_return_variable): Same.
15887 (expand_call_inline): Same.
15888 * tree-nested.c (lookup_field_for_decl): Same.
15889 * tree-sra.c (create_access_replacement): Same.
15890 (generate_subtree_copies): Same.
15891 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
15892 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
15893 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
15894 * tree-ssa-loop-im.c (execute_sm): Same.
15895 * tree-ssa-phiopt.c (cond_store_replacement): Same.
15896 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
15897 (handle_builtin_strcpy): Same.
15898 (maybe_diag_stxncpy_trunc): Same.
15899 (handle_builtin_stxncpy_strncat): Same.
15900 (handle_builtin_strcat): Same.
15901 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
15902 (set_no_uninit_warning): Same.
15903 (uninit_undefined_value_p): Same.
15904 (warn_uninit): Same.
15905 (maybe_warn_operand): Same.
15906 * tree-vrp.c (compare_values_warnv): Same.
15907 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
15908 (test_for_singularity): Same.
15909 * gimple.h (warning_suppressed_p): New function.
15910 (suppress_warning): Same.
15911 (copy_no_warning): Same.
15912 (gimple_set_block): Call gimple_set_location.
15913 (gimple_set_location): Call copy_warning.
15915 2021-06-25 Martin Sebor <msebor@redhat.com>
15917 * tree.h (warning_suppressed_at, copy_warning,
15918 warning_suppressed_p, suppress_warning): New functions.
15920 2021-06-25 Martin Sebor <msebor@redhat.com>
15922 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
15923 * gengtype.c (open_base_files): Add diagnostic-spec.h.
15924 * diagnostic-spec.c: New file.
15925 * diagnostic-spec.h: New file.
15926 * tree.h (no_warning, all_warnings, suppress_warning_at): New
15928 * warning-control.cc: New file.
15930 2021-06-25 liuhongt <hongtao.liu@intel.com>
15933 * config/i386/i386.c (x86_order_regs_for_local_alloc):
15936 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15938 PR tree-optimization/101189
15939 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
15940 LHS range of condition to postfold routine.
15941 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
15942 FALSE edge if the LHS range supports it being taken.
15943 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
15945 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15947 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
15948 (relation_oracle::find_relation_block): Check correct bitmap.
15949 (relation_oracle::dump): Do not dump NULL blocks.
15951 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15953 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
15954 range_on_edge instead of manually calculating.
15956 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15958 * range-op.cc: Fix comment.
15960 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
15963 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
15964 Handle V8QI and V4HI modes.
15965 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
15967 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
15968 (mmxpackmode): New mode attribute.
15969 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
15970 (mmxunpackmode): New mode attribute.
15971 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
15972 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
15973 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
15974 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
15975 * config/i386/i386.md (extsuffix): Move from ...
15976 * config/i386/sse.md: ... here.
15978 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
15980 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
15981 (dwarf2out_finish): ...instead of here.
15983 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
15985 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
15986 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
15987 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
15988 * configure: Regenerate.
15990 2021-06-24 Richard Biener <rguenther@suse.de>
15992 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
15993 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
15994 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
15996 2021-06-24 Richard Biener <rguenther@suse.de>
15998 * config/i386/sse.md (avx_addsubv4df3): Rename to
16000 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
16001 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
16002 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
16003 * config/i386/i386-builtin.def: Adjust.
16004 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
16005 * optabs.def (vec_addsub_optab): New optab.
16006 * tree-vect-slp-patterns.c (class addsub_pattern): New.
16007 (slp_patterns): Add addsub_pattern.
16008 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
16009 across CFN_VEC_ADDSUB.
16010 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
16012 * doc/md.texi (vec_addsub<mode>3): Document.
16014 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16016 PR middle-end/101170
16017 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
16018 where regno + subreg_regno_offset wraps around use 0 as starting
16021 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16023 PR middle-end/101172
16024 * stor-layout.c (finish_bitfield_representative): If nextf has
16025 error_mark_node type, set repr type to error_mark_node too.
16027 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
16029 * config/s390/s390.c (s390_function_profiler): Ignore labelno
16031 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
16033 2021-06-24 Richard Biener <rguenther@suse.de>
16035 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
16036 across operations that have different semantics on different
16039 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16041 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
16042 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
16043 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
16044 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
16045 OMP_TARGET user outer_ctx instead of ctx for placeholders and
16046 initializer/combiner gimplification.
16047 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
16048 on target constructs.
16049 (lower_rec_input_clauses): Likewise.
16050 (lower_omp_target): Likewise.
16051 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
16052 on target if in_reduction is present.
16054 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
16056 * tree-predcom.c (class pcom_worker): New class.
16057 (release_chain): Renamed to...
16058 (pcom_worker::release_chain): ...this.
16059 (release_chains): Renamed to...
16060 (pcom_worker::release_chains): ...this.
16061 (aff_combination_dr_offset): Renamed to...
16062 (pcom_worker::aff_combination_dr_offset): ...this.
16063 (determine_offset): Renamed to...
16064 (pcom_worker::determine_offset): ...this.
16065 (class comp_ptrs): New class.
16066 (split_data_refs_to_components): Renamed to...
16067 (pcom_worker::split_data_refs_to_components): ...this,
16068 and update with class comp_ptrs.
16069 (suitable_component_p): Renamed to...
16070 (pcom_worker::suitable_component_p): ...this.
16071 (filter_suitable_components): Renamed to...
16072 (pcom_worker::filter_suitable_components): ...this.
16073 (valid_initializer_p): Renamed to...
16074 (pcom_worker::valid_initializer_p): ...this.
16075 (find_looparound_phi): Renamed to...
16076 (pcom_worker::find_looparound_phi): ...this.
16077 (add_looparound_copies): Renamed to...
16078 (pcom_worker::add_looparound_copies): ...this.
16079 (determine_roots_comp): Renamed to...
16080 (pcom_worker::determine_roots_comp): ...this.
16081 (determine_roots): Renamed to...
16082 (pcom_worker::determine_roots): ...this.
16083 (single_nonlooparound_use): Renamed to...
16084 (pcom_worker::single_nonlooparound_use): ...this.
16085 (remove_stmt): Renamed to...
16086 (pcom_worker::remove_stmt): ...this.
16087 (execute_pred_commoning_chain): Renamed to...
16088 (pcom_worker::execute_pred_commoning_chain): ...this.
16089 (execute_pred_commoning): Renamed to...
16090 (pcom_worker::execute_pred_commoning): ...this.
16091 (struct epcc_data): New member worker.
16092 (execute_pred_commoning_cbck): Call execute_pred_commoning
16093 with pcom_worker pointer.
16094 (find_use_stmt): Renamed to...
16095 (pcom_worker::find_use_stmt): ...this.
16096 (find_associative_operation_root): Renamed to...
16097 (pcom_worker::find_associative_operation_root): ...this.
16098 (find_common_use_stmt): Renamed to...
16099 (pcom_worker::find_common_use_stmt): ...this.
16100 (combinable_refs_p): Renamed to...
16101 (pcom_worker::combinable_refs_p): ...this.
16102 (reassociate_to_the_same_stmt): Renamed to...
16103 (pcom_worker::reassociate_to_the_same_stmt): ...this.
16104 (stmt_combining_refs): Renamed to...
16105 (pcom_worker::stmt_combining_refs): ...this.
16106 (combine_chains): Renamed to...
16107 (pcom_worker::combine_chains): ...this.
16108 (try_combine_chains): Renamed to...
16109 (pcom_worker::try_combine_chains): ...this.
16110 (prepare_initializers_chain): Renamed to...
16111 (pcom_worker::prepare_initializers_chain): ...this.
16112 (prepare_initializers): Renamed to...
16113 (pcom_worker::prepare_initializers): ...this.
16114 (prepare_finalizers_chain): Renamed to...
16115 (pcom_worker::prepare_finalizers_chain): ...this.
16116 (prepare_finalizers): Renamed to...
16117 (pcom_worker::prepare_finalizers): ...this.
16118 (tree_predictive_commoning_loop): Renamed to...
16119 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
16120 some calls and remove some cleanup code.
16121 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
16122 (static variable looparound_phis): Remove.
16123 (static variable name_expansions): Remove.
16125 2021-06-24 Richard Biener <rguenther@suse.de>
16127 * tree-vect-slp.c (slpg_vertex): New struct.
16128 (vect_slp_build_vertices): Adjust.
16129 (vect_optimize_slp): Likewise. Maintain an outgoing permute
16130 and a materialized one.
16132 2021-06-24 Richard Biener <rguenther@suse.de>
16134 PR tree-optimization/101105
16135 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
16136 Only ignore steps when they are equal or scalar order is preserved.
16138 2021-06-24 liuhongt <hongtao.liu@intel.com>
16141 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
16142 Adjust comments for ix86_expand_vecop_qihi2.
16143 (ix86_expand_vecmul_qihi): Renamed to ..
16144 (ix86_expand_vecop_qihi2): Adjust function prototype to
16145 support shift operation, add static to definition.
16146 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
16147 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
16148 ix86_expand_vec_shift_qihi_constant.
16149 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
16150 (ix86_expand_vec_shift_qihi_constant): Deleted.
16151 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
16153 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
16154 condition TARGET_64BIT.
16155 (mul<mode>3): Ditto.
16156 (<insn><mode>3): Ditto.
16157 (vlshr<mode>3): Extend to support avx512 vlshr.
16158 (v<insn><mode>3): New expander for
16160 (v<insn>v8qi3): Ditto.
16161 (vashrv8hi3<mask_name>): Renamed to ..
16162 (vashr<mode>3): And extend to support V16QImode for avx512.
16163 (vashrv16qi3): Deleted.
16164 (vashrv2di3<mask_name>): Extend expander to support avx512
16167 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
16169 * doc/lto.texi (Design Overview): Update that slim objects are
16172 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
16174 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
16175 of OTHER_POWER10_MASKS so it will not be enabled by default.
16177 2021-06-23 Richard Biener <rguenther@suse.de>
16178 Martin Jambor <mjambor@suse.cz>
16180 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
16181 param replacement unconditionally. Adjust comment.
16183 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16185 * Makefile.in (OBJS): Add gimple-range-fold.o
16186 * gimple-range-fold.cc: New.
16187 * gimple-range-fold.h: New.
16188 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
16189 (gimple_range_calc_op2): Ditto.
16190 * gimple-range-gori.h: Move prototypes to here.
16191 * gimple-range.cc: Adjust include files.
16192 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
16193 (fur_source::get_operand): Ditto.
16194 (fur_source::get_phi_operand): Ditto.
16195 (fur_source::query_relation): Ditto.
16196 (fur_source::register_relation): Ditto.
16197 (class fur_edge): Ditto.
16198 (fur_edge::fur_edge): Ditto.
16199 (fur_edge::get_operand): Ditto.
16200 (fur_edge::get_phi_operand): Ditto.
16201 (fur_stmt::fur_stmt): Ditto.
16202 (fur_stmt::get_operand): Ditto.
16203 (fur_stmt::get_phi_operand): Ditto.
16204 (fur_stmt::query_relation): Ditto.
16205 (class fur_depend): Relocate to gimple-range-fold.h.
16206 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
16207 (fur_depend::register_relation): Ditto.
16208 (fur_depend::register_relation): Ditto.
16209 (class fur_list): Ditto.
16210 (fur_list::fur_list): Ditto.
16211 (fur_list::get_operand): Ditto.
16212 (fur_list::get_phi_operand): Ditto.
16213 (fold_range): Ditto.
16214 (adjust_pointer_diff_expr): Ditto.
16215 (gimple_range_adjustment): Ditto.
16216 (gimple_range_base_of_assignment): Ditto.
16217 (gimple_range_operand1): Ditto.
16218 (gimple_range_operand2): Ditto.
16219 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
16220 (gimple_range_calc_op2): Ditto.
16221 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
16222 (fold_using_range::range_of_range_op): Ditto.
16223 (fold_using_range::range_of_address): Ditto.
16224 (fold_using_range::range_of_phi): Ditto.
16225 (fold_using_range::range_of_call): Ditto.
16226 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
16227 (fold_using_range::range_of_builtin_call): Ditto.
16228 (fold_using_range::range_of_cond_expr): Ditto.
16229 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
16230 (fold_using_range::relation_fold_and_or): Ditto.
16231 (fold_using_range::postfold_gcond_edges): Ditto.
16232 * gimple-range.h: Add gimple-range-fold.h to include files. Change
16233 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
16234 (gimple_range_handler): Relocate to gimple-range-fold.h.
16235 (gimple_range_ssa_p): Ditto.
16236 (range_compatible_p): Ditto.
16237 (class fur_source): Ditto.
16238 (class fur_stmt): Ditto.
16239 (class fold_using_range): Ditto.
16240 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
16241 (gimple_range_calc_op2): Ditto.
16243 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16245 PR tree-optimization/101148
16246 PR tree-optimization/101014
16247 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
16248 (ranger_cache::~ranger_cache): Adjust.
16249 (ranger_cache::block_range): Check if propagation disallowed.
16250 (ranger_cache::propagate_cache): Disallow propagation if new value
16251 can't be stored properly.
16252 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
16254 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16256 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
16257 (sbr_vector::set_bb_range): Return true.
16258 (class sbr_sparse_bitmap): Adjust.
16259 (sbr_sparse_bitmap::set_bb_range): Return value.
16260 (block_range_cache::set_bb_range): Return value.
16261 (ranger_cache::propagate_cache): Use return value to print msg.
16262 * gimple-range-cache.h (class block_range_cache): Adjust.
16264 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16266 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
16268 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
16270 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
16271 so this pattern can be used for test/compare removal. Pass
16272 current insn to compute_logical_op_length and output_logical_op.
16273 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
16274 (h8300_and_costs): Add argument to compute_logical_op_length.
16275 (output_logical_op): Add new argument. Use it to determine if the
16276 condition codes are used and adjust the output accordingly.
16277 (compute_logical_op_length): Add new argument and update length
16278 computations when condition codes are used.
16279 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
16281 (output_logical_op): Likewise.
16283 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16286 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
16287 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
16288 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
16289 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
16290 * config/i386/sse.md (unspec): ... here.
16292 2021-06-23 Martin Liska <mliska@suse.cz>
16295 * optc-save-gen.awk: Put back arm_fp16_format to
16298 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16301 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
16304 (clz<mode>2): Update RTX pattern for additions.
16306 2021-06-23 Jakub Jelinek <jakub@redhat.com>
16308 PR middle-end/101167
16309 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
16310 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
16312 2021-06-22 Sergei Trofimovich <siarheit@google.com>
16314 * doc/rtl.texi: drop unbalanced parenthesis.
16316 2021-06-22 Richard Biener <rguenther@suse.de>
16318 PR middle-end/101156
16319 * gimplify.c (gimplify_expr): Remove premature incorrect
16322 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16324 PR tree-optimization/101159
16325 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
16328 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16330 PR middle-end/101160
16331 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
16332 clear crtl->return_rtx instead of keeping it referencing a pseudo.
16334 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16335 Andrew Pinski <apinski@marvell.com>
16337 PR tree-optimization/101162
16338 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
16341 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16343 * range-op.cc (range_relational_tests): New.
16344 (range_op_tests): Call range_relational_tests.
16346 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16348 * range-op.cc (operator_cast::lhs_op1_relation): New.
16349 (operator_identity::lhs_op1_relation): Mew.
16351 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16353 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
16355 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16357 * range-op.cc (operator_plus::lhs_op1_relation): New.
16358 (operator_plus::lhs_op2_relation): New.
16360 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16362 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
16363 relation_oracle if dominators exist.
16364 (ranger_cache::~ranger_cache): Dispose of oracle.
16365 (ranger_cache::dump_bb): Dump oracle.
16366 * gimple-range.cc (fur_source::fur_source): New.
16367 (fur_source::get_operand): Use mmeber query.
16368 (fur_source::get_phi_operand): Use member_query.
16369 (fur_source::query_relation): New.
16370 (fur_source::register_dependency): Delete.
16371 (fur_source::register_relation): New.
16372 (fur_edge::fur_edge): Adjust.
16373 (fur_edge::get_phi_operand): Fix comment.
16374 (fur_edge::query): Delete.
16375 (fur_stmt::fur_stmt): Adjust.
16376 (fur_stmt::query): Delete.
16377 (fur_depend::fur_depend): Adjust.
16378 (fur_depend::register_relation): New.
16379 (fur_depend::register_relation): New.
16380 (fur_list::fur_list): Adjust.
16381 (fur_list::get_operand): Use member query.
16382 (fold_using_range::range_of_range_op): Process and query relations.
16383 (fold_using_range::range_of_address): Adjust dependency call.
16384 (fold_using_range::range_of_phi): Ditto.
16385 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
16386 (fold_using_range::relation_fold_and_or): New.
16387 (fold_using_range::postfold_gcond_edges): New.
16388 * gimple-range.h (class gimple_ranger): Adjust.
16389 (class fur_source): Adjust members.
16390 (class fur_stmt): Ditto.
16391 (class fold_using_range): Ditto.
16393 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16395 * range-op.cc (range_operator::wi_fold): Apply relation effect.
16396 (range_operator::fold_range): Adjust and apply relation effect.
16397 (*::fold_range): Add relation parameters.
16398 (*::op1_range): Ditto.
16399 (*::op2_range): Ditto.
16400 (range_operator::lhs_op1_relation): New.
16401 (range_operator::lhs_op2_relation): New.
16402 (range_operator::op1_op2_relation): New.
16403 (range_operator::op1_op2_relation_effect): New.
16404 (relop_early_resolve): New.
16405 (operator_equal::op1_op2_relation): New.
16406 (operator_equal::fold_range): Call relop_early_resolve.
16407 (operator_not_equal::op1_op2_relation): New.
16408 (operator_not_equal::fold_range): Call relop_early_resolve.
16409 (operator_lt::op1_op2_relation): New.
16410 (operator_lt::fold_range): Call relop_early_resolve.
16411 (operator_le::op1_op2_relation): New.
16412 (operator_le::fold_range): Call relop_early_resolve.
16413 (operator_gt::op1_op2_relation): New.
16414 (operator_gt::fold_range): Call relop_early_resolve.
16415 (operator_ge::op1_op2_relation): New.
16416 (operator_ge::fold_range): Call relop_early_resolve.
16417 * range-op.h (class range_operator): Adjust parameters and methods.
16419 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16421 * Makefile.in (OBJS): Add value-relation.o.
16422 * gimple-range.h: Adjust include files.
16423 * tree-data-ref.c: Adjust include file order.
16424 * value-query.cc (range_query::get_value_range): Default to no oracle.
16425 (range_query::query_relation): New.
16426 (range_query::query_relation): New.
16427 * value-query.h (class range_query): Adjust.
16428 * value-relation.cc: New.
16429 * value-relation.h: New.
16431 2021-06-22 Richard Biener <rguenther@suse.de>
16433 PR tree-optimization/101151
16434 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
16437 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
16439 * config/riscv/riscv.c (thead_c906_tune_info): New.
16440 (riscv_tune_info_table): Use new tune.
16442 2021-06-22 Richard Biener <rguenther@suse.de>
16444 PR tree-optimization/101158
16445 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
16446 checking after checking for matching operation.
16448 2021-06-22 Richard Biener <rguenther@suse.de>
16450 PR tree-optimization/101159
16451 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
16452 missing NULL vectype check.
16454 2021-06-22 Richard Biener <rguenther@suse.de>
16456 PR tree-optimization/101154
16457 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
16459 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16462 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
16463 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
16464 * config/i386/i386.c (ix86_expand_prologue): Clear it.
16465 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
16466 Remove "" from match_operand. Emit new insns using emit_move_insn and
16467 set ix86_last_zero_store_uid to INSN_UID of the last store.
16468 Add peephole2s for 1/2/4 stores of const0_rtx following previous
16471 2021-06-22 Martin Liska <mliska@suse.cz>
16473 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
16476 2021-06-22 Martin Liska <mliska@suse.cz>
16478 * gcov-io.h: Remove padding entries.
16480 2021-06-22 liuhongt <hongtao.liu@intel.com>
16482 PR tree-optimization/97770
16483 * tree-vect-patterns.c (vect_recog_popcount_pattern):
16485 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
16487 2021-06-22 liuhongt <hongtao.liu@intel.com>
16490 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
16491 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
16492 (expand<mode>_mask): this ..
16493 (*expand<mode>_mask): New pre_reload splitter to transform
16494 v{,p}expand* to vmov* when mask is zero, all ones, or has all
16495 ones in it's lower part, otherwise still generate
16498 2021-06-22 liuhongt <hongtao.liu@intel.com>
16501 * config/i386/i386-expand.c
16502 (ix86_expand_special_args_builtin): Keep constm1_operand only
16503 if it satisfies insn's operand predicate.
16505 2021-06-21 Jason Merrill <jason@redhat.com>
16508 * df-scan.c (df_ref_record): Check that regno < endregno.
16509 * function.c (assign_parms, expand_function_end): Do nothing with a
16510 TYPE_EMPTY_P result.
16512 2021-06-21 Richard Biener <rguenther@suse.de>
16514 PR tree-optimization/101120
16515 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
16517 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
16518 DR chain DCE capability.
16519 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
16520 * tree-vect-stmts.c (vectorizable_load): Remove unused
16521 loads in the DR chain for SLP.
16523 2021-06-21 Jakub Jelinek <jakub@redhat.com>
16525 PR inline-asm/100785
16526 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
16527 output or input operands were already error_mark_node.
16528 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
16529 remove all inputs, outputs and clobbers from the asm and
16530 set template to "".
16532 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16534 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
16535 (vceq_s16): Likewise.
16536 (vceq_s32): Likewise.
16537 (vceq_u8): Likewise.
16538 (vceq_u16): Likewise.
16539 (vceq_u32): Likewise.
16540 (vceq_p8): Likewise.
16541 (vceqq_s8): Likewise.
16542 (vceqq_s16): Likewise.
16543 (vceqq_s32): Likewise.
16544 (vceqq_u8): Likewise.
16545 (vceqq_u16): Likewise.
16546 (vceqq_u32): Likewise.
16547 (vceqq_p8): Likewise.
16548 (vceq_f32): Gate __a == __b on __FAST_MATH__.
16549 (vceqq_f32): Likewise.
16550 (vceq_f16): Likewise.
16551 (vceqq_f16): Likewise.
16553 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16556 * config/arm/iterators.md (NEON_VACMP): Remove.
16557 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
16559 (neon_vca<cmp_op><mode>_insn): Likewise.
16560 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
16563 2021-06-21 Richard Biener <rguenther@suse.de>
16565 PR tree-optimization/101121
16566 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
16567 when we just lack a stmt with the desired op when doing permutation.
16568 (vect_build_slp_tree): When caching a failed SLP build attempt
16569 assert that at least one lane is marked as not matching.
16571 2021-06-21 liuhongt <hongtao.liu@intel.com>
16574 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
16575 register alternative.
16576 (*and<mode>_1): Ditto.
16578 (*andn<mode>_1): Ditto.
16579 (*<code><mode>_1): Ditto.
16580 (*<code>qi_1): Ditto.
16581 (*one_cmpl<mode>2_1): Ditto.
16582 (*one_cmplsi2_1_zext): Ditto.
16583 (*one_cmplqi2_1): Ditto.
16584 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
16585 the order of mask registers to be before general registers.
16587 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
16590 * config/i386/i386.md: New define_peephole2s to shrink writing
16591 1, 2 or 4 consecutive zeros to memory when optimizing for size.
16593 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
16595 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
16596 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
16597 more efficient code when the source can be trivially simplified.
16599 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
16601 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
16602 a range if global is not available.
16603 (ranger_cache::entry_range): Fallback to range_of_def.
16604 * gimple-range-cache.h (range_of_def): Adjust prototype.
16606 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
16608 PR tree-optimization/101014
16609 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
16611 (ranger_cache::~ranger_cache): Ditto.
16612 (ranger_cache::enable_new_values): Delete.
16613 (ranger_cache::push_poor_value): Delete.
16614 (ranger_cache::range_of_def): Remove poor value processing.
16615 (ranger_cache::entry_range): Ditto.
16616 (ranger_cache::fill_block_cache): Ditto.
16617 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
16618 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
16619 * gimple-range.h (class gimple_ranger): Adjust.
16621 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16624 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
16625 derived from arm_canon_arch.
16626 (arm_canon_arch_option): Call it.
16627 (arm_canon_arch_multilib_option): New function.
16628 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
16629 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
16630 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
16631 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
16632 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
16633 * config/arm/arm.opt (mlibarch): New option.
16634 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
16635 of march on RHS with mlibarch.
16637 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
16639 * config.in: Regenerate.
16640 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
16642 * configure: Regenerate.
16643 * configure.ac: Fix for global_load assembler functions.
16645 2021-06-18 Richard Biener <rguenther@suse.de>
16647 PR tree-optimization/101112
16648 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
16649 to lookup a pattern stmt def.
16651 2021-06-18 Jakub Jelinek <jakub@redhat.com>
16653 PR middle-end/101062
16654 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
16655 representatives in QUAL_UNION_TYPE.
16657 2021-06-18 Andrew Pinski <apinski@marvell.com>
16659 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
16660 Add counting of how many times it is done.
16661 (factor_out_conditional_conversion): Likewise.
16662 (match_simplify_replacement): Likewise.
16663 (value_replacement): Likewise.
16664 (spaceship_replacement): Likewise.
16665 (cond_store_replacement): Likewise.
16666 (cond_if_else_store_replacement_1): Likewise.
16667 (hoist_adjacent_loads): Likewise.
16669 2021-06-18 Andrew Pinski <apinski@marvell.com>
16671 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
16672 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
16673 (verify_gimple_assign_binary): Reject point and offset types on
16674 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
16675 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
16676 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
16678 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
16680 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
16681 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
16682 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
16685 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
16687 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
16688 earlyclobber to alts 0/1.
16689 (gen_addadd): Add earlyclobber to alts 0/1.
16690 * config/rs6000/fusion.md: Regenerate file.
16692 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16694 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
16696 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16698 * gimple-range-cache.cc: Comment cleanups.
16699 * gimple-range-gori.cc: Comment cleanups.
16700 * gimple-range.cc: Comment/spacing cleanups
16701 * value-range.h: Comment cleanups.
16703 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
16706 * calls.c (expand_call): Replace PUSH_ARGS with
16707 targetm.calls.push_argument (0).
16708 (emit_library_call_value_1): Likewise.
16709 * defaults.h (PUSH_ARGS): Removed.
16710 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
16711 targetm.calls.push_argument (0).
16712 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
16713 (emit_push_insn): Pass the number bytes to push to
16714 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
16715 * hooks.c (hook_bool_uint_true): New.
16716 * hooks.h (hook_bool_uint_true): Likewise.
16717 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
16718 targetm.calls.push_argument (0).
16719 * target.def (push_argument): Add a targetm.calls hook.
16720 * targhooks.c (default_push_argument): New.
16721 * targhooks.h (default_push_argument): Likewise.
16722 * config/bpf/bpf.h (PUSH_ARGS): Removed.
16723 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
16724 * config/cr16/cr16.h (PUSH_ARGS): Removed.
16725 * config/i386/i386.c (ix86_push_argument): New.
16726 (TARGET_PUSH_ARGUMENT): Likewise.
16727 * config/i386/i386.h (PUSH_ARGS): Removed.
16728 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
16729 * config/m32c/m32c.h (PUSH_ARGS): Removed.
16730 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
16731 * config/pru/pru.h (PUSH_ARGS): Likewise.
16732 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
16733 TARGET_PUSH_ARGUMENT hook.
16734 * doc/tm.texi: Regenerated.
16736 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
16739 * config/i386/i386-expand.c (expand_vector_set_var):
16740 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
16741 ix86_expand_vector_init_duplicate.
16742 (ix86_expand_vector_init_duplicate): Emit insv_1 for
16743 QImode for !TARGET_PARTIAL_REG_STALL.
16744 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
16745 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
16746 as operand 2 predicate. Call ix86_expand_vector_set_var
16747 for non-constant index operand.
16748 (vec_setv2si): Ditto.
16749 (vec_setv4hi): Ditto.
16750 (vec_setv8qi): ditto.
16752 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
16754 PR tree-optimization/100790
16755 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
16758 2021-06-17 Martin Liska <mliska@suse.cz>
16760 * doc/invoke.texi: Use consistently -O1 instead of -O.
16762 2021-06-17 Martin Liska <mliska@suse.cz>
16764 * gcov-io.h: Update documentation entry about string format.
16766 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
16769 * config/s390/vecintrin.h (vec_doublee): Fix to use
16770 __builtin_s390_vflls.
16771 (vec_floate): Fix to use __builtin_s390_vflrd.
16773 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16775 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
16776 * dominance.h (get_dominated_to_depth): Likewise.
16777 (get_all_dominated_blocks): Likewise.
16778 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
16779 * gcse.c (hoist_code): Likewise.
16780 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
16781 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
16782 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
16783 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
16785 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16787 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
16788 * dominance.h (get_dominated_by_region): Likewise.
16789 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
16790 (gimple_duplicate_sese_tail): Likewise.
16791 (move_sese_region_to_fn): Likewise.
16793 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16795 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
16796 * dominance.h (get_dominated_by): Likewise.
16797 * auto-profile.c (afdo_find_equiv_class): Adjust.
16798 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
16799 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
16800 * tree-cfg.c (test_linear_chain): Likewise.
16801 (test_diamond): Likewise.
16803 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16805 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
16806 * cfgloopanal.c (get_loop_hot_path): Likewise.
16807 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
16809 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16811 * cgraph.c (cgraph_node::collect_callers): Return
16812 auto_vec<cgraph_edge *>.
16813 * cgraph.h (cgraph_node::collect_callers): Likewise.
16814 * ipa-cp.c (create_specialized_node): Adjust.
16815 (decide_about_value): Likewise.
16816 (decide_whether_version_node): Likewise.
16817 * ipa-sra.c (process_isra_node_results): Likewise.
16819 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16821 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
16822 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
16824 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
16827 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
16829 * gimple-range.cc (debug_seed_ranger): New.
16830 (dump_ranger): New.
16831 (debug_ranger): New.
16833 2021-06-17 Richard Biener <rguenther@suse.de>
16835 PR tree-optimization/54400
16836 * tree-vectorizer.h (enum slp_instance_kind): Add
16837 slp_inst_kind_bb_reduc.
16838 (reduction_fn_for_scalar_code): Declare.
16839 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
16840 Check SLP_INSTANCE_KIND instead of looking at the
16842 (vect_slp_analyze_instance_alignment): Likewise.
16843 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
16844 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
16845 chain linearization from vect_build_slp_tree_2 and generalize
16846 for the use of BB reduction vectorization.
16847 (vect_build_slp_tree_2): Adjust accordingly.
16848 (vect_optimize_slp): Elide permutes at the root of BB reduction
16850 (vectorizable_bb_reduc_epilogue): New function.
16851 (vect_slp_prune_covered_roots): Likewise.
16852 (vect_slp_analyze_operations): Use them.
16853 (vect_slp_check_for_constructors): Recognize associatable
16854 chains for BB reduction vectorization.
16855 (vectorize_slp_instance_root_stmt): Generate code for the
16856 BB reduction epilogue.
16858 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16860 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
16862 (gori_compute::may_recompute_p): New.
16863 (gori_compute::outgoing_edge_range_p): Perform recomputations.
16864 * gimple-range-gori.h (class gori_compute): Add prototype.
16866 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16868 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
16869 true when a range can be calculated.
16870 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
16872 2021-06-16 Martin Sebor <msebor@redhat.com>
16874 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
16875 Correct documented defaults.
16877 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
16879 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
16880 m_new_value_p directly.
16882 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
16885 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
16886 Handle 64bit modes for TARGET_SSE4_1.
16887 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
16888 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
16889 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
16890 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
16891 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
16892 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
16894 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16896 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
16897 Change to an expander that emits the correct instruction
16898 depending on endianness.
16899 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
16900 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
16902 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16904 * config/aarch64/aarch64-simd-builtins.def: Split generator
16905 for aarch64_<su>qmovn builtins into scalar and vector
16907 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
16909 (aarch64_<su>qmovn<mode>_insn_be): Define.
16910 (aarch64_<su>qmovn<mode>): Split into scalar and vector
16911 variants. Change vector variant to an expander that emits the
16912 correct instruction depending on endianness.
16914 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16916 * config/aarch64/aarch64-simd-builtins.def: Split generator
16917 for aarch64_sqmovun builtins into scalar and vector variants.
16918 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
16919 Split into scalar and vector variants. Change vector variant
16920 to an expander that emits the correct instruction depending
16922 (aarch64_sqmovun<mode>_insn_le): Define.
16923 (aarch64_sqmovun<mode>_insn_be): Define.
16925 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16927 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
16928 Define - modeling zero-high-half semantics.
16929 (aarch64_xtn<mode>): Change to an expander that emits the
16930 appropriate instruction depending on endianness.
16931 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
16933 (aarch64_xtn2<mode>_le): Rename to...
16934 (aarch64_xtn2<mode>_insn_le): This.
16935 (aarch64_xtn2<mode>_be): Rename to...
16936 (aarch64_xtn2<mode>_insn_be): This.
16937 (vec_pack_trunc_<mode>): Emit truncation instruction instead
16939 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
16940 attribute iterator.
16942 2021-06-16 Martin Jambor <mjambor@suse.cz>
16944 PR tree-optimization/100453
16945 * tree-sra.c (create_access): Disqualify any const candidates
16946 which are written to.
16947 (sra_modify_expr): Do not store sub-replacements back to a const base.
16948 (handle_unscalarized_data_in_subtree): Likewise.
16949 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
16950 instead of constant_decl_p.
16952 2021-06-16 Jakub Jelinek <jakub@redhat.com>
16954 PR middle-end/101062
16955 * stor-layout.c (finish_bitfield_representative): For fields in unions
16956 assume nextf is always NULL.
16957 (finish_bitfield_layout): Compute bit field representatives also in
16958 unions, but handle it as if each bitfield was the only field in the
16961 2021-06-16 Richard Biener <rguenther@suse.de>
16963 PR tree-optimization/101088
16964 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
16965 supported refs on edges. Do not assert same ref but
16966 different kind stores are unsuported but mark them so.
16967 (hoist_memory_references): Only look for supported refs
16970 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
16972 PR rtl-optimization/46235
16973 * config/i386/i386.md: New define_split for bt followed by cmov.
16974 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
16975 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
16976 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
16977 by setnc with zero extension.
16979 2021-06-16 Richard Biener <rguenther@suse.de>
16981 PR tree-optimization/101083
16982 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
16983 vectype as argument.
16984 (vect_build_slp_tree_2): Adjust.
16986 2021-06-15 Martin Sebor <msebor@redhat.com>
16988 PR middle-end/100876
16989 * builtins.c: (gimple_call_return_array): Account for size_t
16990 mangling as either unsigned int or unsigned long
16992 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
16994 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
16995 up before eliminating comparisons.
16997 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
16999 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
17000 nonzero|X is nonzero.
17001 (range_op_bitwise_and_tests): Add tests for above.
17003 2021-06-15 Carl Love <cel@us.ibm.com>
17006 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
17008 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
17011 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
17014 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
17015 (struct gimplify_omp_ctx): Extend defaultmap array by one.
17016 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
17017 (omp_notice_variable): Update type classification for Fortran.
17018 (gimplify_scan_omp_clauses): Update calls for new argument; handle
17019 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
17020 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
17021 * langhooks.c (lhd_omp_scalar_p): Likewise.
17022 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
17023 (LANG_HOOKS_DECLS): Add them.
17024 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
17025 omp_scalar_p pointer type to include the new bool argument.
17027 2021-06-15 David Malcolm <dmalcolm@redhat.com>
17029 * doc/analyzer.texi
17030 (Special Functions for Debugging the Analyzer): Add
17031 __analyzer_dump_capacity.
17033 2021-06-15 Jakub Jelinek <jakub@redhat.com>
17036 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
17037 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
17039 2021-06-15 Richard Biener <rguenther@suse.de>
17041 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
17042 check to identify loop latches.
17043 * cfgloop.c (verify_loop_structure): Likewise.
17044 * loop-init.c (apply_loop_flags): Allow marked irreducible
17045 regions even with multiple latches.
17046 * predict.c (rebuild_frequencies): Simplify.
17048 2021-06-15 Richard Biener <rguenther@suse.de>
17050 * tree-ssa-threadupdate.c
17051 (jump_thread_path_registry::mark_threaded_blocks): Assert we
17052 have marked irreducible regions.
17054 2021-06-14 Martin Sebor <msebor@redhat.com>
17057 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
17058 Handle calls to placement new.
17059 (ndecl_dealloc_argno): Avoid placement delete.
17061 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
17064 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
17065 create_tmp_reg_or_ssa_name().
17067 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
17069 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
17070 (ranger_cache::enable_new_values): Set to specified value and
17071 return the old value.
17072 (ranger_cache::disable_new_values): Delete.
17073 (ranger_cache::fill_block_cache): Disable non 1st order derived
17075 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
17076 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
17078 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
17081 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
17082 Return true early when testing with V2HImode.
17083 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
17085 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
17087 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
17088 (mve_vec_unpack<US>_hi_<mode>): New pattern.
17089 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
17090 (mve_vmovntq_<supf><mode>): Prefix with '@'.
17091 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
17093 (vec_unpack<US>_lo_<mode>): Likewise.
17094 (vec_pack_trunc_<mode>): Rename to
17095 neon_quad_vec_pack_trunc_<mode>.
17096 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
17098 (vec_unpack<US>_lo_<mode>): New.
17099 (vec_pack_trunc_<mode>): New.
17101 2021-06-14 Richard Biener <rguenther@suse.de>
17103 PR tree-optimization/100934
17104 * tree-ssa-dom.c (pass_dominator::execute): Properly
17105 mark irreducible regions.
17107 2021-06-14 Martin Liska <mliska@suse.cz>
17109 * doc/invoke.texi: Put r{...} on the same line as @item.
17111 2021-06-14 Martin Liska <mliska@suse.cz>
17113 * doc/invoke.texi: Add missing newline.
17115 2021-06-14 Martin Liska <mliska@suse.cz>
17117 * doc/invoke.texi: Remove '+' charasters.
17119 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
17121 * config.gcc (arc): Add support for with_cpu option.
17122 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
17124 2021-06-14 Richard Biener <rguenther@suse.de>
17126 PR tree-optimization/101031
17127 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
17128 instead of size when accounting for a possibly string
17131 2021-06-14 Martin Liska <mliska@suse.cz>
17133 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
17135 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
17137 * value-query.cc (gimple_range_global): Call get_range_global
17138 if called after inlining.
17140 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
17143 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
17144 Emit constant permutation insn directly from here.
17146 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
17148 * attribs.c (find_attribute_namespace): Iterate over vec<> with
17150 * auto-profile.c (afdo_find_equiv_class): Likewise.
17151 * gcc.c (do_specs_vec): Likewise.
17152 (do_spec_1): Likewise.
17153 (driver::set_up_specs): Likewise.
17154 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
17155 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
17156 (imm_store_chain_info::try_coalesce_bswap): Likewise.
17157 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
17158 (get_location_for_stmts): Likewise.
17159 * graphite-poly.c (print_iteration_domains): Likewise.
17160 (free_poly_bb): Likewise.
17161 (remove_gbbs_in_scop): Likewise.
17162 (free_scop): Likewise.
17163 (dump_gbb_cases): Likewise.
17164 (dump_gbb_conditions): Likewise.
17165 (print_pdrs): Likewise.
17166 (print_scop): Likewise.
17167 * ifcvt.c (cond_move_process_if_block): Likewise.
17168 * lower-subreg.c (decompose_multiword_subregs): Likewise.
17169 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
17170 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
17171 * sel-sched-dump.c (dump_insn_vector): Likewise.
17172 * store-motion.c (store_ops_ok): Likewise.
17173 (store_killed_in_insn): Likewise.
17174 * timevar.c (timer::named_items::print): Likewise.
17175 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
17176 (cleanup_tree_cfg_noloop): Likewise.
17177 * tree-data-ref.c (dump_data_references): Likewise.
17178 (print_dir_vectors): Likewise.
17179 (print_dist_vectors): Likewise.
17180 (dump_data_dependence_relations): Likewise.
17181 (dump_dist_dir_vectors): Likewise.
17182 (dump_ddrs): Likewise.
17183 (create_runtime_alias_checks): Likewise.
17184 (free_subscripts): Likewise.
17185 (save_dist_v): Likewise.
17186 (save_dir_v): Likewise.
17187 (invariant_access_functions): Likewise.
17188 (same_access_functions): Likewise.
17189 (access_functions_are_affine_or_constant_p): Likewise.
17190 (find_data_references_in_stmt): Likewise.
17191 (graphite_find_data_references_in_stmt): Likewise.
17192 (free_dependence_relations): Likewise.
17193 (free_data_refs): Likewise.
17194 * tree-inline.c (copy_debug_stmts): Likewise.
17195 * tree-into-ssa.c (dump_currdefs): Likewise.
17196 (rewrite_update_phi_arguments): Likewise.
17197 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
17198 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
17200 (vect_slp_analyze_node_dependences): Likewise.
17201 (vect_slp_analyze_instance_dependence): Likewise.
17202 (vect_record_base_alignments): Likewise.
17203 (vect_get_peeling_costs_all_drs): Likewise.
17204 (vect_peeling_supportable): Likewise.
17205 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
17206 (vec_info::free_stmt_vec_infos): Likewise.
17208 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
17210 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
17211 (andqi3_1<cczn>): Removed.
17212 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
17213 (H8/SX bit logicals): Split out from other patterns.
17214 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
17215 mulqihi3_const_clobber_flags.
17216 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
17218 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
17221 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
17222 to true if red zone is used.
17223 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
17224 ix86_red_zone_used.
17225 * config/i386/i386.h (machine_function): Add red_zone_used.
17226 (ix86_red_zone_size): Removed.
17227 (ix86_red_zone_used): New.
17228 * config/i386/i386.md (peephole2 patterns): Replace
17229 ix86_red_zone_size with ix86_red_zone_used.
17231 2021-06-12 Jason Merrill <jason@redhat.com>
17233 * doc/extend.texi (unused variable attribute): Applies to
17234 structure fields as well.
17236 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
17238 * auto-profile.c (read_profile): fix a typo in an error string
17240 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
17242 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
17244 * tree-pretty-print.c (dump_omp_clauses): Update.
17245 (dump_generic_node) <OMP_CLAUSE>: Use it.
17247 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17250 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
17251 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
17252 the polymorphic variants matching code.
17253 (__arm_vld1q_z): Likewise.
17254 (__arm_vld2q): Likewise.
17255 (__arm_vld4q): Likewise.
17256 (__arm_vldrbq_gather_offset): Likewise.
17257 (__arm_vldrbq_gather_offset_z): Likewise.
17259 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17261 PR tree-optimization/96392
17262 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
17264 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17266 PR tree-optimization/96392
17267 * fold-const.c (fold_real_zero_addition_p): Take both arguments
17268 of the addition or subtraction, not just the zero. Use this
17269 other argument in tests for signaling NaNs and signed zeros.
17270 (tree_expr_maybe_real_minus_zero_p): New predicate.
17271 * fold-const.h (fold_real_zero_addition_p): Update prototype.
17272 (tree_expr_maybe_real_minus_zero_p): New function prototype.
17273 * match.pd: Update calls to fold_real_zero_addition_p.
17274 Replace HONOR_NANS with tree_expr_maybe_nan_p.
17275 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
17276 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
17277 * tree-ssa-reassoc.c (eliminate_using_constants): Update
17278 call to fold_real_zero_addition_p.
17280 2021-06-11 Richard Biener <rguenther@suse.de>
17282 PR tree-optimization/101025
17283 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
17284 all refs that require dependence checking.
17286 2021-06-11 Richard Biener <rguenther@suse.de>
17288 PR tree-optimization/101028
17289 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
17290 reassoc discovery fails fatally, mark appropriate lanes
17293 2021-06-11 Richard Biener <rguenther@suse.de>
17295 PR tree-optimization/101026
17296 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
17297 have a representative for the associated chain nodes.
17299 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17301 PR rtl-optimization/101008
17302 * simplify-rtx.c (relational_result): New function.
17303 (simplify_logical_relational_operation,
17304 simplify_relational_operation): Use it.
17306 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17309 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
17311 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
17314 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
17315 false if the permutation can be implemented with constant
17316 permutation instruction in wider mode.
17317 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
17318 Handle V8QImode and V4HImode.
17320 2021-06-11 Martin Liska <mliska@suse.cz>
17322 PR gcov-profile/100788
17323 * common.opt: Add new option.
17324 * coverage.c (coverage_begin_function): Emit warning instead on
17325 the internal compiler error.
17326 * doc/invoke.texi: Document the option.
17327 * toplev.c (process_options): Enable it by default.
17329 2021-06-11 Richard Biener <rguenther@suse.de>
17331 PR middle-end/101009
17332 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
17333 to set *init_b to true when we encounter a constant equal
17335 (compute_affine_dependence): Also dump the actual DR_REF.
17337 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17339 PR tree-optimization/100984
17340 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
17341 replacements table.
17342 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
17344 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
17346 * config/rs6000/rs6000.md
17347 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
17348 define_insn_and_split.
17350 2021-06-11 Richard Biener <rguenther@suse.de>
17352 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
17353 to sort operands of the associative chain.
17355 2021-06-11 Richard Biener <rguenther@suse.de>
17357 * system.h (gcc_stablesort_r): Declare.
17358 * sort.cc (gcc_sort_r): Support stable sort.
17359 (gcc_stablesort_r): Define.
17360 * vec.h (vec<>::stablesort): Add.
17362 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
17365 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
17366 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
17367 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
17368 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
17369 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
17370 (expand_vec_perm_interleave2): Handle 64bit modes.
17371 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
17372 (expand_vec_perm_even_odd_1): Ditto.
17373 (ix86_vectorize_vec_perm_const): Ditto.
17374 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
17375 * config/i386/sse.md: ... here.
17376 * config/i386/mmx.md (*vec_interleave_lowv2sf):
17377 New insn_and_split pattern.
17378 (*vec_interleave_highv2sf): Ditto.
17379 (mmx_pshufbv8qi3): New insn pattern.
17380 (*mmx_pblendw): Ditto.
17382 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
17384 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
17385 (build_acc): Likewise.
17386 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
17387 source operands in little-endian mode.
17388 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
17389 (mma_init_builtins): Likewise.
17390 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
17391 ordering for the MMA assemble and build source operands.
17392 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
17394 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
17397 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
17399 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
17401 * config/h8300/extensions.md: Replace _clobber_flags patterns
17404 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
17406 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
17407 (vcond_mask_<mode><tointvec>): this.
17409 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
17410 Thomas Schwinge <thomas@codesourcery.com>
17412 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
17413 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
17414 * gimple.h (enum gf_mask): Split
17415 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
17416 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
17417 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
17418 (is_gimple_omp_oacc): Update.
17419 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
17420 * gimplify.c (gimplify_omp_target_update): Likewise.
17421 * omp-expand.c (expand_omp_target, build_omp_regions_1)
17422 (omp_make_gimple_edges): Likewise.
17423 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
17426 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17428 * value-query.cc (value_query::value_on_edge): Rename name to
17430 (range_query::range_on_edge): Same.
17431 (range_query::value_of_expr): Same.
17432 (range_query::value_on_edge): Same.
17433 * value-query.h (class value_query): Same.
17434 (class range_query): Same.
17436 2021-06-10 Richard Biener <rguenther@suse.de>
17438 PR tree-optimization/101003
17439 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
17440 use the pattern stmt defs when linearizing a chain.
17442 2021-06-10 Jakub Jelinek <jakub@redhat.com>
17445 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
17446 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
17448 2021-06-10 Clement Chigot <clement.chigot@atos.net>
17450 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
17451 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
17453 2021-06-09 Andrew Pinski <apinski@marvell.com>
17455 PR tree-optimization/100925
17456 * match.pd (a ? CST1 : CST2): Limit transformations
17457 that would produce a negative to integeral types only.
17458 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
17460 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17463 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17465 * doc/tm.texi: Correctly update.
17467 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17469 * doc/tm.texi: Correctly update.
17471 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17474 * doc/tm.texi.in (Trampolines): Add a missing blank line.
17476 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
17479 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
17480 and -ftrampolines work only with Ada.
17481 * doc/tm.texi.in (Trampolines): Likewise.
17482 * doc/tm.texi: Regenerated.
17484 2021-06-09 Carl Love <cel@us.ibm.com>
17486 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
17487 Add define for new builtins.
17488 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
17489 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
17490 overloaded builtin definitions.
17491 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
17492 VSIGNEXTSD2Q): Add builtin expansions.
17493 (SIGNEXT): Add P10 overload definition.
17494 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
17495 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
17496 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
17497 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
17498 vsignextend_si_v2di)[VIlong]: Add define_expand.
17499 Make define_insn vsx_sign_extend_si_v2di visible.
17500 * doc/extend.texi: Add documentation for the vec_signexti,
17501 vec_signextll builtins and vec_signextq.
17503 2021-06-09 Carl Love <cel@us.ibm.com>
17505 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
17506 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
17507 __floattikf_sw, __floatuntikf_sw respectively.
17508 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
17509 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
17510 define_insn for mode IEEE 128.
17512 2021-06-09 Carl Love <cel@us.ibm.com>
17514 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
17515 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
17516 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
17517 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
17518 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
17519 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
17520 uses of VSX_TI to VEC_TI.
17522 2021-06-09 Carl Love <cel@us.ibm.com>
17524 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
17526 2021-06-09 Carl Love <cel@us.ibm.com>
17528 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
17530 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
17531 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
17532 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
17533 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
17534 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
17535 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
17536 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
17538 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
17539 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
17540 altivec_vrlqnm): New define_expands.
17541 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
17542 VCMPGTUT_P): Add macro expansions.
17543 (BU_P10V_AV_P): Add builtin predicate definition.
17544 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
17545 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
17546 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
17547 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
17548 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
17549 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
17550 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
17551 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
17552 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
17553 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
17554 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
17555 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
17556 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
17557 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
17558 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
17559 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
17560 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
17561 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
17562 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
17563 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
17564 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
17565 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
17566 P10V_BUILTIN_MODU_V1TI):
17567 New overloaded definitions.
17568 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
17569 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
17570 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
17571 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
17572 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
17573 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
17575 (altivec_init_builtins): New E_V1TImode case statement.
17576 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
17577 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
17578 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
17579 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
17580 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
17581 E_V1TImode]: New case statements.
17582 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
17583 value RS6000_BTI_bool_V1TI.
17584 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
17585 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
17586 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
17587 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
17588 vlshrv1ti3, vashrv1ti3): New define_expands.
17589 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
17590 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
17591 UNSPEC_VSX_MODUQ): New unspecs.
17592 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
17593 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
17595 (vcmpnet): New define_expand.
17596 * doc/extend.texi: Add documentation for the new builtins vec_rl,
17597 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
17598 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
17599 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
17600 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
17601 vec_any_ge, vec_any_le.
17603 2021-06-09 Carl Love <cel@us.ibm.com>
17605 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
17606 bug in argument generation.
17608 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
17610 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
17612 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
17613 remove <supf> iterator.
17614 (mve_vclzq_u<mode>): New.
17615 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
17616 (neon_vclz<mode): Move to ...
17617 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
17618 * config/arm/vec-common.md: ... here. Add support for MVE.
17620 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
17622 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
17623 (@mve_vrhaddq_<supf><mode): Likewise.
17624 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
17625 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
17626 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
17628 2021-06-09 imba-tjd <109224573@qq.com>
17630 * doc/invoke.texi: Fix typo.
17632 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
17634 PR middle-end/53267
17635 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
17636 Support evaluation of fmod/fmodf/fmodl at compile-time.
17638 2021-06-09 Richard Biener <rguenther@suse.de>
17640 PR tree-optimization/100981
17641 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
17642 gimple_get_lhs to also handle calls.
17643 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
17646 2021-06-09 Richard Biener <rguenther@suse.de>
17648 PR tree-optimization/97832
17649 * tree-vectorizer.h (_slp_tree::failed): New.
17650 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
17652 (_slp_tree::~_slp_tree): Free failed.
17653 (vect_build_slp_tree): Retain failed nodes and record
17654 matches in them, copying that back out when running
17655 into a cached fail. Dump start and end of discovery.
17656 (dt_sort_cmp): New.
17657 (vect_build_slp_tree_2): Handle associatable chains
17658 together doing more aggressive operand swapping.
17660 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17663 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
17665 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
17666 Linux and GNU targets.
17668 2021-06-09 Richard Biener <rguenther@suse.de>
17670 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
17673 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17675 * config/arc/arc.md (loop_end): Change it to
17676 define_insn_and_split.
17678 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17680 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
17681 (machi): New pattern.
17682 (umaddhisi4): Use VMAC2HU instruction.
17683 (umachi): New pattern.
17685 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17687 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
17688 * config/arc/arc.c (arc_split_move_p): New function.
17689 (arc_split_move): Clean up.
17690 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
17691 (movdf_insn): Likewise.
17692 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
17694 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
17697 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
17698 argument to "raw". Do not emit segment overrides when "raw" is true.
17700 2021-06-09 Martin Liska <mliska@suse.cz>
17702 * doc/gcov.texi: Create a proper JSON files.
17703 * doc/invoke.texi: Remove dots in order to make it a valid
17706 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
17708 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
17709 (insn_is_load_p): Use pattern_is_rotate64.
17710 (insn_is_swap_p): Likewise.
17711 (quad_aligned_load_p): Likewise.
17712 (const_load_sequence_p): Likewise.
17713 (replace_swapped_aligned_load): Likewise.
17714 (recombine_lvx_pattern): Likewise.
17715 (recombine_stvx_pattern): Likewise.
17717 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
17719 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
17720 fur_stmt source record.
17721 * gimple-range.cc (fur_source::get_operand): Generic range query.
17722 (fur_source::get_phi_operand): New.
17723 (fur_source::register_dependency): New.
17724 (fur_source::query): New.
17725 (class fur_edge): New. Edge source for operands.
17726 (fur_edge::fur_edge): New.
17727 (fur_edge::get_operand): New.
17728 (fur_edge::get_phi_operand): New.
17729 (fur_edge::query): New.
17730 (fur_stmt::fur_stmt): New.
17731 (fur_stmt::get_operand): New.
17732 (fur_stmt::get_phi_operand): New.
17733 (fur_stmt::query): New.
17734 (class fur_depend): New. Statement source and process dependencies.
17735 (fur_depend::fur_depend): New.
17736 (fur_depend::register_dependency): New.
17737 (class fur_list): New. List source for operands.
17738 (fur_list::fur_list): New.
17739 (fur_list::get_operand): New.
17740 (fur_list::get_phi_operand): New.
17741 (fold_range): New. Instantiate appropriate fur_source class and fold.
17742 (fold_using_range::range_of_range_op): Use new API.
17743 (fold_using_range::range_of_address): Ditto.
17744 (fold_using_range::range_of_phi): Ditto.
17745 (imple_ranger::fold_range_internal): Use fur_depend class.
17746 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
17747 * gimple-range.h (class fur_source): Now a base class.
17748 (class fur_stmt): New.
17749 (fold_range): New prototypes.
17750 (fur_source::fur_source): Delete.
17752 2021-06-08 Andrew Pinski <apinski@marvell.com>
17754 PR tree-optimization/25290
17755 * tree-ssa-phiopt.c (xor_replacement): Delete.
17756 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
17757 (match_simplify_replacement): Allow one cheap preparation
17758 statement that can be moved to before the if.
17760 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
17762 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
17763 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
17765 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
17767 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
17768 Create length attribute on define_insn_and_split. Only split for cases which we
17770 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
17771 fix length computation.
17772 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
17774 2021-06-08 Richard Biener <rguenther@suse.de>
17776 PR tree-optimization/100923
17777 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
17778 the operand vector to be valueized.
17779 (valueize_refs): Likewise.
17780 (valueize_shared_reference_ops_from_ref): Adjust.
17781 (valueize_shared_reference_ops_from_call): Likewise.
17782 (vn_reference_lookup_3): Likewise.
17783 (vn_reference_lookup_pieces): Likewise. Re-valueize
17784 with honoring availability when we are about to create
17785 the ao_ref and valueized before.
17786 (vn_reference_lookup): Likewise.
17787 (vn_reference_insert_pieces): Adjust.
17789 2021-06-08 Richard Biener <rguenther@suse.de>
17791 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
17792 (_slp_instance::root_stmts): ... a vector.
17793 (SLP_INSTANCE_ROOT_STMT): Rename to ...
17794 (SLP_INSTANCE_ROOT_STMTS): ... this.
17795 (slp_root::root): Change to...
17796 (slp_root::roots): ... a vector.
17797 (slp_root::slp_root): Adjust.
17798 * tree-vect-slp.c (_slp_instance::location): Adjust.
17799 (vect_free_slp_instance): Release the root stmt vector.
17800 (vect_build_slp_instance): Adjust.
17801 (vect_analyze_slp): Likewise.
17802 (_bb_vec_info::~_bb_vec_info): Likewise.
17803 (vect_slp_analyze_operations): Likewise.
17804 (vect_bb_vectorization_profitable_p): Likewise. Adjust
17805 costs for the root stmt.
17806 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
17808 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
17810 (vectorize_slp_instance_root_stmt): Adjust.
17811 (vect_schedule_slp): Likewise.
17813 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
17815 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
17816 (ssa_equiv_stack::ssa_equiv_stack): New.
17817 (ssa_equiv_stack::~ssa_equiv_stack): New.
17818 (ssa_equiv_stack::enter): New.
17819 (ssa_equiv_stack::leave): New.
17820 (ssa_equiv_stack::push_replacement): New.
17821 (ssa_equiv_stack::get_replacement): New.
17822 (is_pointer_ssa): New.
17823 (class pointer_equiv_analyzer): New.
17824 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
17825 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
17826 (pointer_equiv_analyzer::set_global_equiv): New.
17827 (pointer_equiv_analyzer::set_cond_equiv): New.
17828 (pointer_equiv_analyzer::get_equiv): New.
17829 (pointer_equiv_analyzer::enter): New.
17830 (pointer_equiv_analyzer::leave): New.
17831 (pointer_equiv_analyzer::get_equiv_expr): New.
17832 (pta_valueize): New.
17833 (pointer_equiv_analyzer::visit_stmt): New.
17834 (pointer_equiv_analyzer::visit_edge): New.
17835 (hybrid_folder::value_of_expr): Call PTA.
17836 (hybrid_folder::value_on_edge): Same.
17837 (hybrid_folder::pre_fold_bb): New.
17838 (hybrid_folder::post_fold_bb): New.
17839 (hybrid_folder::pre_fold_stmt): New.
17840 (rvrp_folder::pre_fold_bb): New.
17841 (rvrp_folder::post_fold_bb): New.
17842 (rvrp_folder::pre_fold_stmt): New.
17843 (rvrp_folder::value_of_expr): Call PTA.
17844 (rvrp_folder::value_on_edge): Same.
17846 2021-06-08 Jakub Jelinek <jakub@redhat.com>
17849 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
17850 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
17852 2021-06-08 Richard Biener <rguenther@suse.de>
17854 PR middle-end/100951
17855 * tree-vect-generic.c (expand_vector_piecewise): Build a
17856 VECTOR_CST if all elements are constant.
17857 (expand_vector_condition): Likewise.
17858 (lower_vec_perm): Likewise.
17859 (expand_vector_conversion): Likewise.
17861 2021-06-08 Martin Liska <mliska@suse.cz>
17863 * doc/invoke.texi: Document new param evrp-sparse-threshold.
17865 2021-06-08 Martin Liska <mliska@suse.cz>
17867 * genautomata.c (create_automata): Fix typo.
17869 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
17871 PR tree-optimization/100794
17872 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
17873 allow_unroll_p and only allow unrolling when it's true.
17874 (tree_predictive_commoning): Add parameter allow_unroll_p and
17876 (run_tree_predictive_commoning): Likewise.
17877 (pass_predcom::gate): Check flag_tree_loop_vectorize and
17878 global_options_set.x_flag_predictive_commoning.
17879 (pass_predcom::execute): Adjust for allow_unroll_p.
17881 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
17883 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
17884 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
17885 lambda function cleanup, remove scev_reset call, and adjust return
17887 (tree_predictive_commoning): Adjust for different changed values,
17888 only set flag TODO_update_ssa_only_virtuals if changed.
17889 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
17890 from todo_flags_finish.
17892 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
17894 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
17895 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
17896 (sbr_sparse_bitmap::bitmap_set_quad): New.
17897 (sbr_sparse_bitmap::bitmap_get_quad): New.
17898 (sbr_sparse_bitmap::set_bb_range): New.
17899 (sbr_sparse_bitmap::get_bb_range): New.
17900 (sbr_sparse_bitmap::bb_range_p): New.
17901 (block_range_cache::block_range_cache): initialize bitmap obstack.
17902 (block_range_cache::~block_range_cache): Destruct obstack.
17903 (block_range_cache::set_bb_range): Decide when to utilze the
17904 sparse on entry cache.
17905 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
17906 * params.opt (-param=evrp-sparse-threshold): New.
17908 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
17910 * bitmap.c (bitmap_set_aligned_chunk): New.
17911 (bitmap_get_aligned_chunk): New.
17912 (test_aligned_chunk): New.
17913 (bitmap_c_tests): Call test_aligned_chunk.
17914 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
17916 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
17919 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
17921 (ix86_expand_vector_init_one_nonzero): Ditto.
17922 (ix86_expand_vector_init_one_var): Ditto.
17923 (ix86_expand_vector_init_general): Ditto.
17924 * config/i386/mmx.md (vec_initv4qiqi): New expander.
17926 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
17928 * config/h8300/movepush.md: Change most _clobber_flags
17929 patterns to instead use <cczn> subst.
17930 (movsi_cczn): New pattern with usable CC cases split out.
17931 (movsi_h8sx_cczn): Likewise.
17933 2021-06-07 Martin Liska <mliska@suse.cz>
17935 * common/common-target.def: Split long lines and replace them
17937 * target.def: Likewise.
17938 * doc/tm.texi: Re-generated.
17940 2021-06-07 Jakub Jelinek <jakub@redhat.com>
17943 * fold-const.c (fold_read_from_vector): Return NULL if trying to
17944 read from a CONSTRUCTOR with vector type elements.
17946 2021-06-07 Jakub Jelinek <jakub@redhat.com>
17948 PR middle-end/100898
17949 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
17950 should copy any arguments. Don't call gimple_call_num_args
17951 on id->call_stmt or call_stmt more than once.
17953 2021-06-07 liuhongt <hongtao.liu@intel.com>
17956 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
17958 (<insn>v4siv4di2): Delete constraints for define_expand.
17960 2021-06-07 liuhongt <hongtao.liu@intel.com>
17963 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
17964 assignment of cfun->machine->has_explicit_vzeroupper.
17965 * config/i386/i386-features.c
17966 (ix86_add_reg_usage_to_vzerouppers): Delete.
17967 (ix86_add_reg_usage_to_vzeroupper): Ditto.
17968 (rest_of_handle_insert_vzeroupper): Remove
17969 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
17971 (gate): Remove cfun->machine->has_explicit_vzeroupper.
17972 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
17974 * config/i386/i386.c (ix86_insn_callee_abi): New function.
17975 (ix86_initialize_callee_abi): Ditto.
17976 (ix86_expand_avx_vzeroupper): Ditto.
17977 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
17979 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
17980 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
17982 * config/i386/i386.h (struct GTY(()) machine_function): Delete
17983 has_explicit_vzeroupper.
17984 * config/i386/i386.md (enum unspec): New member
17986 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
17987 define_constants for insn callee abi index.
17988 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
17989 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
17990 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
17991 (*avx_vzeroupper): Rename to ..
17992 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
17993 call_insn which has a special vzeroupper ABI.
17994 (*avx_vzeroupper_1): Deleted.
17996 2021-06-07 liuhongt <hongtao.liu@intel.com>
17999 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
18000 it won't use stack pointer reg.
18001 * final.c (leaf_function_p): When call_insn is a fake call, it
18002 won't affect caller as a leaf function.
18003 * reg-stack.c (callee_clobbers_any_stack_reg): New.
18004 (subst_stack_regs): When call_insn doesn't clobber any stack
18005 reg, don't clear the arguments.
18006 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
18008 * shrink-wrap.c (requires_stack_frame_p): No need for stack
18009 frame for a fake call.
18010 * rtl.h (FAKE_CALL_P): New macro.
18012 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
18014 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
18016 (sparc_order_regs_for_local_alloc): ...this.
18017 (sparc_leaf_reg_remap): Declare.
18018 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
18019 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
18020 * config/sparc/sparc.c (leaf_reg_remap): Delete.
18021 (order_regs_for_local_alloc): Rename to...
18022 (sparc_order_regs_for_local_alloc): ...this.
18023 (sparc_leaf_reg_remap): New function.
18024 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
18026 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
18028 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
18029 Use assemble_name to output BSS section name.
18031 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
18033 * config/i386/constraints.md (Bs):
18034 Remove boolean operators from match_test RTX.
18037 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
18040 2021-06-06 Martin Liska <mliska@suse.cz>
18042 * doc/extend.texi: Add missing @headitem.
18043 * doc/invoke.texi: Likewise.
18044 * doc/objc.texi: Likewise.
18046 2021-06-06 Martin Liska <mliska@suse.cz>
18048 * genhooks.c (emit_findices): Remove unused function.
18049 (emit_documentation): Do not call emit_findices
18050 and do not search for @Fcode directives.
18052 2021-06-06 Martin Liska <mliska@suse.cz>
18054 * doc/invoke.texi: Remove extra character.
18056 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18058 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
18060 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18062 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
18063 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
18064 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
18065 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
18067 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18069 * config/or1k/or1k.md (*movdi): Fix empty split condition.
18071 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18073 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
18076 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18078 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
18079 *zero_extendsidi2): Fix empty split condition.
18081 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
18083 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
18085 * config/h8300/bitfield.md: Likewise.
18086 * config/h8300/combiner.md: Likewise.
18087 * config/h8300/divmod.md: Likewise.
18088 * config/h8300/extensions.md: Likewise.
18089 * config/h8300/jumpcall.md: Likewise.
18090 * config/h8300/movepush.md: Likewise.
18091 * config/h8300/multiply.md: Likewise.
18092 * config/h8300/other.md: Likewise.
18093 * config/h8300/shiftrotate.md: Likewise.
18094 * config/h8300/logical.md: Likewise. Fix split pattern to use
18095 code iterator that somehow slipped through.
18097 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
18099 PR middle-end/100905
18100 * tree-nested.c (convert_nonlocal_omp_clauses,
18101 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
18103 2021-06-04 Martin Sebor <msebor@redhat.com>
18105 PR middle-end/100732
18106 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
18107 with either source or destination argument of invalid type.
18108 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
18109 calls with arguments of invalid type.
18111 2021-06-04 Martin Sebor <msebor@redhat.com>
18113 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
18115 (attr_access::vla_bounds): Also handle VLA bounds.
18117 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
18119 * config/i386/predicates.md (GOT_memory_operand):
18120 Implement using match_code RTXes.
18121 (GOT32_symbol_operand): Ditto.
18123 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
18126 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18128 (ix86_expand_vector_init_general): Ditto.
18129 Use SImode instead of word_mode for logic operations
18130 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
18131 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
18132 implemented by expand_vec_perm_1.
18133 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
18134 should be implemented using standard shuffle patterns.
18135 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
18136 V2HI modes to modes, implementable with shuffle for one operand.
18137 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
18138 (*pshufw_1): New insn pattern.
18139 (*vec_dupv2hi): Ditto.
18140 (vec_initv2hihi): New expander.
18142 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
18144 * config/arm/vfp.md (no_literal_pool_df_immediate,
18145 no_literal_pool_sf_immediate): Fix empty split condition.
18147 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
18149 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
18150 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
18151 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
18152 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
18153 *<sse4_1_avx2>_pblendvb_lt): Likewise.
18155 2021-06-04 Jakub Jelinek <jakub@redhat.com>
18158 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
18159 concatenation from half-sized modes with TImode elements.
18161 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
18163 * config/arc/arc.c (arc_override_options): Disable millicode
18164 thunks when RF16 is on.
18166 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18168 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
18170 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18172 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
18173 Replace PROMOTE_MODE marco with its content.
18175 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18177 * config/cris/cris.md (*addi_reload): Fix empty split condition.
18179 2021-06-03 Jim Wilson <jimw@sifive.com>
18181 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
18182 turn it on for all riscv targets.
18184 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
18187 * config/i386/i386-expand.c (ix86_expand_vector_set):
18188 Handle V2HI and V4QI modes.
18189 (ix86_expand_vector_extract): Ditto.
18190 * config/i386/mmx.md (*pinsrw): New insn pattern.
18193 (*pextrw_zext): Ditto.
18195 (*pextrb_zext): Ditto.
18196 (vec_setv2hi): New expander.
18197 (vec_extractv2hihi): Ditto.
18198 (vec_setv4qi): Ditto.
18199 (vec_extractv4qiqi): Ditto.
18200 (vec_setv8qi): Enable only for TARGET_SSE4_1.
18201 (vec_extractv8qiqi): Ditto.
18203 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
18205 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
18206 order to subf instruction.
18207 * config/rs6000/fusion.md: Regenerate.
18209 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18211 * calls.c (get_size_range): Use range_of_expr instead of
18212 determine_value_range.
18213 * tree-affine.c (expr_to_aff_combination): Same.
18214 * tree-data-ref.c (split_constant_offset): Same.
18215 * tree-vrp.c (determine_value_range_1): Remove.
18216 (determine_value_range): Remove.
18217 * tree-vrp.h (determine_value_range): Remove.
18219 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18221 * function-tests.c (test_ranges): Call gimple_range_tests.
18222 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
18224 * gimple-range.cc (fur_source::get_operand): Do not call
18225 get_tree_range or gimple_range_global.
18227 (get_tree_range): Move to value-query.cc.
18228 Call get_arith_expr_range.
18229 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
18230 Include gimple-range-tests.cc.
18231 * gimple-range.h (fold_range): Add argument.
18232 (get_tree_range): Remove.
18233 * selftest.h (gimple_range_tests): New.
18234 * value-query.cc (global_range_query::range_of_expr): Add
18236 (range_query::get_tree_range): Move from gimple-range.cc.
18237 * value-query.h (class range_query): Add get_tree_range and
18238 get_arith_expr_range. Make fur_source a friend.
18239 * vr-values.c (vr_values::range_of_expr): Pass stmt to
18241 * gimple-range-tests.cc: New file.
18243 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18245 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
18246 update_global_range.
18247 * value-query.cc (update_global_range): New.
18248 * value-query.h (update_global_range): New.
18250 2021-06-03 David Malcolm <dmalcolm@redhat.com>
18252 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
18253 printing the same location twice if there are fix-it hints,
18254 multiple locations, or a label.
18256 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
18258 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
18259 thresholds to narrow the upper bound on epilogue iterations.
18261 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
18263 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
18264 (mve_vabsq_s<mode>): Likewise.
18265 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
18266 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
18267 * config/arm/vec-common.md (neg<mode>2): Rename to
18268 <absneg_str><mode>2.
18270 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
18272 * common/config/arc/arc-common.c (arc_option_optimization_table):
18273 Remove malign-call.
18274 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
18275 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
18276 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
18277 * config/arc/arc.md (abssi2_mixed): Remove pattern.
18278 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
18279 (malign-call): Likewise.
18280 (mmixed-code): Likewise.
18281 * doc/invoke.texi (ARC): Update doc.
18283 2021-06-03 Martin Liska <mliska@suse.cz>
18285 * common.opt: Use proper Enum values.
18286 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
18287 (parse_sanitizer_options): Handle only sanitizer_opts.
18288 (common_handle_option): Just assign value.
18290 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18293 * tree-inline.c (inline_forbidden_p): Remove test on return type.
18295 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18297 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
18298 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
18299 (type_byte_size): Inline into...
18300 (add_byte_size_attribute): ...this and call add_scalar_info.
18302 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18304 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
18305 (typed_binop_from_tree): New function.
18306 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
18307 turn a divide by a power of 2 into a shift.
18308 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
18309 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
18310 typed divide by calling typed_binop_from_tree.
18312 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18314 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
18315 (is_handled_procedure_type): Likewise.
18316 (struct loc_descr_context): Add strict_signedness field.
18317 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
18318 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
18319 (resolve_args_picking): Minor tweak.
18320 (function_to_dwarf_procedure): Initialize strict_signedness field.
18321 (type_byte_size): Likewise.
18322 (field_byte_offset): Likewise.
18323 (gen_descr_array_type_die): Likewise.
18324 (gen_variant_part): Likewise.
18325 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
18326 to true when a context is present before evaluating the arguments.
18327 <COND_EXPR>: Do not generate a useless comparison with zero.
18328 When dereferencing an address, if strict_signedness is true and the
18329 type is small and signed, use DW_OP_deref_type to do the dereference
18330 and then DW_OP_convert to convert back to the generic type.
18332 2021-06-03 Jakub Jelinek <jakub@redhat.com>
18335 * tree-inline.c (copy_tree_body_r): Handle iterators on
18336 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
18338 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18340 * config/arc/arc.md (*bbit_di): Remove.
18342 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
18344 PR rtl-optimization/100264
18345 * ree.c (get_sub_rtx): Ignore SET expressions without register
18346 destinations and remove assertion, as it is not valid anymore
18347 with this new behaviour.
18348 (merge_def_and_ext): Eliminate destination check for register
18349 as such SET expressions can't occur anymore.
18350 (combine_reaching_defs): Likewise.
18352 2021-06-02 Jakub Jelinek <jakub@redhat.com>
18355 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
18356 -Wtype-limits warnings.
18357 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
18358 in operands to avoid -Wsign-compare warnings.
18360 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
18362 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
18365 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
18367 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
18369 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
18371 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
18373 * config/s390/subst.md(cconly_subst): Use a single constraint
18374 in (match_scratch).
18376 2021-06-02 Martin Liska <mliska@suse.cz>
18378 * ipa-icf.h: Use auto_vec for memory_access_types.
18380 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
18382 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
18383 argument from prototype.
18384 (output_logical_op): Add rtx_code argument.
18385 (compute_logical_op_length): Likewise.
18386 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
18387 to compute_a_shift_length.
18388 (output_logical_op); New argument with the rtx code rather than
18389 extracting it from an operand. Handle QImode too.
18390 (compute_logical_op_length): Similary.
18391 (compute_a_shift_length): Drop unused argument.
18392 * config/h8300/h8300.md (logicals): New code iterator.
18393 * config/h8300/logical.md (<code><mode>3 expander): Combine
18394 the "and" expander with the "ior"/"xor" expander.
18395 (bclr<mode>msx): Combine the QI/HI mode patterns.
18396 (<logical><mode>3 insns): Use code iterator rather than match_operator.
18397 Handle QImode as well. Update call to output_logical_op and
18398 compute_logical_op_length to pass in rtx_code
18399 Fix split condition on all define_insn_and_split patterns.
18400 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
18401 the flags and setting ZN via existing define_subst.
18402 * config/h8300/shiftrotate.md: Drop unused argument from
18403 calls to compute_a_shift_length.
18404 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
18406 2021-06-01 Andrew Pinski <apinski@marvell.com>
18408 PR tree-optimization/25290
18409 * tree-ssa-phiopt.c (match_simplify_replacement):
18411 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
18412 (two_value_replacement): Change the comment about
18413 conditional_replacement.
18414 (conditional_replacement): Delete.
18416 2021-06-01 Andrew Pinski <apinski@marvell.com>
18418 PR tree-optimization/95481
18419 * tree-tailcall.c (find_tail_calls): Handle empty typed
18422 2021-06-01 Andrew Pinski <apinski@marvell.com>
18424 * gimplify.c (zero_sized_field_decl): Delete
18425 (zero_sized_type): Delete
18426 (gimplify_init_ctor_eval): Use is_empty_type instead
18427 of zero_sized_field_decl.
18428 (gimplify_modify_expr): Use is_empty_type instead of
18431 2021-06-01 Jason Merrill <jason@redhat.com>
18434 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
18436 2021-06-01 Jason Merrill <jason@redhat.com>
18439 * diagnostic.h (warning_enabled_at): Declare.
18440 * diagnostic.c (diagnostic_enabled): Factor out from...
18441 (diagnostic_report_diagnostic): ...here.
18442 (warning_enabled_at): New.
18444 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
18446 * gimple-ssa-evrp.c: Enable exporting of global ranges.
18448 2021-06-01 Martin Liska <mliska@suse.cz>
18451 * doc/invoke.texi: Mention that -fgcse-after-reload
18452 is enabled with -O3.
18454 2021-06-01 liuhongt <hongtao.liu@intel.com>
18456 PR tree-optimization/98365
18457 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
18458 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
18459 (convert_scalar_cond_reduction): Ditto.
18460 (predicate_scalar_phi): Ditto.
18462 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18464 PR tree-optimization/100781
18465 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
18466 value calculation by default.
18467 (ranger_cache::enable_new_values): New.
18468 (ranger_cache::disable_new_values): New.
18469 (ranger_cache::push_poor_value): Check if new values are allowed.
18470 * gimple-range-cache.h (class ranger_cache): New member/methods.
18471 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
18472 statement, and disable/renable new value calculation.
18474 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18476 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
18477 (ranger_cache::range_of_def): New.
18478 (ranger_cache::entry_range): New.
18479 (ranger_cache::exit_range): New.
18480 (ranger_cache::range_of_expr): Adjust.
18481 (ranger_cache::range_on_edge): Adjust.
18482 (ranger_cache::propagate_cache): Call exit_range directly.
18483 * gimple-range-cache.h (class ranger_cache): Adjust.
18485 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18487 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
18488 gori_compute being a member rather than base class.
18489 dervied call to member call.
18490 (ranger_cache::dump): No longer dump gori_map.
18491 (ranger_cache::dump_bb): New.
18492 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
18493 being a member rather than base class.
18494 (ranger_cache::set_global_range): Ditto.
18495 (ranger_cache::ssa_range_in_bb): Ditto.
18496 (ranger_cache::range_of_expr): New.
18497 (ranger_cache::range_on_edge): New.
18498 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
18499 (ranger_cache::propagate_cache): Adjust debugging output.
18500 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
18502 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
18503 member, and inherit from range_query instead.
18504 (ranger_cache::dump_bb): New. split from dump.
18505 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
18506 (gori_compute::expr_range_at_stmt): Delete.
18507 (gori_compute::compute_name_range_op): Delete.
18508 (gori_compute::compute_operand_range_switch): Add fur_source.
18509 (gori_compute::compute_operand_range): Add fur_source param, inline
18510 old compute_name_range_op and optimize_logical_operands.
18511 (struct tf_range): Delete.
18512 (gori_compute::logical_combine): Adjust
18513 (gori_compute::optimize_logical_operands): Delete.
18514 (gori_compute::compute_logical_operands_in_chain): Delete.
18515 (gori_compute::compute_logical_operands): Adjust.
18516 (gori_compute::compute_operand1_range): Adjust to fur_source.
18517 (gori_compute::compute_operand2_range): Ditto.
18518 (gori_compute::compute_operand1_and_operand2_range): Ditto.
18519 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
18520 and adjust to fur_source.
18521 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
18522 range_query and fur_source.
18523 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
18524 from the ranger_cache..
18525 (gimple_ranger::fold_range_internal): Adjust to base class change of
18527 (gimple_ranger::dump_bb): Adjust dump.
18528 * gimple-range.h (gimple_ranger):export gori computes object.
18530 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18532 PR tree-optimization/100774
18533 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
18534 Constant values are also not stale.
18535 (ranger_cache::set_global_range): Range invariant values should also
18536 have the correct timestamp.
18538 2021-05-31 Martin Liska <mliska@suse.cz>
18540 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
18541 Unpack FUNCTION_DECL_DECL_TYPE.
18542 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
18543 Stream FUNCTION_DECL_DECL_TYPE instead of
18544 DECL_IS_OPERATOR_NEW_P.
18545 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
18547 (DECL_IS_OPERATOR_NEW_P): Likewise.
18548 (DECL_IS_OPERATOR_DELETE_P): Likewise.
18549 (DECL_LAMBDA_FUNCTION_P): Likewise.
18551 2021-05-31 Richard Biener <rguenther@suse.de>
18554 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
18555 * internal-fn.def (SHUFFLEVECTOR): New.
18556 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
18557 * doc/extend.texi: Document __builtin_shufflevector.
18559 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
18562 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
18563 indexed form addresses.
18565 2021-05-29 Jeff Law <jlaw@tachyum.com>
18567 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
18568 parameter. Call callers fixed.
18570 (output_plussi): Add FALLTHRU markers.
18571 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
18573 2021-05-29 Jakub Jelinek <jakub@redhat.com>
18575 PR middle-end/99928
18576 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
18577 combined with parallel, make sure to add shared clause to
18578 parallel for explicit linear clause.
18580 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
18582 PR tree-optimization/100787
18583 * gimple-ssa-evrp.c: Disable exporting of global ranges.
18585 2021-05-28 Jason Merrill <jason@redhat.com>
18587 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
18588 operator--, operator*, operator==, and operator!=.
18589 (class tsi_range): New.
18591 2021-05-28 Richard Biener <rguenther@suse.de>
18593 PR tree-optimization/100778
18594 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
18595 trapping ops in different BBs.
18597 2021-05-28 Richard Biener <rguenther@suse.de>
18600 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
18601 copy fntype from original call.
18603 2021-05-28 Martin Liska <mliska@suse.cz>
18605 PR gcov-profile/100751
18606 * doc/gcov.texi: Revert partially a hunk that was wrong.
18608 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
18610 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
18612 (HAVE_sync_compare_and_swaphi): Likewise.
18613 (HAVE_sync_compare_and_swapsi): Likewise.
18615 2021-05-28 Jakub Jelinek <jakub@redhat.com>
18617 PR middle-end/99928
18618 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
18620 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
18622 * gimplify.c (gimplify_omp_affinity): New.
18623 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
18624 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
18625 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
18626 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
18627 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
18629 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
18630 Richard Biener <rguenther@suse.de>
18632 * match.pd <popcount & / + pattern matching>:
18633 When generating popcount directly fails, try doing it in two halves.
18635 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
18637 * Makefile.in (generated_files): Add gimple-match.c and
18640 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
18642 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
18644 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
18646 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
18648 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
18650 PR tree-optimization/99398
18651 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
18652 where the fed operands are CTOR/CST and propagated through
18653 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
18654 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
18656 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
18659 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18661 * config/i386/mmx.md (addv2sf3): Do not call
18662 ix86_fixup_binary_operands_no_copy.
18665 (<smaxmin:code>v2sf3): Ditto.
18666 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
18667 (<plusminus:insn><VI_32:mode>3): Remove expander.
18668 (<plusminus:insn><VI_32:mode>3): Rename from
18669 "*<plusminus:insn><VI_32:mode>3".
18670 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
18671 (mulv2hi3): Remove expander.
18672 (mulv2hi3): Rename from *mulv2hi3.
18673 (<s>mulv2hi3_highpart): Remove expander.
18674 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
18675 (<smaxmin:code><MMXMODE14:mode>3): Rename from
18676 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
18677 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
18678 (SMAXMIN_MMXMODEI): Remove mode iterator.
18679 (<smaxmin:code>v4hi3): New expander.
18680 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
18681 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
18682 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
18683 (SMAXMIN_VI_32): Remove mode iterator.
18684 (<umaxmin:code><MMXMODE24:mode>3): Rename from
18685 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
18686 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
18687 (UMAXMIN_MMXMODEI): Remove mode iterator.
18688 (<umaxmin:code>v8qi3): New expander.
18689 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
18690 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
18691 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
18692 (UMAXMIN_VI_32): Remove mode iterator.
18693 (<any_shift:insn>v2hi3): Remove expander.
18694 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
18695 (<any_logic:code><MMXMODEI:mode>3): Do not call
18696 ix86_fixup_binary_operands_no_copy.
18697 (<any_logic:code><VI_32:mode>3): Remove expander.
18698 (<any_logic:code><VI_32:mode>3): Rename from
18699 "*<any_logic:code><VI_32:mode>3".
18700 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
18701 * config/i386/sse.md (div<VF2:mode>3): Do not call
18702 ix86_fixup_binary_operands_no_copy.
18703 (div<VF1:mode>3): Ditto.
18704 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
18705 (smulhrsv4hi3): Ditto.
18706 (smulhrsv2hi3): Ditto.
18708 2021-05-27 Martin Sebor <msebor@redhat.com>
18710 * ggc.h (gt_ggc_mx): Add overloads for all integers.
18712 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
18714 (hash_map::operator==): New function.
18716 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18719 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
18720 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
18721 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
18722 (*xop_maskcmp<VI_32:mode>3): Ditto.
18723 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
18724 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
18726 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
18729 * config/arm/arm.c (arm_configure_build_target): Remove parameter
18730 opts_set, directly check opts parameters for being non-null.
18731 (arm_option_restore): Update call to arm_configure_build_target.
18732 (arm_option_override): Likewise.
18733 (arm_can_inline_p): Likewise.
18734 (arm_valid_target_attribute_tree): Likewise.
18735 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
18736 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
18739 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
18741 * vr-values.c (simplify_conversion_using_ranges): Use
18742 get_range_query instead of get_global_range_query.
18744 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
18746 * gimple-range.cc (get_range_global): Move to value-query.cc.
18747 (gimple_range_global): Same.
18748 (get_global_range_query): Same.
18749 (global_range_query::range_of_expr): Same.
18750 * gimple-range.h (class global_range_query): Move to
18752 (gimple_range_global): Same.
18753 * tree-ssanames.c (get_range_info): Move to value-query.cc.
18754 (get_ptr_nonnull): Same.
18755 * tree-ssanames.h (get_range_info): Remove.
18756 (get_ptr_nonnull): Remove.
18757 * value-query.cc (get_ssa_name_range_info): Move from
18759 (get_ssa_name_ptr_info_nonnull): Same.
18760 (get_range_global): Move from gimple-range.cc.
18761 (gimple_range_global): Same.
18762 (get_global_range_query): Same.
18763 (global_range_query::range_of_expr): Same.
18764 * value-query.h (class global_range_query): Move from
18766 (gimple_range_global): Same.
18768 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18771 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
18772 (uavgv2hi3_ceil): Ditto.
18774 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
18777 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
18779 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18781 * tree-ssanames.c (get_range_info): Merge both copies of
18782 get_range_info into one that works with irange.
18783 * tree-ssanames.h (get_range_info): Remove version that works on
18786 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18788 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
18789 (expand_builtin_strnlen): Same.
18790 (determine_block_size): Same.
18791 * fold-const.c (expr_not_equal_to): Same.
18792 * gimple-fold.c (size_must_be_zero_p): Same.
18793 * gimple-match-head.c: Include gimple-range.h.
18794 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
18795 * gimple-ssa-warn-restrict.c
18796 (builtin_memref::extend_offset_range): Same.
18797 * graphite-sese-to-poly.c (add_param_constraints): Same.
18798 * internal-fn.c (get_min_precision): Same.
18799 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
18800 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
18802 * tree-data-ref.c (split_constant_offset): Same.
18803 (dr_step_indicator): Same.
18804 * tree-dfa.c (get_ref_base_and_extent): Same.
18805 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
18806 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
18807 (determine_value_range): Same.
18808 (record_nonwrapping_iv): Same.
18809 (infer_loop_bounds_from_signedness): Same.
18810 (scev_var_range_cant_overflow): Same.
18811 * tree-ssa-phiopt.c (two_value_replacement): Same.
18812 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
18813 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
18814 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
18816 (dump_strlen_info): Same.
18817 (set_strlen_range): Same.
18818 (maybe_diag_stxncpy_trunc): Same.
18819 (get_len_or_size): Same.
18820 (handle_integral_assign): Same.
18821 * tree-ssa-structalias.c (find_what_p_points_to): Same.
18822 * tree-ssa-uninit.c (find_var_cmp_const): Same.
18823 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
18824 * tree-vect-patterns.c (vect_get_range_info): Same.
18825 (vect_recog_divmod_pattern): Same.
18826 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
18827 (register_edge_assert_for_2): Same.
18828 (determine_value_range_1): Same.
18829 * tree.c (get_range_pos_neg): Same.
18830 * vr-values.c (vr_values::get_lattice_entry): Same.
18831 (vr_values::update_value_range): Same.
18832 (simplify_conversion_using_ranges): Same.
18834 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18836 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
18837 get_range_query instead of query argument.
18838 (pass_walloca::execute): Enable and disable global ranger.
18840 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18842 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
18844 (rvrp_folder::~rvrp_folder): Call disable_ranger.
18845 (hybrid_folder::hybrid_folder): Call enable_ranger.
18846 (hybrid_folder::~hybrid_folder): Call disable_ranger.
18848 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18850 * function.c (allocate_struct_function): Set cfun->x_range_query.
18851 * function.h (struct function): Declare x_range_query.
18852 (get_range_query): New.
18853 (get_global_range_query): New.
18854 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
18855 Remove call to safe_grow_cleared.
18856 * gimple-range.cc (get_range_global): New.
18857 (gimple_range_global): Move from gimple-range.h.
18858 (get_global_range_query): New.
18859 (global_range_query::range_of_expr): New.
18860 (enable_ranger): New.
18861 (disable_ranger): New.
18862 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
18863 (class global_range_query): New.
18864 (enable_ranger): New.
18865 (disable_ranger): New.
18866 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
18867 dump_all_value_ranges to dump.
18868 * tree-vrp.c (vrp_prop::finalize): Same.
18869 * value-query.cc (range_query::dump): New.
18870 * value-query.h (range_query::dump): New.
18871 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
18872 (vr_values::dump): ...this.
18873 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
18874 dump and make virtual.
18876 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
18878 * config/i386/i386.c (ix86_autovectorize_vector_modes):
18879 Add V4QImode and V16QImode for TARGET_SSE2.
18880 * doc/sourcebuild.texi (Vector-specific attributes):
18881 Add vect64 and vect32 description.
18883 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
18885 * gimple-range-gori.cc (range_def_chain::register_dependency):
18886 Resize m_def_chain when needed.
18888 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
18890 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
18891 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
18892 * config/arm/vec-common.md: .. here. Add support for MVE.
18894 2021-05-26 Jakub Jelinek <jakub@redhat.com>
18896 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
18898 * config/microblaze/microblaze.c (microblaze_legitimize_address,
18900 microblaze_option_override, print_operand): Likewise.
18901 * config/microblaze/microblaze.md (call_internal_plt,
18902 call_value_intern_plt, call_value_intern): Likewise.
18903 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
18904 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
18905 call_value_multiple_internal1): Likewise.
18906 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
18908 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
18910 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
18911 arc_ccfsm_advance, symbolic_reference_mentioned_p,
18912 arc_raw_symbolic_reference_mentioned_p): Remove register
18915 2021-05-26 Jakub Jelinek <jakub@redhat.com>
18918 * omp-low.c: Include omp-offload.h.
18919 (create_omp_child_function): If current_function_decl has
18920 "omp declare target" attribute and is_gimple_omp_offloaded,
18921 remove that attribute from the copy of attribute list and
18922 add "omp target entrypoint" attribute instead.
18923 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
18924 variables for offloading if in omp_maybe_offloaded_ctx.
18925 * omp-offload.c (pass_omp_target_link::execute): Nullify second
18926 argument to GOMP_target_data_ext in offloaded code.
18928 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18930 * config/csky/csky.c (csky_can_change_mode_class): Delete.
18931 For csky, HF/SF mode use the low bits of VREGS.
18933 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
18935 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
18936 DECL which is a reference for OMP.
18938 2021-05-26 Martin Liska <mliska@suse.cz>
18940 PR gcov-profile/100751
18941 * doc/gcov.texi: Document that __gcov_dump can be called just
18942 once and that __gcov_reset resets run-time counters.
18944 2021-05-26 Martin Liska <mliska@suse.cz>
18946 * doc/install.texi: Port relevant part from install-old.texi
18947 and re-generate list of CPUs and systems.
18949 2021-05-26 Martin Liska <mliska@suse.cz>
18951 * Makefile.in: Remove it.
18952 * doc/include/fdl.texi: Update next/previous chapters.
18953 * doc/install.texi: Likewise.
18954 * doc/install-old.texi: Removed.
18956 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18958 * config/csky/csky.c (ck810_legitimate_index_p): Support
18959 "base + index" with DF mode.
18960 * config/csky/constraints.md ("Y"): New constraint for memory operands
18961 without index register.
18962 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
18963 when mov between memory and general registers, and lower their priority.
18964 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
18966 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18968 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
18970 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18972 * config/csky/csky.md (untyped_call): Emit clobber for return
18973 registers to mark them used.
18975 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18977 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
18979 2021-05-26 Andrew Pinski <apinski@marvell.com>
18981 * match.pd (x < 0 ? ~y : y): New patterns.
18983 2021-05-26 Andrew Pinski <apinski@marvell.com>
18985 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
18986 A?POW2:0 and A?0:POW2.
18988 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18990 * gimple-range-gori.cc (class logical_stmt_cache): Delete
18991 (logical_stmt_cache::logical_stmt_cache ): Delete.
18992 (logical_stmt_cache::~logical_stmt_cache): Delete.
18993 (logical_stmt_cache::cache_entry::dump): Delete.
18994 (logical_stmt_cache::get_range): Delete.
18995 (logical_stmt_cache::cached_name ): Delete.
18996 (logical_stmt_cache::same_cached_name): Delete.
18997 (logical_stmt_cache::cacheable_p): Delete.
18998 (logical_stmt_cache::slot_diagnostics ): Delete.
18999 (logical_stmt_cache::dump): Delete.
19000 (gori_compute_cache::gori_compute_cache): Delete.
19001 (gori_compute_cache::~gori_compute_cache): Delete.
19002 (gori_compute_cache::compute_operand_range): Delete.
19003 (gori_compute_cache::cache_stmt): Delete.
19004 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
19006 (class gori_compute_cache): Delete.
19008 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19010 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
19012 (fold_using_range::range_of_address): Adjust.
19013 (fold_using_range::range_of_phi): Adjust.
19014 * gimple-range.h (class fur_source): Adjust.
19015 (fur_source::fur_source): Adjust.
19017 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19019 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
19020 from expr_range_in_bb and adjust.
19021 (gori_compute::compute_name_range_op): Adjust.
19022 (gori_compute::optimize_logical_operands): Adjust.
19023 (gori_compute::compute_logical_operands_in_chain): Adjust.
19024 (gori_compute::compute_operand1_range): Adjust.
19025 (gori_compute::compute_operand2_range): Adjust.
19026 (ori_compute_cache::cache_stmt): Adjust.
19027 * gimple-range-gori.h (gori_compute): Rename prototype.
19029 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19031 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
19032 checked only after range_of_stmt, not range_on_entry.
19033 (gimple_ranger::range_on_entry): Check for non-null in any
19034 predecessor block, if it is not already non-null.
19035 (gimple_ranger::range_on_exit): DOnt check for non-null after
19036 range on entry call.
19037 (gimple_ranger::dump_bb): New. Split from dump.
19038 (gimple_ranger::dump): Adjust.
19039 * gimple-range.h (class gimple_ranger): Adjust.
19041 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19043 * gimple-range-cache.cc (struct range_timestamp): Delete.
19044 (class temporal_cache): Adjust.
19045 (temporal_cache::get_timestamp): Delete.
19046 (temporal_cache::set_dependency): Delete.
19047 (temporal_cache::temporal_value): Adjust.
19048 (temporal_cache::current_p): Take dependencies as params.
19049 (temporal_cache::set_timestamp): Adjust.
19050 (temporal_cache::set_always_current): Adjust.
19051 (ranger_cache::get_non_stale_global_range): Adjust.
19052 (ranger_cache::register_dependency): Delete.
19053 * gimple-range-cache.h (class range_cache): Adjust.
19055 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19057 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
19059 (range_def_chain::~range_def_chain): Dispose of obstack rather than
19060 each individual bitmap.
19061 (range_def_chain::set_import): New.
19062 (range_def_chain::get_imports): New.
19063 (range_def_chain::chain_import_p): New.
19064 (range_def_chain::register_dependency): Rename from build_def_chain
19066 (range_def_chain::def_chain_in_bitmap_p): New.
19067 (range_def_chain::add_def_chain_to_bitmap): New.
19068 (range_def_chain::has_def_chain): Just check first depenedence.
19069 (range_def_chain::get_def_chain): Process imports, use generic
19070 register_dependency routine.
19071 (range_def_chain::dump): New.
19072 (gori_map::gori_map): Allocate import list.
19073 (gori_map::~gori_map): Release imports.
19074 (gori_map::exports): Check for past allocated block size.
19075 (gori_map::imports): New.
19076 (gori_map::def_chain_in_export_p): Delete.
19077 (gori_map::is_import_p): New.
19078 (gori_map::maybe_add_gori): Handle imports.
19079 (gori_map::dump): Adjust output, add imports.
19080 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
19081 (gori_export_iterator::gori_export_iterator): New.
19082 (gori_export_iterator::next): New.
19083 (gori_export_iterator::get_name): New.
19084 * gimple-range-gori.h (range_def_chain): Add imports and direct
19085 dependecies via struct rdc.
19086 (range_def_chain::depend1): New.
19087 (range_def_chain::depend2): New.
19088 (class gori_map): Adjust.
19089 (FOR_EACH_GORI_IMPORT_NAME): New.
19090 (FOR_EACH_GORI_EXPORT_NAME): New.
19091 (class gori_export_iterator): New.
19093 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19095 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
19096 export cache filling to here.
19097 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
19099 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19101 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
19102 (gori_map): Move to gimple-range-gori.h.
19103 (gori_compute::gori_compute): Adjust.
19104 (gori_compute::~gori_compute): Delete.
19105 (gori_compute::compute_operand_range_switch): Adjust.
19106 (gori_compute::compute_operand_range): Adjust.
19107 (gori_compute::compute_logical_operands): Adjust.
19108 (gori_compute::has_edge_range_p ): Adjust.
19109 (gori_compute::set_range_invariant): Delete.
19110 (gori_compute::dump): Adjust.
19111 (gori_compute::outgoing_edge_range_p): Adjust.
19112 * gimple-range-gori.h (class range_def_chain): Relocate here.
19113 (class gori_map): Relocate here.
19114 (class gori_compute): Inherit from gori_map, and adjust.
19116 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
19118 * value-range.cc (range_tests_legacy): Use
19119 build_nonstandard_integer_type instead of int and short.
19121 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
19123 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
19124 when really creating an initialization statement for it.
19126 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
19128 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
19130 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
19132 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
19134 2021-05-25 Martin Liska <mliska@suse.cz>
19136 PR tree-optimization/92860
19138 * optc-save-gen.awk: Remove exceptions.
19140 2021-05-25 Martin Liska <mliska@suse.cz>
19142 * asan.h (sanitize_coverage_p): New function.
19143 * doc/extend.texi: Document it.
19144 * fold-const.c (fold_range_test): Use sanitize_flags_p
19145 instead of flag_sanitize_coverage.
19146 (fold_truth_andor): Likewise.
19147 * sancov.c: Likewise.
19148 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
19149 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
19150 -fsanitize-coverage when inlining.
19152 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19154 * config/csky/csky-modes.def : Fix copyright.
19156 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19158 * config/csky/csky-modes.def : Amend copyright.
19159 * config/csky/csky_insn_fpuv2.md : Likewise.
19160 * config/csky/csky_insn_fpuv3.md : Likewise.
19162 2021-05-25 Richard Biener <rguenther@suse.de>
19164 PR middle-end/100727
19165 * calls.c (initialize_argument_information): Explicitely test
19166 for WITH_SIZE_EXPR.
19167 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
19169 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19171 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
19172 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
19173 the signle definition. The signle definition may not work well
19174 at simplify_subreg_regno().
19175 (HARD_FRAME_POINTER_REGNUM): New.
19176 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
19177 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
19178 csky_initial_elimination_offset, csky_expand_prologue,
19179 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
19181 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19183 * config/csky/csky.c (csky_option_override):
19184 Init csky_arch_isa_features[] in advance, so TARGET_DSP
19185 and TARGET_DIV can be set well.
19187 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19189 * config/csky/constraints.md ("l", "h"): Delete.
19190 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
19191 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
19192 * config/csky/csky.c (regno_reg_classm,
19193 csky_secondary_reload, csky_register_move_cost):
19194 Use HILO_REGS instead of LO_REGS and HI_REGS.
19196 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19198 * config/csky/constraints.md ("W"): New constriant for mem operand
19199 with base reg, index register.
19200 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
19201 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
19203 ("Dv"): New constraint for const double value that can be used at
19205 * config/csky/csky-modes.def (HFmode): New mode.
19206 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
19207 to "csky_valid_mem_constraint_operand" and support new constraint
19209 (csky_get_movedouble_length): New.
19210 (fpuv3_output_move): New.
19211 (fpuv3_const_double): New.
19212 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
19213 (decompose_csky_address): Refine.
19214 (csky_print_operand): New "CONST_DOUBLE" operand.
19215 (csky_output_move): Support fpv3 instructions.
19216 (csky_get_movedouble_length): New.
19217 (fpuv3_output_move): New.
19218 (fpuv3_const_double): New.
19219 (csky_emit_compare): Cover float comparsion.
19220 (csky_emit_compare_float): Refine.
19221 (csky_vaild_fpuv2_mem_operand): Rename to
19222 "csky_valid_mem_constraint_operand" and support new constraint "W".
19223 (ck860_rtx_costs): New.
19224 (csky_rtx_costs): Add the cost calculation of CK860.
19225 (regno_reg_class): New vregs for fpuv3.
19226 (csky_dbx_regno): Likewise.
19227 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
19228 (csky_conditional_register_usage): Suporrot fpuv3.
19229 (csky_dwarf_register_span): Suporrot fpuv3.
19230 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
19231 (ck810_legitimate_index_p): Support fp16.
19232 * config/csky/csky.h (TARGET_TLS): ADD CK860.
19233 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
19234 (TARGET_SINGLE_FPU): Support fpuv3.
19235 (TARGET_SUPPORT_FPV3): New.
19236 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
19237 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
19238 REG_CLASS_CONTENTS): Support fpuv3.
19239 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
19240 (csky_movsf_fpv2): Likewise.
19241 (ck801_movsf): Likewise.
19242 (csky_movsf): Likewise.
19244 (csky_movdf_fpv2): Likewise.
19245 (ck801_movdf): Likewise.
19246 (csky_movdf): Likewise.
19247 (movsicc): Refine. Use "comparison_operatior" instead of
19248 "ordered_comparison_operatior".
19249 (addsicc): Likewise.
19250 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
19251 (call_value_internal_vh): New.
19252 * config/csky/csky_cores.def (CK860): New arch and cpu.
19257 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
19258 into emit-patterns and match-patterns, remain the emit-patterns here,
19259 and move the match-patterns to csky_insn_fpuv2.md or
19260 csky_insn_fpuv3.md.
19261 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
19262 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
19264 * config/csky/csky_isa.def (fcr): New.
19269 (CK860): New definition for ck860.
19270 * config/csky/csky_tables.opt (ck860): New processors ck860,
19271 ck860f. And new arch ck860.
19276 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
19277 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
19278 * config/csky/t-csky-elf: Support 860.
19279 * config/csky/t-csky-linux: Likewise.
19280 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
19282 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
19284 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
19285 add generation of logical-add and add-logical fusion pairs.
19286 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
19288 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
19289 logical-add and add-logical fusion by default.
19290 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
19291 -mpower10-fusion-add-logical options.
19292 * config/rs6000/fusion.md: Regenerate file.
19294 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
19296 * value-range.cc (irange::legacy_equal_p): Check type when
19297 comparing VR_VARYING types.
19298 (range_tests_legacy): Test comparing VARYING ranges of different
19301 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19303 * config/aarch64/aarch64.c (neoversen1_tunings):
19304 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
19306 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19308 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
19309 extern weak symbols. Limit symbol offsets for non-GOT symbols with
19312 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19314 * config/arm/neon.md (vec_load_lanesxi<mode>)
19315 (vec_store_lanexoi<mode>): Move ...
19316 * config/arm/vec-common.md: here.
19318 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19320 * config/arm/neon.md (vec_load_lanesoi<mode>)
19321 (vec_store_lanesoi<mode>): Move ...
19322 * config/arm/vec-common.md: here.
19324 2021-05-24 liuhongt <hongtao.liu@intel.com>
19327 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
19328 stmt with GIMPLE_NOP when lhs doesn't exist.
19330 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
19333 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
19334 New instruction pattern.
19335 (*push<VI_32:mode>2): Ditto.
19336 (push splitter for SSE registers): New splitter.
19338 2021-05-23 Andrew Pinski <apinski@marvell.com>
19340 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
19342 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
19344 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
19345 * config/rs6000/fusion.md: Regenerate file.
19347 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19349 * config/rs6000/genfusion.pl (gen_addadd): New function.
19350 * config/rs6000/fusion.md: Regenerate file.
19351 * config/rs6000/rs6000-cpus.def: Add
19352 OPTION_MASK_P10_FUSION_2ADD to masks.
19353 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19354 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
19355 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
19357 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19359 PR middle-end/99928
19360 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
19361 * gimplify.c (enum gimplify_omp_var_data): Fix up
19362 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
19363 (omp_lastprivate_for_combined_outer_constructs): If combined target
19364 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
19365 GOVD_MAP | GOVD_SEEN.
19366 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
19367 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
19368 (gimplify_adjust_omp_clauses): For firstprivate clauses with
19369 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
19370 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
19371 let it be replaced by implicit map clause.
19373 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19375 PR middle-end/99928
19376 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
19378 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
19379 (gimplify_omp_for): Likewise.
19381 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19383 PR middle-end/90115
19384 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
19385 'external' in blocks.
19387 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19389 PR middle-end/90115
19390 * flag-types.h (enum openacc_privatization): New.
19391 * params.opt (-param=openacc-privatization): New.
19392 * doc/invoke.texi (openacc-privatization): Document it.
19393 * omp-general.h (get_openacc_privatization_dump_flags): New
19395 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
19396 * omp-offload.c (execute_oacc_device_lower)
19397 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
19398 * target.def (goacc.adjust_private_decl): Add 'location_t'
19400 * doc/tm.texi: Regenerate.
19401 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
19402 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
19403 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
19404 Likewise. Preserve it for...
19405 (nvptx_goacc_expand_var_decl): ... use here.
19407 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19409 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
19412 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19414 PR middle-end/90115
19415 * omp-low.c (oacc_privatization_candidate_p): New function.
19416 (oacc_privatization_scan_clause_chain)
19417 (oacc_privatization_scan_decl_chain): Use it. Also
19418 'gcc_checking_assert' that we're not seeing duplicates.
19420 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19422 PR middle-end/90115
19423 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
19426 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19428 PR middle-end/90115
19429 * omp-offload.c (execute_oacc_device_lower): Explain.
19431 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19433 PR middle-end/90115
19434 * omp-offload.c (execute_oacc_device_lower)
19435 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
19437 * internal-fn.c (expand_UNIQUE): Don't expect
19438 'IFN_UNIQUE_OACC_PRIVATE'.
19440 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19442 PR middle-end/90115
19443 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
19445 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19447 PR middle-end/90115
19448 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
19449 (nvptx_goacc_expand_var_decl): Tighten.
19451 2021-05-21 Julian Brown <julian@codesourcery.com>
19452 Chung-Lin Tang <cltang@codesourcery.com>
19453 Thomas Schwinge <thomas@codesourcery.com>
19455 PR middle-end/90115
19456 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
19457 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
19458 * doc/tm.texi: Regenerate.
19459 * expr.c (expand_expr_real_1): Expand decls using the
19460 expand_var_decl OpenACC hook if defined.
19461 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
19462 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
19463 * omp-low.c (omp_context): Add oacc_privatization_candidates
19465 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
19467 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
19468 private marker's gimple call arguments, and pass it to
19469 lower_oacc_reductions.
19470 (oacc_privatization_scan_clause_chain)
19471 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
19473 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
19474 * omp-offload.c (convert.h): Include.
19475 (oacc_loop_xform_head_tail): Treat private-variable markers like
19476 fork/join when transforming head/tail sequences.
19477 (struct var_decl_rewrite_info): Add struct.
19478 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
19479 (execute_oacc_device_lower): Support rewriting gang-private
19480 variables using target hook, and fix up addr_expr and var_decl
19482 * target.def (adjust_private_decl, expand_var_decl): New hooks.
19483 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
19485 (gcn_goacc_adjust_private_decl): ...this.
19486 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
19488 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
19489 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
19490 definition using gcn_goacc_adjust_gangprivate_decl...
19491 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
19492 gcn_goacc_adjust_private_decl.
19493 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
19494 (gang_private_shared_size): New global variable.
19495 (gang_private_shared_align): Likewise.
19496 (gang_private_shared_sym): Likewise.
19497 (gang_private_shared_hmap): Likewise.
19498 (nvptx_option_override): Initialize these.
19499 (nvptx_file_end): Output gang_private_shared_sym.
19500 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
19502 (nvptx_set_current_function): Clear gang_private_shared_hmap.
19503 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
19504 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
19506 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
19508 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
19510 2021-05-21 Richard Biener <rguenther@suse.de>
19511 H.J. Lu <hjl.tools@gmail.com>
19513 PR middle-end/90773
19514 * expr.c (expand_constructor): Elide expand_constructor if
19515 move by pieces is preferred.
19517 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19519 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
19520 Take a flag and mode value as arguments.
19521 (aarch64_modifies_global_state_p): Likewise.
19522 (aarch64_reads_global_state_p): Likewise.
19523 (aarch64_could_trap_p): Likewise.
19524 (aarch64_get_attributes): Likewise.
19525 (aarch64_init_simd_builtins): Adjust callsite of above.
19526 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
19527 function attributes to apply to builtins.
19528 (aarch64_init_crc32_builtins): Likewise.
19529 (aarch64_init_builtin_rsqrt): Likewise.
19531 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19533 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
19534 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
19535 (gen_2logical): Use new fusion types.
19536 * config/rs6000/fusion.md: Regenerate.
19538 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19541 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
19542 Handle V4QI and V2HI modes.
19543 (ix86_expand_sse_movcc): Ditto.
19544 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
19545 New instruction pattern.
19546 (*eq<VI_32:mode>3): Ditto.
19547 (*gt<VI_32:mode>3): Ditto.
19548 (*xop_pcmov_<VI_32:mode>): Ditto.
19549 (mmx_pblendvb32): Ditto.
19550 (mmx_pblendvb64): Rename from mmx_pblendvb.
19551 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
19552 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
19553 (vcond<VI_32:mode><VI_32:mode>): Ditto.
19554 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
19555 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
19557 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19559 PR tree-optimization/94589
19560 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
19561 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
19562 not the same as cmp3) treat <= the same as < and >= the same as >.
19563 Don't require that cond2_phi_edge is true edge, instead take
19564 false/true edges into account based on cmp1/cmp2 comparison kinds.
19566 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19569 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
19570 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
19571 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
19572 using SMAXMIN_MMXMODEI mode iterator.
19573 (*<smaxmin:code>v4qi3): New insn pattern.
19574 (*<smaxmin:code>v2hi3): Ditto.
19575 (SMAXMIN_VI_32): New mode iterator.
19576 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
19577 (UMAXMIN_MMXMODEI): New mode iterator.
19578 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
19579 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
19580 using UMAXMIN_MMXMODEI mode iterator.
19581 (*<umaxmin:code>v4qi3): New insn pattern.
19582 (*<umaxmin:code>v2hi3): Ditto.
19583 (UMAXMIN_VI_32): New mode iterator.
19584 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
19585 (abs<VI_32:mode>2): New insn pattern.
19586 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
19587 * config/i386/sse.md: ... here.
19589 2021-05-20 Clement Chigot <clement.chigot@atos.net>
19590 David Edelsohn <dje.gcc@gmail.com>
19592 * collect2.c (scan_prog_file): Issue non-fatal warning for
19595 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
19597 * doc/invoke.texi (-Wno-c++11-extensions)
19598 (-Wno-c++14-extensions, -Wno-c++17-extensions)
19599 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
19602 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
19604 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
19605 * config/darwin.c (darwin_override_options): Likewise.
19606 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
19607 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
19608 (DWARF2_FRAME_REG_OUT): Likewise.
19609 * config/mips/mips.c (mips_output_filename): Likewise.
19610 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
19612 (rs6000_dbx_register_number): Likewise.
19613 * dbxout.c: Include flags.h.
19614 * dwarf2cfi.c (cfi_label_required_p): Likewise.
19615 (dwarf2out_do_frame): Likewise.
19616 * except.c: Include flags.h.
19617 * final.c (dwarf2_debug_info_emitted_p): Likewise.
19618 (final_scan_insn_1): Likewise.
19619 * flags.h (dwarf_debuginfo_p): New function declaration.
19620 * opts.c (dwarf_debuginfo_p): New function definition.
19621 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
19622 * toplev.c (process_options): Likewise.
19624 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
19626 * common.opt: Change type to support bitmasks.
19627 * flag-types.h (enum debug_info_type): Rename enumerator constants.
19628 (NO_DEBUG): New bitmask.
19629 (DBX_DEBUG): Likewise.
19630 (DWARF2_DEBUG): Likewise.
19631 (XCOFF_DEBUG): Likewise.
19632 (VMS_DEBUG): Likewise.
19633 (VMS_AND_DWARF2_DEBUG): Likewise.
19634 * flags.h (debug_set_to_format): New function declaration.
19635 (debug_set_count): Likewise.
19636 (debug_set_names): Likewise.
19637 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
19638 (debug_set_to_format): New function definition.
19639 (debug_set_count): Likewise.
19640 (debug_set_names): Likewise.
19641 (set_debug_level): Update access to debug_type_names.
19642 * toplev.c: Likewise.
19644 2021-05-20 Martin Sebor <msebor@redhat.com>
19646 PR middle-end/100684
19647 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
19649 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
19652 * config/i386/i386.md (isa): Remove x64_bmi.
19653 (enabled): Remove x64_bmi.
19654 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
19655 Remove general register alternative.
19656 (*andnot<VI_32:mode>3): Ditto.
19657 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
19658 (*<any_logic:code><VI_32:mode>3): Ditto.
19660 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
19662 * config/arm/arm.c: Include head files tree-vectorizer.h and
19665 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
19668 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
19669 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
19670 (<s>mulv4hi3_highpart): New expander.
19671 (*<s>mulv2hi3_highpart): New insn pattern.
19672 (<s>mulv2hi3_higpart): New expander.
19673 (*<any_shift:insn>v2hi3): New insn pattern.
19674 (<any_shift:insn>v2hi3): New expander.
19675 * config/i386/sse.md (smulhrsv2hi3): New expander.
19676 (*smulhrsv2hi3): New insn pattern.
19678 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
19680 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
19682 * params.opt (vect-inner-loop-cost-factor): New.
19683 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
19684 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
19685 tree-vectorizer.h and its required ones.
19686 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
19687 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
19688 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
19689 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
19690 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
19691 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19693 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
19694 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
19695 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
19697 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
19698 Torbjörn Svensson <torbjorn.svensson@st.com>
19701 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
19703 2021-05-20 Jakub Jelinek <jakub@redhat.com>
19705 PR middle-end/99928
19706 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
19707 explicit linear clause when combined with target, make it map(tofrom:)
19708 instead of no clause or firstprivate.
19710 2021-05-20 Jakub Jelinek <jakub@redhat.com>
19712 PR tree-optimization/94589
19713 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
19714 of integral conversions.
19716 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
19718 * gimple-range.cc (fur_source::get_operand): New.
19719 (gimple_range_fold): Delete.
19720 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
19721 (fold_using_range::range_of_range_op): Move from gimple_ranger.
19722 (fold_using_range::range_of_address): Ditto.
19723 (fold_using_range::range_of_phi): Ditto.
19724 (fold_using_range::range_of_call): Ditto.
19725 (fold_using_range::range_of_builtin_ubsan_call): Move from
19726 range_of_builtin_ubsan_call.
19727 (fold_using_range::range_of_builtin_call): Move from
19728 range_of_builtin_call.
19729 (gimple_ranger::range_of_builtin_call): Delete.
19730 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
19731 (gimple_ranger::fold_range_internal): New.
19732 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
19733 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
19734 gimple_ranger. Improve ranges of SSA_NAMES when possible.
19735 * gimple-range.h (gimple_ranger): Remove various range_of routines.
19736 (class fur_source): New.
19737 (class fold_using_range): New.
19738 (fur_source::fur_source): New.
19740 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
19741 instead of range_of_builtin_call.
19743 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
19745 * doc/cpp.texi (Common Predefined Macros): Update documentation
19746 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
19748 2021-05-19 Alex Coplan <alex.coplan@arm.com>
19751 * config/arm/arm.md (nonsecure_call_internal): Always ensure
19752 callee's address is in a register.
19754 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
19756 * common/config/riscv/riscv-common.c
19757 (riscv_subset_list::parsing_subset_version): Properly parse the letter
19759 (riscv_subset_list::parse_std_ext,
19760 riscv_subset_list::parse_multiletter_ext): To handle errors generated
19761 in riscv_subset_list::parsing_subset_version.
19763 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19765 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
19766 type attribute in patterns generating XTN(2).
19768 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19770 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
19771 Remove as duplicate of...
19772 (aarch64_xtn<mode>): This.
19773 (aarch64_xtn2<mode>_le): Move position in file.
19774 (aarch64_xtn2<mode>_be): Move position in file.
19775 (aarch64_xtn2<mode>): Move position in file.
19776 (vec_pack_trunc_<mode>): Define as an expander.
19778 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19780 * config/aarch64/aarch64-simd-builtins.def: Split builtin
19781 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
19782 separate scalar and vector generators.
19783 * config/aarch64/aarch64-simd.md
19784 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
19786 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
19787 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
19788 * config/aarch64/iterators.md: Define SD_HSDI iterator.
19790 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
19792 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
19794 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
19796 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19798 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
19799 Implement as an expand emitting a big/little endian
19800 instruction pattern.
19801 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
19802 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
19804 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19806 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
19807 Implement as an expand emitting a big/little endian
19808 instruction pattern.
19809 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
19810 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
19811 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
19812 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
19814 2021-05-19 Richard Biener <rguenther@suse.de>
19816 PR middle-end/100672
19817 * fold-const.c (fold_negate_expr_1): Use element_precision.
19818 (negate_expr_p): Likewise.
19820 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
19822 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
19823 (pred_load): New int attribute.
19824 * config/aarch64/aarch64-sve.md
19825 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
19826 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
19827 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
19828 code_for_aarch64_load.
19830 2021-05-19 Richard Biener <rguenther@suse.de>
19832 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
19833 sure TARGET_MEM_REF bases are expanded as memory.
19834 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
19835 Do not mark TARGET_MEM_REF bases addressable.
19836 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
19837 TARGET_MEM_REF bases as never rewritable.
19838 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
19839 walk TARGET_MEM_REF bases as address-takens.
19840 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
19842 2021-05-19 Richard Biener <rguenther@suse.de>
19844 * builtins.c (get_object_alignment_1): Strip outer
19846 * tree-dfa.c (get_ref_base_and_extent): Handle outer
19847 WITH_SIZE_EXPR for size processing and process the
19849 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
19850 outer WITH_SIZE_EXPR.
19851 (ao_ref_base_alias_ptr_type): Likewise.
19852 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
19853 and handle that accordingly, stripping it for the
19854 core alias workers.
19855 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
19856 looking through it instead of returning NULL.
19858 2021-05-19 Jakub Jelinek <jakub@redhat.com>
19860 PR middle-end/100576
19861 * builtins.c (check_read_access): Convert bound to size_type_node if
19864 2021-05-19 Richard Biener <rguenther@suse.de>
19866 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
19867 (verify_types_in_gimple_reference): ... here. Sanitize.
19868 (verify_gimple_call): Verify references in LHS and arguments.
19869 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
19871 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
19873 * config/i386/i386.h (VALID_INT_MODE_P):
19874 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
19875 * config/i386/i386.md (isa): Add x64_bmi.
19876 (enabled): Handle x64_bmi.
19877 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
19878 Add alternative using 64bit general registers.
19879 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
19881 2021-05-19 Jakub Jelinek <jakub@redhat.com>
19883 PR middle-end/99928
19884 * tree.h (OMP_MASTER_COMBINED): Define.
19885 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
19886 handling for outer combined/composite constructs to a loop.
19887 Handle lastprivate on combined target.
19888 (gimplify_expr): Formatting fix.
19890 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
19892 * passes.def: Add sink_code pass before store_merging.
19893 * tree-ssa-sink.c (pass_sink_code:clone): New.
19895 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
19897 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
19898 rs6000_special_adjust_field_align_p.
19899 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
19900 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
19901 Remove ABI warning.
19902 (rs6000_function_arg): Likewise.
19903 * config/rs6000/rs6000-protos.h
19904 (rs6000_special_adjust_field_align_p): Remove prototype.
19905 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
19907 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
19908 rs6000_special_adjust_field_align_p.
19910 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
19913 * config/i386/i386.h (VALID_SSE2_REG_MODE):
19914 Add V4QI and V2HI modes.
19915 (VALID_INT_MODE_P): Ditto.
19916 * config/i386/mmx.md (VI_32): New mode iterator.
19917 (mmxvecsize): Handle V4QI and V2HI.
19919 (mov<VI_32:mode>): New expander.
19920 (*mov<mode>_internal): New insn pattern.
19921 (movmisalign<VI_32:mode>): New expander.
19922 (neg<VI_32:mode>): New expander.
19923 (<plusminus:insn><VI_32:mode>3): New expander.
19924 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
19925 (mulv2hi3): New expander.
19926 (*mulv2hi3): New insn pattern.
19927 (one_cmpl<VI_32:mode>2): New expander.
19928 (*andnot<VI_32:mode>3): New insn pattern.
19929 (<any_logic:code><VI_32:mode>3): New expander.
19930 (*<any_logic:code><VI_32:mode>3): New insn pattern.
19932 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
19934 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
19935 Fix a mode mismatch with operand 1.
19937 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
19940 * config/i386/i386-expand.c (split_double_mode): Return
19941 temporary register when simplify_gen_subreg fails with
19942 the high half od the paradoxical subreg.
19944 2021-05-18 Richard Biener <rguenther@suse.de>
19946 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
19947 and honor it when expanding.
19948 (expand_used_vars_for_block): Pass through forced_stack_var.
19949 (expand_used_vars): Likewise.
19950 (discover_nonconstant_array_refs_r): Set bits in
19951 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
19952 (avoid_type_punning_on_regs): Likewise.
19953 (discover_nonconstant_array_refs): Likewise.
19954 (pass_expand::execute): Create and pass down forced_stack_var
19955 bitmap. For parameters and returns temporarily set
19956 TREE_ADDRESSABLE when expand_function_start.
19958 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
19960 * doc/sourcebuild.texi: Document 'dg-note'.
19962 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
19965 * configure: Regenerate.
19966 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
19968 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
19970 * gimple.h (is_gimple_omp_oacc): Tighten.
19971 * omp-low.c (check_omp_nesting_restrictions): Adjust.
19973 2021-05-18 Richard Biener <rguenther@suse.de>
19975 * tree-ssa-operands.c (mark_address_taken): Simplify.
19977 2021-05-18 Martin Liska <mliska@suse.cz>
19979 * config/gcn/mkoffload.c (STR): Redefine.
19980 * config/i386/intelmic-mkoffload.c (STR): Likewise.
19981 * config/nvptx/mkoffload.c (STR): Likewise.
19983 2021-05-18 Martin Liska <mliska@suse.cz>
19985 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
19986 Use startswith function instead of strncmp.
19987 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
19988 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
19989 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
19990 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
19991 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
19992 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
19993 * config/arm/arm.c (arm_file_start): Likewise.
19994 (arm_valid_target_attribute_rec): Likewise.
19995 (thumb1_md_asm_adjust): Likewise.
19996 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
19997 * config/avr/avr.c (STR_PREFIX_P): Likewise.
19998 (avr_set_current_function): Likewise.
19999 (avr_handle_addr_attribute): Likewise.
20000 (avr_asm_output_aligned_decl_common): Likewise.
20001 (avr_asm_named_section): Likewise.
20002 (avr_section_type_flags): Likewise.
20003 (avr_asm_select_section): Likewise.
20004 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
20005 (c6x_section_type_flags): Likewise.
20006 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
20007 (darwin_objc_declare_unresolved_class_reference): Likewise.
20008 (darwin_objc_declare_class_definition): Likewise.
20009 * config/darwin.c (indirect_data): Likewise.
20010 (darwin_encode_section_info): Likewise.
20011 (darwin_objc2_section): Likewise.
20012 (darwin_objc1_section): Likewise.
20013 (machopic_select_section): Likewise.
20014 (darwin_globalize_label): Likewise.
20015 (darwin_label_is_anonymous_local_objc_name): Likewise.
20016 (darwin_asm_named_section): Likewise.
20017 (darwin_asm_output_dwarf_offset): Likewise.
20018 * config/frv/frv.c (frv_string_begins_with): Likewise.
20019 (frv_in_small_data_p): Likewise.
20020 * config/gcn/mkoffload.c (STR): Likewise.
20022 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
20023 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
20024 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
20025 (ix86_md_asm_adjust): Likewise.
20026 * config/i386/intelmic-mkoffload.c (STR): Likewise.
20027 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
20028 (i386_pe_file_end): Likewise.
20029 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
20030 (ia64_section_type_flags): Likewise.
20031 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
20032 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
20033 (mips16_stub_function_p): Likewise.
20034 (mips_function_rodata_section): Likewise.
20035 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
20036 (msp430_function_section): Likewise.
20037 (msp430_section_type_flags): Likewise.
20038 (msp430_expand_helper): Likewise.
20039 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
20040 (nios2_valid_target_attribute_rec): Likewise.
20041 * config/nvptx/mkoffload.c (process): Likewise.
20043 * config/pa/som.h: Likewise.
20044 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
20045 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
20046 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
20047 (rs6000_inner_target_options): Likewise.
20048 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
20049 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
20050 * config/vax/vax.c (vax_output_int_move): Likewise.
20051 * config/vms/vms-ld.c (startswith): Likewise.
20052 (process_args): Likewise.
20054 * config/vms/vms.c: Likewise.
20056 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20058 PR rtl-optimization/100590
20059 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
20060 they are NONJUMP_INSN_P.
20062 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20065 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
20066 DECL_ASSEMBLER_NAME on the fn_decl.
20068 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20070 PR tree-optimization/94589
20071 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
20072 phi result used in (res & ~1) == 0 comparison as res >= 0 as
20073 res == 2 would be UB with -ffinite-math-only.
20075 2021-05-18 Martin Liska <mliska@suse.cz>
20077 * Makefile.in: genversion.o should depend on DATESTAMP.
20079 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
20081 * config/arc/simdext.md (negv2si2): Remove round bracket.
20083 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
20085 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
20086 _Bool as macro expanding to _Bool.
20088 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
20091 * tree.c (build_reference_type_for_mode)
20092 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
20094 (build_reference_type, build_pointer_type): Invoke
20095 build_*_type_for_mode with VOIDmode.
20097 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
20099 PR tree-optimization/100512
20100 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
20101 and non-zero pointer ranges as invariant.
20102 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
20103 processing from here.
20105 2021-05-17 Tom de Vries <tdevries@suse.de>
20108 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
20109 * config/nvptx/nvptx.c (nvptx_output_barrier)
20110 (nvptx_output_atomic_insn): New function.
20111 (nvptx_print_operand): Add support for 'B'.
20112 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
20115 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
20117 PR tree-optimization/100349
20118 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
20121 2021-05-17 Tamar Christina <tamar.christina@arm.com>
20123 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
20124 (host_detect_local_cpu): Use it.
20126 2021-05-17 Martin Liska <mliska@suse.cz>
20128 * doc/invoke.texi: Add 2 missing dots.
20130 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
20132 PR bootstrap/100552
20133 * configure.ac: Replace pattern substitution with call to sed.
20134 * configure: Regenerate.
20136 2021-05-17 Richard Biener <rguenther@suse.de>
20138 PR middle-end/100582
20139 * tree.c (array_at_struct_end_p): Get to the base of the
20140 reference before looking for the underlying decl.
20142 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
20144 * genoutput.c (validate_insn_alternatives) Make "wrong number of
20145 alternatives" message more specific, and remove assumption on where
20148 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20150 * config/arm/iterators.md (V16): New iterator.
20151 (VH_cvtto): New iterator.
20152 (v_cmp_result): Added V4HF and V8HF support.
20153 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
20154 (vcond<mode><mode>): Likewise.
20155 (vcond_mask_<mode><v_cmp_result>): Likewise.
20156 (vcond<VH_cvtto><mode>): New expander.
20158 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20160 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
20162 * config/arm/arm.c (arm_expand_vector_compare): Add support for
20164 (arm_expand_vcond): Likewise.
20165 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
20166 VCMPEQQ_N_S, VCMPNEQ_N_S.
20167 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
20168 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
20169 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
20170 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
20171 (@mve_vpselq_<supf><mode>): Likewise.
20172 (@mve_vpselq_f<mode>"): Likewise.
20173 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
20174 and move to vec-common.md.
20175 (vec_cmpu<mode><mode>): Likewise.
20176 (vcond<mode><mode>): Likewise.
20177 (vcond<V_cvtto><mode>): Likewise.
20178 (vcondu<mode><v_cmp_result>): Likewise.
20179 (vcond_mask_<mode><v_cmp_result>): Likewise.
20180 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
20181 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
20182 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
20183 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
20184 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
20185 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
20187 (vec_cmpu<mode><mode>): Likewise.
20188 (vcond<mode><mode>): Likewise.
20189 (vcond<V_cvtto><mode>): Likewise.
20190 (vcondu<mode><v_cmp_result>): Likewise.
20191 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
20194 2021-05-17 liuhongt <hongtao.liu@intel.com>
20197 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
20198 gsi_insert_seq_before instead.
20200 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20202 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
20203 (arm_sat_ok): ...this.
20205 2021-05-17 Martin Liska <mliska@suse.cz>
20207 * lto-wrapper.c (merge_flto_options): Factor out a new function.
20208 (merge_and_complain): Use it.
20209 (run_gcc): Merge also linker command line -flto=foo argument
20212 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
20214 * config/arm/arm.h (CPP_SPEC): Remove error message about
20215 -mlittle-endian/-mbig-endian conflict.
20217 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20219 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
20220 __ROP_PROTECT__ if -mrop-protect is selected.
20222 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20224 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
20225 rop_hash_save_offset and rop_hash_size.
20226 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
20227 rop_hash_size and rop_hash_save_offset.
20228 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
20229 (rs6000_emit_prologue): Emit hashst[p] in prologue.
20230 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
20231 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
20233 (hashst): New define_insn.
20234 (hashchk): Likewise.
20236 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20238 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20239 Disable shrink wrap when inserting ROP-protect instructions.
20240 * config/rs6000/rs6000.opt (mrop-protect): New option.
20241 (mprivileged): Likewise.
20242 * doc/invoke.texi: Document mrop-protect and mprivileged.
20244 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
20246 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
20249 2021-05-15 Martin Jambor <mjambor@suse.cz>
20252 2021-05-13 Martin Jambor <mjambor@suse.cz>
20254 PR tree-optimization/100453
20255 * tree-sra.c (sra_modify_assign): All const base accesses do not
20256 need refreshing, not just those from decl_pool.
20257 (sra_modify_assign): Do not refresh into a const base decl.
20259 2021-05-15 Jakub Jelinek <jakub@redhat.com>
20261 PR rtl-optimization/100342
20262 * regcprop.c (copy_value): When copying a source reg in a wider
20263 mode than it has recorded for the value, adjust recorded destination
20264 mode too or punt if !REG_CAN_CHANGE_MODE_P.
20266 2021-05-14 Jason Merrill <jason@redhat.com>
20268 * intl.h: Add comments.
20270 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20272 * config/aarch64/aarch64-simd.md
20273 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
20274 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
20275 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
20276 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
20277 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
20278 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
20279 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
20280 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
20281 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
20283 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20286 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
20287 boolean logic equivalent.
20288 (vtst_s16): Likewise.
20289 (vtst_s32): Likewise.
20290 (vtst_u8): Likewise.
20291 (vtst_u16): Likewise.
20292 (vtst_u32): Likewise.
20293 (vtst_p8): Likewise.
20294 (vtst_p16): Likewise.
20295 (vtstq_s8): Likewise.
20296 (vtstq_s16): Likewise.
20297 (vtstq_s32): Likewise.
20298 (vtstq_u8): Likewise.
20299 (vtstq_u16): Likewise.
20300 (vtstq_u32): Likewise.
20301 (vtstq_p8): Likewise.
20302 (vtstq_p16): Likewise.
20303 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
20304 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
20306 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20308 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
20309 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
20310 (aarch64_sqdmlsl2<mode>): Delete.
20311 (aarch64_sqdmlal2_lane<mode>): Merge this...
20312 (aarch64_sqdmlsl2_lane<mode>): ... And this...
20313 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
20314 (aarch64_sqdmlal2_laneq<mode>): Merge this...
20315 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
20316 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
20317 (aarch64_sqdmlal2_n<mode>): Merge this...
20318 (aarch64_sqdmlsl2_n<mode>): ... And this...
20319 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
20321 2021-05-13 Martin Sebor <msebor@redhat.com>
20323 PR middle-end/100574
20324 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
20325 all null arguments.
20327 2021-05-13 Martin Sebor <msebor@redhat.com>
20329 PR tree-optimization/93100
20330 PR middle-end/98583
20331 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
20332 don't modify referenced objects.
20334 2021-05-13 Martin Jambor <mjambor@suse.cz>
20336 PR tree-optimization/100453
20337 * tree-sra.c (sra_modify_assign): All const base accesses do not
20338 need refreshing, not just those from decl_pool.
20339 (sra_modify_assign): Do not refresh into a const base decl.
20341 2021-05-13 Martin Liska <mliska@suse.cz>
20343 * tree-ssa-dom.c: Remove m_simplifier.
20345 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
20348 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
20349 canonicalize DImode inequality comparisons against the
20350 maximum integral value.
20352 2021-05-13 Jakub Jelinek <jakub@redhat.com>
20354 PR tree-optimization/98856
20355 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
20356 Expect V2DI and V4DI arithmetic right shifts to be emulated.
20357 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
20359 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
20360 expand_vec_perm_2perm_pblendv): New functions.
20361 (ix86_expand_vec_perm_const_1): Use them.
20362 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
20363 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
20364 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
20365 (ashrv4di3): New define_expand.
20366 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
20367 and !TARGET_AVX512VL expansion.
20369 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
20372 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
20373 sizes < 16 to a register when constructing vpcmov pattern.
20374 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
20376 2021-05-13 Martin Liska <mliska@suse.cz>
20378 * gcov-io.c (gcov_write_block): Remove.
20379 (gcov_write_words): Likewise.
20380 (gcov_read_words): Re-implement using gcov_read_bytes.
20381 (gcov_allocate): Remove.
20382 (GCOV_BLOCK_SIZE): Likewise.
20383 (struct gcov_var): Remove most of the fields.
20384 (gcov_position): Implement with ftell.
20385 (gcov_rewrite): Remove setting of start and offset fields.
20386 (from_file): Re-format.
20387 (gcov_open): Remove setbuf call. It should not be needed.
20388 (gcov_close): Remove internal buffer handling.
20389 (gcov_magic): Use __builtin_bswap32.
20390 (gcov_write_counter): Use directly gcov_write_unsigned.
20391 (gcov_write_string): Use direct fwrite and do not round
20393 (gcov_seek): Use directly fseek.
20394 (gcov_write_tag): Use gcov_write_unsigned directly.
20395 (gcov_write_length): Likewise.
20396 (gcov_write_tag_length): Likewise.
20397 (gcov_read_bytes): Use directly fread.
20398 (gcov_read_unsigned): Use gcov_read_words.
20399 (gcov_read_counter): Likewise.
20400 (gcov_read_string): Use gcov_read_bytes.
20401 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
20402 that size is not in bytes, but words (4B).
20403 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
20404 (GCOV_TAG_ARCS_LENGTH): Likewise.
20405 (GCOV_TAG_ARCS_NUM): Likewise.
20406 (GCOV_TAG_COUNTER_LENGTH): Likewise.
20407 (GCOV_TAG_COUNTER_NUM): Likewise.
20408 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
20410 2021-05-13 liuhongt <hongtao.liu@intel.com>
20413 * config/i386/sse.md (ssedoublevecmode): Add attribute for
20414 V64QI/V32HI/V16SI/V4DI.
20415 (ssehalfvecmode): Add attribute for V2DI/V2DF.
20416 (*vec_concatv4si_0): Extend to VI124_128.
20417 (*vec_concat<mode>_0): New pre-reload splitter.
20418 * config/i386/predicates.md (movq_parallel): New predicate.
20420 2021-05-13 Alexandre Oliva <oliva@adacore.com>
20422 * targhooks.c (default_zero_call_used_regs): Retry using
20423 successfully-zeroed registers as sources.
20425 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
20427 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
20429 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
20432 * gimple-range.cc (range_of_builtin_call): Skip out on
20433 processing __builtin_clz when varying.
20435 2021-05-12 Tom de Vries <tdevries@suse.de>
20438 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
20439 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
20440 to ptx_version_option.
20441 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
20442 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
20443 (define_insn "nvptx_vote_ballot"): Use sync variant for
20445 * config/nvptx/nvptx.opt (ptx_version): Add enum.
20446 (mptx): Add option.
20447 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
20449 2021-05-12 Richard Biener <rguenther@suse.de>
20451 PR tree-optimization/100566
20452 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
20453 allow_back for all edge queries.
20455 2021-05-12 liuhongt <hongtao.liu@intel.com>
20458 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
20459 splitters for pblendvb of NOT mask register.
20461 2021-05-12 Richard Biener <rguenther@suse.de>
20463 PR tree-optimization/100519
20464 * tree-ssa-reassoc.c (can_associate_p): Split into...
20465 (can_associate_op_p): ... this
20466 (can_associate_type_p): ... and this.
20467 (is_reassociable_op): Call can_associate_op_p.
20468 (break_up_subtract_bb): Call the appropriate predicates.
20469 (reassociate_bb): Likewise.
20471 2021-05-12 Martin Liska <mliska@suse.cz>
20473 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
20474 (run_gcc): Use -flto argument detection for merged
20477 2021-05-12 Martin Liska <mliska@suse.cz>
20479 * lto-wrapper.c (print_lto_docs_link): New function.
20480 (run_gcc): Print warning about missing job server detection
20481 after we know NR of partitions. Do the same for -flto{,=1}.
20482 * opts.c (get_option_html_page): Support -flto option.
20484 2021-05-12 Martin Liska <mliska@suse.cz>
20486 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
20488 (append_option): Remove.
20489 (find_option): Rework to use the vector type.
20490 (remove_option): Remove.
20491 (merge_and_complain): Use vectors for cl_decoded_option data
20493 (append_compiler_options): Likewise.
20494 (append_diag_options): Likewise.
20495 (append_linker_options): Likewise.
20496 (append_offload_options): Likewise.
20497 (compile_offload_image): Likewise.
20498 (compile_images_for_offload_targets): Likewise.
20499 (find_and_merge_options): Likewise.
20500 (run_gcc): Likewise.
20502 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
20505 * dwarf2out.c (dwarf2out_finish): Set
20506 have_multiple_function_sections with multi-range text_section.
20508 2021-05-12 Martin Liska <mliska@suse.cz>
20510 PR bootstrap/100560
20511 * Makefile.in: Remove version.h from linker command line.
20513 2021-05-12 Richard Biener <rguenther@suse.de>
20515 PR middle-end/100547
20516 * rtl.h (rtvec_alloc): Make argument size_t.
20517 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
20519 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20521 PR middle-end/100508
20522 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
20523 type, don't reuse DECL_RTL if it has different mode, instead force
20524 creation of a new DEBUG_EXPR.
20526 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20527 Marc Glisse <marc.glisse@inria.fr>
20529 PR tree-optimization/94589
20530 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
20531 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
20533 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
20536 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
20537 * config/i386/mmx.md (MMXMODE124): New mode iterator.
20539 (mmxintvecmode): New mode attribute.
20540 (mmxintvecmodelower): Ditto.
20541 (*mmx_maskcmpv2sf3_comm): New insn pattern.
20542 (*mmx_maskcmpv2sf3): Ditto.
20543 (vec_cmpv2sfv2si): New expander.
20544 (vcond<V2FI:mode>v2si): Ditto.
20545 (mmx_vlendvps): New insn pattern.
20546 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
20547 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
20548 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
20550 2021-05-11 Martin Sebor <msebor@redhat.com>
20552 PR middle-end/21433
20553 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
20555 2021-05-11 Richard Biener <rguenther@suse.de>
20557 * gimple-fold.c (gimple_fold_call): Do not call
20558 maybe_fold_reference on call arguments or the static chain.
20559 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
20562 2021-05-11 Martin Liska <mliska@suse.cz>
20564 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
20565 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
20566 (DEF_HSAIL_SAT_BUILTIN): Likewise.
20567 (DEF_HSAIL_INTR_BUILTIN): Likewise.
20568 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
20569 * doc/frontends.texi: Remove BRIG.
20570 * doc/install.texi: Likewise.
20571 * doc/invoke.texi: Likewise.
20572 * doc/standards.texi: Likewise.
20573 * brig-builtins.def: Removed.
20574 * brig/ChangeLog: Removed.
20575 * brig/Make-lang.in: Removed.
20576 * brig/brig-builtins.h: Removed.
20577 * brig/brig-c.h: Removed.
20578 * brig/brig-lang.c: Removed.
20579 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
20580 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
20581 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
20582 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
20583 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
20584 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
20585 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
20586 * brig/brigfrontend/brig-comment-handler.cc: Removed.
20587 * brig/brigfrontend/brig-control-handler.cc: Removed.
20588 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
20589 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
20590 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
20591 * brig/brigfrontend/brig-function-handler.cc: Removed.
20592 * brig/brigfrontend/brig-function.cc: Removed.
20593 * brig/brigfrontend/brig-function.h: Removed.
20594 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
20595 * brig/brigfrontend/brig-label-handler.cc: Removed.
20596 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
20597 * brig/brigfrontend/brig-machine.c: Removed.
20598 * brig/brigfrontend/brig-machine.h: Removed.
20599 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
20600 * brig/brigfrontend/brig-module-handler.cc: Removed.
20601 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
20602 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
20603 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
20604 * brig/brigfrontend/brig-to-generic.cc: Removed.
20605 * brig/brigfrontend/brig-to-generic.h: Removed.
20606 * brig/brigfrontend/brig-util.cc: Removed.
20607 * brig/brigfrontend/brig-util.h: Removed.
20608 * brig/brigfrontend/brig-variable-handler.cc: Removed.
20609 * brig/brigfrontend/hsa-brig-format.h: Removed.
20610 * brig/brigfrontend/phsa.h: Removed.
20611 * brig/brigspec.c: Removed.
20612 * brig/config-lang.in: Removed.
20613 * brig/gccbrig.texi: Removed.
20614 * brig/lang-specs.h: Removed.
20615 * brig/lang.opt: Removed.
20617 2021-05-11 Richard Biener <rguenther@suse.de>
20620 * ipa-param-manipulation.c
20621 (ipa_param_body_adjustments::modify_call_stmt): Avoid
20622 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
20623 via gimple_call_lhs_ptr.
20625 2021-05-11 Alex Coplan <alex.coplan@arm.com>
20628 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
20629 Avoid emitting CFA adjusts on the sp if we have the fp.
20631 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
20633 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
20634 (VMULD): New iterator.
20635 (VCOND): Handle V4HF and V8HF.
20636 (VCONQ): Fix entry for V2SF.
20637 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
20638 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
20639 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
20640 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
20641 Use a 128-bit vector mode for the indexed operand. Use stype for
20642 the scheduling type.
20644 2021-05-11 Richard Biener <rguenther@suse.de>
20646 * gimple-fold.c (maybe_fold_reference): Only return
20647 is_gimple_min_invariant values.
20649 2021-05-11 Richard Biener <rguenther@suse.de>
20651 PR middle-end/100509
20652 * gimple-fold.c (fold_gimple_assign): Only call
20653 get_symbol_constant_value on register type symbols.
20655 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
20656 Joe Ramsay <joe.ramsay@arm.com>
20659 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
20660 (__arm_vcmpneq): Remove duplicate definition.
20661 (__arm_vstrwq_scatter_offset_p): Likewise.
20662 (__arm_vmaxq_x): Likewise.
20663 (__arm_vmlsdavaq): Likewise.
20664 (__arm_vmlsdavaxq): Likewise.
20665 (__arm_vmlsdavq_p): Likewise.
20666 (__arm_vmlsdavxq_p): Likewise.
20667 (__arm_vrmlaldavhaq): Likewise.
20668 (__arm_vstrbq_p): Likewise.
20669 (__arm_vstrbq_scatter_offset): Likewise.
20670 (__arm_vstrbq_scatter_offset_p): Likewise.
20671 (__arm_vstrdq_scatter_offset): Likewise.
20672 (__arm_vstrdq_scatter_offset_p): Likewise.
20673 (__arm_vstrdq_scatter_shifted_offset): Likewise.
20674 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
20676 2021-05-11 Jakub Jelinek <jakub@redhat.com>
20678 PR middle-end/100471
20679 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
20680 is 0, bypass the reduction loop including
20681 GOMP_taskgroup_reduction_unregister call.
20683 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20685 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
20686 costing_for_scalar.
20687 (rs6000_density_test): Early return if costing_for_scalar is true.
20688 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
20690 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20692 * doc/tm.texi: Regenerated.
20693 * target.def (init_cost): Add new parameter costing_for_scalar.
20694 * targhooks.c (default_init_cost): Adjust for new parameter.
20695 * targhooks.h (default_init_cost): Likewise.
20696 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
20697 (vect_compute_single_scalar_iteration_cost): Likewise.
20698 (vect_analyze_loop_2): Likewise.
20699 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
20700 (vect_bb_vectorization_profitable_p): Likewise.
20701 * tree-vectorizer.h (init_cost): Likewise.
20702 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
20703 * config/i386/i386.c (ix86_init_cost): Likewise.
20704 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
20706 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20708 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
20709 vect_nonmem and moved into...
20710 (struct rs6000_cost_data): ...here.
20711 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
20712 (rs6000_add_stmt_cost): Likewise.
20713 (rs6000_finish_cost): Likewise.
20715 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
20717 * range-op.cc (get_bool_state): Adjust head comment.
20718 (operator_not_equal::op1_range): Fix comment.
20719 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
20721 2021-05-10 Martin Sebor <msebor@redhat.com>
20723 PR middle-end/100425
20724 PR middle-end/100510
20725 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
20726 (pass_walloca::xlimit_certain_p): ...to this.
20727 (pass_walloca::gate): Execute for any kind of handled warning.
20728 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
20729 warnings when xlimit_certain_p is set.
20731 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
20733 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
20734 Return ALTIVEC_REGS if that is best_class.
20735 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
20737 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20739 * config/arm/arm.h (CPP_SPEC): Remove error message about
20742 2021-05-10 Martin Jambor <mjambor@suse.cz>
20744 * ipa-prop.h (IPA_NODE_REF): Removed.
20745 (IPA_NODE_REF_GET_CREATE): Likewise.
20746 (IPA_EDGE_REF): Likewise.
20747 (IPA_EDGE_REF_GET_CREATE): Likewise.
20748 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
20749 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
20750 use of ipa_node_params_sum.
20751 (ipcp_versionable_function_p): Likewise.
20752 (push_node_to_stack): Likewise.
20753 (pop_node_from_stack): Likewise.
20754 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
20755 direct use of ipa_node_params_sum.
20756 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
20757 ipa_node_params_sum.
20758 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
20760 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
20761 use of ipa_node_params_sum.
20762 (self_recursively_generated_p): Likewise.
20763 (propagate_scalar_across_jump_function): Likewise.
20764 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
20765 direct use of ipa_edge_args_sum, moved the lookup after the early
20766 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
20767 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
20768 direct uses of ipa_node_params_sum.
20769 (propagate_vr_across_jump_function): Likewise.
20770 (propagate_aggregate_lattice): Likewise.
20771 (propagate_aggs_across_jump_function): Likewise.
20772 (propagate_constants_across_call): Likewise, also replaced
20773 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
20774 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
20775 of ipa_node_params_sum.
20776 (estimate_local_effects): Likewise.
20777 (add_all_node_vals_to_toposort): Likewise.
20778 (propagate_constants_topo): Likewise.
20779 (ipcp_propagate_stage): Likewise.
20780 (ipcp_discover_new_direct_edges): Likewise.
20781 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
20782 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
20783 (get_info_about_necessary_edges): Likewise.
20784 (want_remove_some_param_p): Likewise.
20785 (create_specialized_node): Likewise.
20786 (self_recursive_pass_through_p): Likewise.
20787 (self_recursive_agg_pass_through_p): Likewise.
20788 (find_more_scalar_values_for_callers_subset): Likewise and also
20789 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
20790 case replacing two of those with a single query.
20791 (find_more_contexts_for_caller_subset): Likewise for the
20792 ipa_polymorphic_call_context overload.
20793 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
20794 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
20795 ipa_node_params_sum.
20796 (find_aggregate_values_for_callers_subset): Likewise, also reusing
20797 results of ipa_edge_args_sum->get.
20798 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
20799 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
20800 direct use of ipa_edge_args_sum.
20801 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
20802 summary query after the early exit and reused the result later.
20803 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
20804 ipa_node_params_sum.
20805 (decide_whether_version_node): Likewise. Removed re-querying for
20806 summaries after cloning.
20807 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
20808 ipa_node_params_sum.
20809 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
20811 (identify_dead_nodes): Likewise.
20812 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
20813 ipa_node_params_sum.
20814 (ipcp_store_vr_results): Likewise.
20815 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
20816 (ipa_fn_summary_t::duplicate): Likewise.
20817 (analyze_function_body): Likewise.
20818 (estimate_calls_size_and_time): Likewise.
20819 (ipa_cached_call_context::duplicate_from): Likewise.
20820 (ipa_call_context::equal_to): Likewise.
20821 (remap_edge_params): Likewise.
20822 (ipa_merge_fn_summary_after_inlining): Likewise.
20823 (inline_read_section): Likewise.
20824 * ipa-icf.c (sem_function::param_used_p): Likewise.
20825 * ipa-modref.c (compute_parm_map): Likewise.
20826 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
20828 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
20829 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
20831 * ipa-profile.c (check_argument_count): Likewise.
20832 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
20833 with a direct use of ipa_node_params_sum.
20834 (ipa_initialize_node_params): Likewise.
20835 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
20836 direct use of ipa_edge_args_sum and reused the query result.
20837 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
20838 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
20839 direct use of ipa_edge_args_sum.
20840 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
20841 ipa_node_params_sum and reused the result of the query.
20842 (ipa_analyze_node): Likewise.
20843 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
20844 of ipa_node_params_sum.
20845 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
20846 direct uses of ipa_edge_args_sum.
20847 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
20848 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
20849 direct use of ipa_edge_args_sum. Removed superficial re-querying the
20851 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
20852 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
20854 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
20855 direct use of ipa_edge_args_sum.
20856 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
20857 use of ipa_node_params_sum.
20858 (ipa_print_node_params): Likewise.
20859 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
20860 direct uses of ipa_edge_args_sum.
20861 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
20863 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
20864 ipa_node_params_sum.
20865 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
20866 scopes where it is used.
20868 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
20870 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
20871 <case E_V2SImode>: Force op_true to register.
20873 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20875 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
20876 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
20877 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
20879 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
20880 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
20881 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
20882 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
20883 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
20884 (mve_vcmpneq_n_f<mode>): Remove.
20885 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
20886 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
20887 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
20889 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20891 * config/arm/iterators.md (MVE_COMPARISONS): New.
20893 (mve_cmp_type): New.
20894 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
20896 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
20897 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
20898 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
20899 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
20900 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
20901 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
20902 (mve_vcmpneq_n_<mode>): Remove.
20904 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20906 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
20907 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
20908 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
20911 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20913 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
20914 (vcmpneq_n_u): Likewise.
20915 (vcmpeqq_u,): Likewise.
20916 (vcmpeqq_n_u): Likewise.
20917 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
20918 VCMPEQQ_N_U and VCMPNEQ_N_U.
20919 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
20920 (mve_vcmpeqq_n): Likewise.
20921 (mve_vcmpeqq): Likewise.
20922 (mve_vcmpneq_n): Likewise.
20924 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20926 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
20927 the 's' version of the builtin.
20929 2021-05-10 Richard Biener <rguenther@suse.de>
20931 PR tree-optimization/100492
20932 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
20933 Find nothing when the loop contains an irreducible region.
20935 2021-05-10 Richard Biener <rguenther@suse.de>
20937 PR middle-end/100464
20939 * gimple-fold.c (canonicalize_constructor_val): Do not set
20942 2021-05-10 Richard Biener <rguenther@suse.de>
20944 PR tree-optimization/100434
20945 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
20947 (dse_optimize_stmt): Handle call LHS by dropping the
20948 LHS or the whole call if it doesn't have other
20950 (pass_dse::execute): Adjust.
20952 2021-05-10 Martin Liska <mliska@suse.cz>
20954 * Makefile.in: Add missing genversion rule.
20956 2021-05-10 Alex Coplan <alex.coplan@arm.com>
20959 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
20960 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
20962 2021-05-10 Martin Liska <mliska@suse.cz>
20964 * builtins.c (is_builtin_name): Use startswith
20965 function instead of strncmp.
20966 * collect2.c (main): Likewise.
20967 (has_lto_section): Likewise.
20968 (scan_libraries): Likewise.
20969 * coverage.c (coverage_checksum_string): Likewise.
20970 (coverage_init): Likewise.
20971 * dwarf2out.c (is_cxx): Likewise.
20972 (gen_compile_unit_die): Likewise.
20973 * gcc-ar.c (main): Likewise.
20974 * gcc.c (init_spec): Likewise.
20975 (read_specs): Likewise.
20976 (execute): Likewise.
20977 (check_live_switch): Likewise.
20978 * genattrtab.c (write_attr_case): Likewise.
20979 (IS_ATTR_GROUP): Likewise.
20980 * gencfn-macros.c (main): Likewise.
20981 * gengtype.c (type_for_name): Likewise.
20982 (gen_rtx_next): Likewise.
20983 (get_file_langdir): Likewise.
20984 (write_local): Likewise.
20985 * genmatch.c (get_operator): Likewise.
20986 (get_operand_type): Likewise.
20987 (expr::gen_transform): Likewise.
20988 * genoutput.c (validate_optab_operands): Likewise.
20989 * incpath.c (add_sysroot_to_chain): Likewise.
20990 * langhooks.c (lang_GNU_C): Likewise.
20991 (lang_GNU_CXX): Likewise.
20992 (lang_GNU_Fortran): Likewise.
20993 (lang_GNU_OBJC): Likewise.
20994 * lto-wrapper.c (run_gcc): Likewise.
20995 * omp-general.c (omp_max_simt_vf): Likewise.
20996 * omp-low.c (omp_runtime_api_call): Likewise.
20997 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
20998 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
20999 * real.c (real_from_string): Likewise.
21000 * selftest.c (assert_str_startswith): Likewise.
21001 * timevar.c (timer::validate_phases): Likewise.
21002 * tree.c (get_file_function_name): Likewise.
21003 * ubsan.c (ubsan_use_new_style_p): Likewise.
21004 * varasm.c (default_function_rodata_section): Likewise.
21005 (incorporeal_function_p): Likewise.
21006 (default_section_type_flags): Likewise.
21007 * system.h (startswith): Define startswith.
21009 2021-05-10 Martin Liska <mliska@suse.cz>
21011 * bitmap.h (class auto_bitmap): Remove
21012 __cplusplus >= 201103.
21013 * config/aarch64/aarch64.c: Likewise.
21014 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
21016 * sbitmap.h: Likewise.
21018 2021-05-10 Martin Liska <mliska@suse.cz>
21020 * Makefile.in: Rename gcov-iov to genversion and depend
21021 on version.h (instead of gcov-iov.h).
21022 * gcov-io.h: Include version.h instread of gcov-iov.h.
21023 * gengtype-state.c (read_state_version): Likewise.
21024 * gcov-iov.c: Moved to...
21025 * genversion.c: ...here.
21026 * lto-streamer.h (LTO_major_version): Define it with
21028 * version.c: Removed.
21029 * version.h: Removed.
21031 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21033 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
21034 * config/arc/simdext.md (VCT): Add predicates for iterator
21037 (voptab): Likewise.
21038 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
21039 (<voptab>v2si3): New patterns.
21041 (reduc_plus_scal_v4hi): Likewise.
21042 (reduc_plus_scal_v2si): Likewise.
21043 (vec_duplicatev2si): Likewise.
21044 (vec_duplicatev4hi): Likewise.
21046 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21048 * config/arc/simdext.md: Format and cleanup file.
21050 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21052 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
21053 only when munaligned-access option is on.
21054 (movmisalign<mode>): Likewise.
21056 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21058 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
21060 * config/arc/arc.c (arc_reorg): Remove underscore from string.
21062 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21064 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21065 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
21066 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
21067 (norm_f): Likewise.
21070 (clzsi2): Use fls instruction when available.
21071 (arc_clzsi2): Likewise.
21073 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21075 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
21077 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21079 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
21081 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
21083 PR middle-end/100467
21084 * toplev.c (compile_file): Call insn_locations_init before
21085 targetm.asm_out.code_end.
21087 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21090 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21092 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21094 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21095 Andrew Stubbs <amd@codesourcery.com>
21098 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
21099 emit_move_insn operands.
21101 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21103 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
21104 location for the outgoing edges of an empty block.
21105 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
21106 type by means of the get_array_descr_info langhook, if it is set and
21107 returns true. Remove obsolete code dealing with unnamed subtypes.
21109 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21111 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
21112 (sbr_vector): Renamed from ssa_block_cache.
21113 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
21114 (ssa_block_ranges::~ssa_block_ranges): Remove.
21115 (sbr_vector::set_bb_range): Use varying and undefined cached values.
21116 (ssa_block_ranges::set_bb_varying): Remove.
21117 (sbr_vector::get_bb_range): Adjust assert.
21118 (sbr_vector::bb_range_p): Adjust assert.
21119 (~block_range_cache): No freeing loop required.
21120 (block_range_cache::get_block_ranges): Remove.
21121 (block_range_cache::set_bb_range): Inline get_block_ranges.
21122 (block_range_cache::set_bb_varying): Remove.
21123 * gimple-range-cache.h (set_bb_varying): Remove prototype.
21124 * value-range.h (irange_allocator::get_memory): New.
21126 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21128 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
21129 dominator tree is available and requested.
21130 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
21131 (ranger_cache::fill_block_cache): Don't search dom tree here either.
21132 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
21134 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21136 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
21137 only PHI nodes better.
21139 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21141 * gimple-range-edge.h (gimple_outgoing_range): Rename from
21143 (gcond_edge_range): Export prototype.
21144 * gimple-range-edge.cc (gcond_edge_range): New.
21145 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
21146 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
21148 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21150 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
21151 default range into a temp and allocate only what is needed.
21153 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21155 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
21157 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21159 * gimple-range.h (gimple_range_global): Pick up parameter initial
21160 values, and use-before defined locals are UNDEFINED.
21162 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21164 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
21166 * tree.h (reverse_storage_order_for_component_p): Return false if
21167 the type is a pointer.
21169 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21171 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21173 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21176 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
21177 Handle V8QI, V4HI and V2SI modes.
21178 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
21179 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
21180 * config/i386/i386.md (unspec): ... here.
21182 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
21183 Tom de Vries <tdevries@suse.de>
21185 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
21186 a truth_value_p reduction variable is nonintegral.
21188 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21191 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
21192 Return false for mode sizes < 16.
21194 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21197 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
21199 2021-05-06 Martin Jambor <mjambor@suse.cz>
21201 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
21202 when there is no function summary.
21203 (ipa_sra_summarize_function): produce edge summaries even when
21206 2021-05-06 Tom Tromey <tom@tromey.com>
21208 * godump.c (string_hash_eq): Remove.
21209 (go_finish): Use htab_eq_string.
21211 2021-05-06 Tom Tromey <tom@tromey.com>
21213 * gengtype-state.c (read_state): Use htab_eq_string.
21214 (string_eq): Remove.
21216 2021-05-06 Tom Tromey <tom@tromey.com>
21218 * gensupport.c (htab_eq_string): Remove.
21220 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
21223 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
21224 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
21225 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
21226 set_ignored_loc callbacks.
21227 * debug.c (do_nothing_debug_hooks): Likewise.
21228 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21229 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
21230 (in_text_section_p, last_text_label, last_cold_label,
21231 switch_text_ranges, switch_cold_ranges): New data items.
21232 (dwarf2out_note_section_used): Remove.
21233 (dwarf2out_begin_prologue): Set fde->ignored_debug and
21235 (mark_ignored_debug_section): New helper function.
21236 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
21237 mark_ignored_debug_section.
21238 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
21239 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
21240 (size_of_aranges): Adjust formula for multi-part text ranges size.
21241 (output_aranges): Output multi-part text ranges.
21242 (dwarf2out_set_ignored_loc): New callback function.
21243 (dwarf2out_finish): Output multi-part text ranges.
21244 (dwarf2out_c_finalize): Clear new data items.
21245 * final.c (final_start_function_1): Call set_ignored_loc callback.
21246 (final_scan_insn_1): Likewise.
21247 * ggc-page.c (gt_ggc_mx): New helper function.
21248 * stringpool.c (gt_pch_nx): Likewise.
21250 2021-05-06 Richard Biener <rguenther@suse.de>
21252 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
21253 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
21254 (TV_TREE_INTO_SSA): New.
21255 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
21256 (rewrite_blocks): Likewise.
21257 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
21259 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21261 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
21262 abs_replacement, xor_replacement,
21263 cond_removal_in_popcount_clz_ctz_pattern,
21264 replace_phi_edge_with_variable): Change type of phi argument from
21265 gimple * to gphi *.
21267 2021-05-06 Richard Biener <rguenther@suse.de>
21269 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
21270 Output an opt-info message.
21271 (do_split_loop_on_cond): Likewise.
21272 (tree_ssa_split_loops): Update SSA form here.
21274 2021-05-06 Richard Biener <rguenther@suse.de>
21276 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
21277 return variable removal.
21279 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
21281 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
21282 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
21283 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
21285 * config/s390/s390.c (s390_const_operand_ok): Remove unused
21288 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21290 PR tree-optimization/94589
21291 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
21292 spaceship_replacement.
21293 (cond_only_block_p, spaceship_replacement): New functions.
21295 2021-05-06 Richard Biener <rguenther@suse.de>
21298 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
21299 dealing with a debug use and only query existing addresses
21301 (lower_emutls_1): Avoid splitting out addresses for debug
21302 stmts, reset the debug stmt when we fail to find existing
21304 (lower_emutls_phi_arg): Set wi.stmt.
21306 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
21309 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
21310 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
21311 (stack_protect_test): Use cbranch helper.
21313 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21316 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21317 always return the establisher frame for __builtin_frame_address (0).
21319 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
21322 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
21323 (ix86_cpu_model_var): Likewise.
21324 (ix86_cpu_features2_type_node): Likewise.
21325 (ix86_cpu_features2_var): Likewise.
21326 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
21329 2021-05-05 Martin Sebor <msebor@redhat.com>
21331 * passes.def (pass_warn_printf): Run after SSA.
21333 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21335 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
21336 * config/arm/predicates.md (minus_one_operand): New predicate.
21338 2021-05-05 Jeff Law <jlaw@tachyum.com>
21340 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
21342 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21344 PR rtl-optimization/100263
21345 * postreload.c (move2add_valid_value_p): Ensure register can
21348 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21350 PR rtl-optimization/100411
21351 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
21352 and beginning of function markers.
21354 2021-05-05 Jeff Law <jlaw@tachyum.com>
21356 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
21357 * config/cr16/cr16.c (notice_update_cc): Remove.
21358 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
21360 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
21363 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
21364 Handle V8QI, V4HI and V2SI modes.
21365 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
21366 (ix86_build_signbit_mask): Ditto.
21367 * config/i386/mmx.md (MMXMODE14): New mode iterator.
21368 (<smaxmin:code><MMXMODE14:mode>3): New expander.
21369 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
21370 (<umaxmin:code><MMXMODE24:mode>3): New expander.
21371 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
21372 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
21373 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21374 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21375 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21376 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21378 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21380 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
21381 not expand the VALUE_EXPR of variables put in the non-local frame.
21382 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
21383 to be ignored for debug info, ensure its variable offsets are not.
21385 2021-05-05 Richard Biener <rguenther@suse.de>
21387 PR tree-optimization/79333
21388 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
21389 Fold stmt following SSA edges.
21391 2021-05-05 Richard Biener <rguenther@suse.de>
21393 PR middle-end/100394
21394 * calls.c (expand_call): Preserve possibly throwing calls.
21395 * cfgexpand.c (expand_call_stmt): When a call can throw signal
21396 RTL expansion there are side-effects.
21397 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
21398 mark all possibly throwing stmts necessary unless we can elide
21400 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
21401 -fdelete-dead-exceptions.
21402 * tree.h (DECL_PURE_P): Add note about exceptions.
21404 2021-05-05 Alexandre Oliva <oliva@adacore.com>
21406 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
21409 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
21411 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
21412 get_fnname_from_decl for name of thunk.
21413 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
21414 and ASM_OUTPUT_LABEL.
21415 (rs6000_xcoff_declare_function_name): Use assemble_name and
21417 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
21418 (rs6000_xcoff_encode_section_info): Don't add mapping class
21419 for aliases. Always add [DS] mapping class to primary
21421 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
21423 2021-05-04 Martin Sebor <msebor@redhat.com>
21425 PR middle-end/100307
21426 * builtins.c (compute_objsize_r): Clear base0 for pointers.
21428 2021-05-04 Jeff Law <jlaw@tachyum.com>
21430 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
21432 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
21434 * caller-save.c: Remove CC0.
21435 * cfgcleanup.c: Remove CC0.
21436 * cfgrtl.c: Remove CC0.
21437 * combine.c: Remove CC0.
21438 * compare-elim.c: Remove CC0.
21439 * conditions.h: Remove CC0.
21440 * config/h8300/h8300.h: Remove CC0.
21441 * config/h8300/h8300-protos.h: Remove CC0.
21442 * config/h8300/peepholes.md: Remove CC0.
21443 * config/i386/x86-tune-sched.c: Remove CC0.
21444 * config/m68k/m68k.c: Remove CC0.
21445 * config/rl78/rl78.c: Remove CC0.
21446 * config/sparc/sparc.c: Remove CC0.
21447 * config/xtensa/xtensa.c: Remove CC0.
21448 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
21449 RTL where that is used as a placeholder only.
21450 * cprop.c: Remove CC0.
21451 * cse.c: Remove CC0.
21452 * cselib.c: Remove CC0.
21453 * df-problems.c: Remove CC0.
21454 * df-scan.c: Remove CC0.
21455 * doc/md.texi: Remove CC0. Adjust an example.
21456 * doc/rtl.texi: Remove CC0. Adjust an example.
21457 * doc/tm.texi: Regenerate.
21458 * doc/tm.texi.in: Remove CC0.
21459 * emit-rtl.c: Remove CC0.
21460 * final.c: Remove CC0.
21461 * fwprop.c: Remove CC0.
21462 * gcse-common.c: Remove CC0.
21463 * gcse.c: Remove CC0.
21464 * genattrtab.c: Remove CC0.
21465 * genconfig.c: Remove CC0.
21466 * genemit.c: Remove CC0.
21467 * genextract.c: Remove CC0.
21468 * gengenrtl.c: Remove CC0.
21469 * genrecog.c: Remove CC0.
21470 * haifa-sched.c: Remove CC0.
21471 * ifcvt.c: Remove CC0.
21472 * ira-costs.c: Remove CC0.
21473 * ira.c: Remove CC0.
21474 * jump.c: Remove CC0.
21475 * loop-invariant.c: Remove CC0.
21476 * lra-constraints.c: Remove CC0.
21477 * lra-eliminations.c: Remove CC0.
21478 * optabs.c: Remove CC0.
21479 * postreload-gcse.c: Remove CC0.
21480 * postreload.c: Remove CC0.
21481 * print-rtl.c: Remove CC0.
21482 * read-rtl-function.c: Remove CC0.
21483 * reg-notes.def: Remove CC0.
21484 * reg-stack.c: Remove CC0.
21485 * reginfo.c: Remove CC0.
21486 * regrename.c: Remove CC0.
21487 * reload.c: Remove CC0.
21488 * reload1.c: Remove CC0.
21489 * reorg.c: Remove CC0.
21490 * resource.c: Remove CC0.
21491 * rtl.c: Remove CC0.
21492 * rtl.def: Remove CC0.
21493 * rtl.h: Remove CC0.
21494 * rtlanal.c: Remove CC0.
21495 * sched-deps.c: Remove CC0.
21496 * sched-rgn.c: Remove CC0.
21497 * shrink-wrap.c: Remove CC0.
21498 * simplify-rtx.c: Remove CC0.
21499 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
21500 CC_STATUS_MDEP, and CC_STATUS.
21501 * target.def: Remove CC0.
21502 * valtrack.c: Remove CC0.
21503 * var-tracking.c: Remove CC0.
21505 2021-05-04 Richard Biener <rguenther@suse.de>
21507 PR tree-optimization/100414
21508 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
21510 (tree_ssa_phiopt_worker): But unconditionally here.
21512 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
21514 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
21515 && and || with floating-point and complex arguments.
21517 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21519 * tree-inline.c (insert_debug_decl_map): Delete.
21520 (copy_debug_stmt): Minor tweak.
21521 (setup_one_parameter): Do not use a variable if the value is either
21522 a read-only DECL or a non-addressable local variable in the caller.
21523 In this case, insert the debug-only variable in the map manually.
21524 (expand_call_inline): Do not generate a CLOBBER for these values.
21525 * tree-inline.h (debug_map): Minor tweak.
21527 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21529 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
21530 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
21532 2021-05-04 Richard Biener <rguenther@suse.de>
21534 PR tree-optimization/100329
21535 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
21537 (insert_stmt_after): Assert we're not running into asm goto.
21539 2021-05-04 Richard Biener <rguenther@suse.de>
21541 PR tree-optimization/100398
21542 * tree-ssa-dse.c (pass_dse::execute): Preserve control
21545 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21547 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
21549 2021-05-04 Alexandre Oliva <oliva@adacore.com>
21551 * builtins.c (try_store_by_multiple_pieces): New.
21552 (expand_builtin_memset_args): Use it. If target_char_cast
21553 fails, proceed as for non-constant val. Pass len's ctz to...
21554 * expr.c (clear_storage_hints): ... this. Try store by
21555 multiple pieces after setmem.
21556 (clear_storage): Adjust.
21557 * expr.h (clear_storage_hints): Likewise.
21558 (try_store_by_multiple_pieces): Declare.
21559 * passes.def: Replace the last copy_prop with ccp.
21561 2021-05-03 Tom de Vries <tdevries@suse.de>
21564 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
21567 2021-05-03 Richard Biener <rguenther@suse.de>
21569 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
21571 2021-05-03 Richard Biener <rguenther@suse.de>
21573 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
21574 (dse_dom_walker): Remove.
21575 (dse_dom_walker::dse_optimize_stmt): Rename...
21576 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
21577 (dse_dom_walker::before_dom_children): Inline ...
21578 (pass_dse::execute): ... here. Perform a reverse program
21581 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
21584 * configure: Regenerated.
21586 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
21589 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
21590 (s390_md_asm_adjust): Handle hard registers.
21592 2021-05-03 Jakub Jelinek <jakub@redhat.com>
21594 PR tree-optimization/100382
21595 * tree-ssa-dse.c: Include tree-eh.h.
21596 (dse_dom_walker::before_dom_children): Don't remove stmts if
21597 stmt_unremovable_because_of_non_call_eh_p is true.
21599 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
21601 * varasm.c (compute_reloc_for_var): Split out from...
21602 (get_variable_section): Use it.
21603 * output.h (compute_reloc_for_var): Declare.
21604 * config/rs6000/rs6000-protos.h
21605 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
21607 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
21608 Don't append storage mapping class to symbol.
21609 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
21610 Don't convert TLS BSS to common.
21611 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
21612 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
21614 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
21616 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
21617 If align is 0 from TLS class, use the same rules as varasm.c
21618 If not common, switch to BSS section manually.
21619 If common, emit appropriate comm or lcomm directive.
21620 (rs6000_xcoff_encode_section_info): Add logic to append all
21621 storage mapping classes.
21622 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
21623 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
21624 rs6000_xcoff_asm_output_aligned_decl_common.
21625 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
21626 rs6000_xcoff_asm_output_aligned_decl_common.
21627 (ASM_OUTPUT_TLS_COMMON): Use
21628 rs6000_xcoff_asm_output_aligned_decl_common.
21630 2021-05-02 Jakub Jelinek <jakub@redhat.com>
21633 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
21634 as first argument of pseudo_node_t constructors.
21636 2021-05-02 Jakub Jelinek <jakub@redhat.com>
21639 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
21641 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
21643 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
21644 (gt_pch_nx (int_range<1> *&)): New.
21645 (gt_ggc_mx (int_range<1> *&)): New.
21646 * value-range.h (class irange): Add GTY support for
21649 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
21651 * doc/options.texi (Negative): Change either or to both and.
21653 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21655 * config/aarch64/aarch64-simd-builtins.def: Add
21656 float_ml[as][q]_laneq builtin generator macros.
21657 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
21658 (aarch64_float_mla_laneq<mode>): Define.
21659 (aarch64_float_mls_laneq<mode>): Define.
21660 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
21661 instead of GCC vector extensions.
21662 (vmlaq_laneq_f32): Likewise.
21663 (vmls_laneq_f32): Likewise.
21664 (vmlsq_laneq_f32): Likewise.
21666 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21668 * config/aarch64/aarch64-simd-builtins.def: Add
21669 float_ml[as]_lane builtin generator macros.
21670 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
21672 (mul_lane<mode>3): This, and re-order arguments.
21673 (aarch64_float_mla_lane<mode>): Define.
21674 (aarch64_float_mls_lane<mode>): Define.
21675 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
21676 instead of GCC vector extensions.
21677 (vmlaq_lane_f32): Likewise.
21678 (vmls_lane_f32): Likewise.
21679 (vmlsq_lane_f32): Likewise.
21681 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21683 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
21684 builtin generator macros.
21685 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
21687 (aarch64_float_mls<mode>): Define.
21688 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
21689 instead of relying on GCC vector extensions.
21690 (vmla_f64): Likewise.
21691 (vmlaq_f32): Likewise.
21692 (vmlaq_f64): Likewise.
21693 (vmls_f32): Likewise.
21694 (vmls_f64): Likewise.
21695 (vmlsq_f32): Likewise.
21696 (vmlsq_f64): Likewise.
21697 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
21699 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21701 * config/aarch64/aarch64-simd-builtins.def: Add
21702 float_ml[as]_n_builtin generator macros.
21703 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
21705 (mul_n<mode>3): This, and re-order arguments.
21706 (aarch64_float_mla_n<mode>): Define.
21707 (aarch64_float_mls_n<mode>): Define.
21708 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
21709 instead of inline asm.
21710 (vmlaq_n_f32): Likewise.
21711 (vmls_n_f32): Likewise.
21712 (vmlsq_n_f32): Likewise.
21714 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
21716 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
21717 builtin generator macros.
21718 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
21719 (aarch64_pmull_hiv16qi_insn): Define.
21720 (aarch64_pmull_hiv16qi): Define.
21721 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
21722 instead of inline asm.
21723 (vmull_p8): Likewise.
21725 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
21727 * config/avr/avr.md: Adjust peepholes to match and
21728 generate parallels with clobber of REG_CC.
21729 (mov<mode>_insn): Rename to mov<mode>_insn_split.
21730 (*mov<mode>_insn): Rename to mov<mode>_insn.
21732 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
21734 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
21735 for VAR_DECLs if -fdata-sections enabled.
21737 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
21739 PR bootstrap/100327
21740 * config/rs6000/rs6000.c
21741 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
21742 (rs6000_libgcc_floating_mode_supported_p): New target hook.
21744 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
21746 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
21748 (class back_threader_registry): ...to here...
21749 (class back_threader_profitability): ...and here...
21750 (thread_jumps::thread_through_all_blocks): Remove argument.
21751 (back_threader_registry::back_threader_registry): New.
21752 (back_threader_registry::~back_threader_registry): New.
21753 (back_threader_registry::thread_through_all_blocks): New.
21754 (thread_jumps::profitable_jump_thread_path): Move from here...
21755 (back_threader_profitability::profitable_path_p): ...to here.
21756 (thread_jumps::find_taken_edge): New.
21757 (thread_jumps::convert_and_register_current_path): Move...
21758 (back_threader_registry::register_path): ...to here.
21759 (thread_jumps::register_jump_thread_path_if_profitable): Move...
21760 (thread_jumps::maybe_register_path): ...to here.
21761 (thread_jumps::handle_phi): Call find_taken_edge and
21762 maybe_register_path.
21763 (thread_jumps::handle_assignment): Same.
21764 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
21765 tree argument to handle_phi and handle_assignment.
21766 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
21767 set of m_speed_p and m_max_threaded_paths.
21768 (pass_thread_jumps::execute): Remove second argument from
21769 find_jump_threads_backwards.
21770 (pass_early_thread_jumps::execute): Same.
21772 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
21774 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
21775 (class dom_opt_dom_walker): Initialize some class variables.
21776 (pass_dominator::execute): Pass evrp_range_analyzer and
21777 dom_jump_threader_simplifier to dom_opt_dom_walker.
21778 Adjust for some functions moving into classes.
21779 (simplify_stmt_for_jump_threading): Adjust and move to...
21780 (jump_threader_simplifier::simplify): ...here.
21781 (dom_opt_dom_walker::before_dom_children): Adjust for
21782 m_evrp_range_analyzer.
21783 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
21784 (test_for_singularity): Place in dom_opt_dom_walker class.
21785 (dom_opt_dom_walker::optimize_stmt): The argument
21786 evrp_range_analyzer is now a class field.
21787 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
21788 (thread_jumps::thread_through_all_blocks): New.
21789 (thread_jumps::convert_and_register_current_path): Use m_registry.
21790 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
21791 being in the threader class.
21792 (pass_early_thread_jumps::execute): Same.
21793 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
21794 (jump_threader::jump_threader): ...here.
21795 (threadedge_finalize_values): Move...
21796 (jump_threader::~jump_threader): ...here.
21797 (jump_threader::remove_jump_threads_including): New.
21798 (jump_threader::thread_through_all_blocks): New.
21799 (record_temporary_equivalences_from_phis): Move...
21800 (jump_threader::record_temporary_equivalences_from_phis): ...here.
21801 (record_temporary_equivalences_from_stmts_at_dest): Move...
21802 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
21804 (simplify_control_stmt_condition_1): Move to jump_threader class.
21805 (simplify_control_stmt_condition): Move...
21806 (jump_threader::simplify_control_stmt_condition): ...here.
21807 (thread_around_empty_blocks): Move...
21808 (jump_threader::thread_around_empty_blocks): ...here.
21809 (thread_through_normal_block): Move...
21810 (jump_threader::thread_through_normal_block): ...here.
21811 (thread_across_edge): Move...
21812 (jump_threader::thread_across_edge): ...here.
21813 (thread_outgoing_edges): Move...
21814 (jump_threader::thread_outgoing_edges): ...here.
21815 * tree-ssa-threadedge.h: Move externally facing functings...
21816 (class jump_threader): ...here...
21817 (class jump_threader_simplifier): ...and here.
21818 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
21819 (jump_thread_path_allocator::jump_thread_path_allocator): New.
21820 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
21821 (jump_thread_path_allocator::allocate_thread_edge): New.
21822 (jump_thread_path_allocator::allocate_thread_path): New.
21823 (jump_thread_path_registry::jump_thread_path_registry): New.
21824 (jump_thread_path_registry::~jump_thread_path_registry): New.
21825 (jump_thread_path_registry::allocate_thread_edge): New.
21826 (jump_thread_path_registry::allocate_thread_path): New.
21827 (dump_jump_thread_path): Make extern.
21828 (debug (const vec<jump_thread_edge *> &path)): New.
21829 (struct removed_edges): Move to tree-ssa-threadupdate.h.
21830 (struct thread_stats_d): Remove.
21831 (remove_ctrl_stmt_and_useless_edges): Make static.
21832 (lookup_redirection_data): Move...
21833 (jump_thread_path_registry::lookup_redirection_data): ...here.
21834 (ssa_redirect_edges): Make static.
21835 (thread_block_1): Move...
21836 (jump_thread_path_registry::thread_block_1): ...here.
21837 (thread_block): Move...
21838 (jump_thread_path_registry::thread_block): ...here.
21839 (thread_through_loop_header): Move...
21840 (jump_thread_path_registry::thread_through_loop_header): ...here.
21841 (mark_threaded_blocks): Move...
21842 (jump_thread_path_registry::mark_threaded_blocks): ...here.
21843 (debug_path): Move...
21844 (jump_thread_path_registry::debug_path): ...here.
21845 (debug_all_paths): Move...
21846 (jump_thread_path_registry::dump): ..here.
21847 (rewire_first_differing_edge): Move...
21848 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
21849 (adjust_paths_after_duplication): Move...
21850 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
21851 (duplicate_thread_path): Move...
21852 (jump_thread_path_registry::duplicate_thread_path): ..here.
21853 (remove_jump_threads_including): Move...
21854 (jump_thread_path_registry::remove_jump_threads_including): ...here.
21855 (thread_through_all_blocks): Move to...
21856 (jump_thread_path_registry::thread_through_all_blocks): ...here.
21857 (delete_jump_thread_path): Remove.
21858 (register_jump_thread): Move...
21859 (jump_thread_path_registry::register_jump_thread): ...here.
21860 * tree-ssa-threadupdate.h: Move externally facing functions...
21861 (class jump_thread_path_allocator): ...here...
21862 (class jump_thread_path_registry): ...and here.
21863 (thread_through_all_blocks): Remove.
21864 (struct removed_edges): New.
21865 (register_jump_thread): Remove.
21866 (remove_jump_threads_including): Remove.
21867 (delete_jump_thread_path): Remove.
21868 (remove_ctrl_stmt_and_useless_edges): Remove.
21869 (free_dom_edge_info): New prototype.
21870 * tree-vrp.c: Remove x_vr_values hack.
21871 (class vrp_jump_threader_simplifier): New.
21872 (vrp_jump_threader_simplifier::simplify): New.
21873 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
21874 Remove m_dummy_cond.
21875 Instantiate m_simplifier and m_threader.
21876 (vrp_jump_threader::thread_through_all_blocks): New.
21877 (vrp_jump_threader::simplify_stmt): Remove.
21878 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
21879 Remove x_vr_values hack.
21880 (execute_vrp): Adjust for thread_through_all_blocks being in a
21883 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
21885 * genflags.c (gen_insn): Print failed expansion string.
21887 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
21889 * expr.c (alignment_for_piecewise_move): Call mode_for_size
21890 without limit to MAX_FIXED_MODE_SIZE.
21892 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
21894 PR middle-end/90773
21895 * builtins.c (builtin_memset_gen_str): Don't use return from
21896 simplify_gen_subreg.
21898 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
21901 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
21902 (*addsi3_carry_zext_0r): Ditto.
21903 (*sub<mode>3_carry_0): Ditto.
21904 (*subsi3_carry_zext_0r): Ditto.
21905 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
21907 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
21908 Also consider ix86_carry_flag_unset_operator to calculate
21909 the cost of adc/sbb insn.
21911 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
21913 PR rtl-optimization/100225
21914 PR rtl-optimization/84878
21915 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
21916 where we have an instruction which touches (writes) any hard
21917 register from df->regular_block_artificial_uses set.
21918 Allow not-single-set instruction only right before basic block
21921 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
21923 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
21925 2021-04-30 LevyHsu <admin@levyhsu.com>
21927 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
21928 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
21929 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
21930 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
21932 2021-04-29 Alexandre Oliva <oliva@adacore.com>
21934 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
21936 2021-04-29 Alexandre Oliva <oliva@adacore.com>
21938 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
21939 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
21940 * config/i386/i386.c: Adjust.
21941 * config/i386/i386.md: Adjust.
21942 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
21943 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21944 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21945 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21946 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21947 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21948 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21949 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21950 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21951 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21952 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
21954 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
21956 * config/i386/i386-expand.c (ix86_expand_int_compare):
21957 Swap operands of GTU and LEU comparison to emit carry flag comparison.
21958 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
21959 predicate to allow more combine opportunities with memory operands.
21960 (*sub<mode>3_carry_0): Ditto.
21962 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
21964 PR rtl-optimization/100303
21965 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
21966 boolean that indicates whether the use will only be used in
21967 debug instructions. Treat it in the same way that existing
21968 cross-EBB debug references would be handled if so.
21969 (function_info::make_uses_available): Likewise.
21970 * rtl-ssa/functions.h (function_info::make_uses_available): Update
21971 prototype accordingly.
21972 (function_info::make_uses_available): Likewise.
21973 * fwprop.c (try_fwprop_subst): Update call accordingly.
21975 2021-04-29 Jeff Law <jlaw@tachyum.com>
21977 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
21980 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
21981 Richard Biener <rguenther@suse.de>
21984 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
21985 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
21986 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
21987 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
21988 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
21989 to PURE_ARGS category.
21990 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
21991 Handle PURE_ARGS category.
21992 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
21994 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
21996 * configure.ac: Check for the presence of sys/locking.h header and
21997 for whether _LK_LOCK is supported by _locking.
21998 * configure: Regenerate.
21999 * config.in: Likewise.
22000 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
22001 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
22002 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
22004 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22006 * config/i386/predicates.md (fcmov_comparison_operator):
22007 Do not check for trivial FP comparison operator.
22008 <case GEU, case LTU>: Allow CCGZmode.
22009 <case GTU, case LEU>: Do not allow CCCmode.
22010 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
22011 (ix86_carry_flag_operator): Match only LTU and UNLT code.
22012 Do not check for trivial FP comparison operator. Allow CCGZmode.
22014 2021-04-29 Tom de Vries <tdevries@suse.de>
22016 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
22017 fd->loop.step by either step or orig_step.
22019 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
22021 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
22022 (load_got_register): Do the PIC dance here.
22023 (sparc_legitimize_tls_address): Simplify.
22024 (sparc_emit_probe_stack_range): Likewise.
22025 (sparc32_initialize_trampoline): Likewise.
22026 (sparc64_initialize_trampoline): Likewise.
22027 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
22028 (probe_stack_range<P:mode>): Likewise.
22029 (flush<P:mode>): Likewise.
22030 (tgd_hi22<P:mode>): Likewise.
22031 (tgd_lo10<P:mode>): Likewise.
22032 (tgd_add<P:mode>): Likewise.
22033 (tgd_call<P:mode>): Likewise.
22034 (tldm_hi22<P:mode>): Likewise.
22035 (tldm_lo10<P:mode>): Likewise.
22036 (tldm_add<P:mode>): Likewise.
22037 (tldm_call<P:mode>): Likewise.
22038 (tldo_hix22<P:mode>): Likewise.
22039 (tldo_lox10<P:mode>): Likewise.
22040 (tldo_add<P:mode>): Likewise.
22041 (tie_hi22<P:mode>): Likewise.
22042 (tie_lo10<P:mode>): Likewise.
22043 (tie_add<P:mode>): Likewise.
22044 (tle_hix22<P:mode>): Likewise.
22045 (tle_lox10<P:mode>): Likewise.
22046 (stack_protect_setsi): Rename to...
22047 (stack_protect_set32): ...this.
22048 (stack_protect_setdi): Rename to...
22049 (stack_protect_set64): ...this.
22050 (stack_protect_set): Adjust calls to above.
22051 (stack_protect_testsi): Rename to...
22052 (stack_protect_test32): ...this.
22053 (stack_protect_testdi): Rename to...
22054 (stack_protect_test64): ...this.
22055 (stack_protect_test): Adjust calls to above.
22057 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
22059 PR middle-end/90773
22060 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
22061 (builtin_strncpy_read_str): Likewise.
22062 (builtin_memset_read_str): Add an argument for the previous RTL
22063 information and generate the new RTL from the previous RTL info.
22064 (builtin_memset_gen_str): Likewise.
22065 * builtins.h (builtin_strncpy_read_str): Update the prototype.
22066 (builtin_memset_read_str): Likewise.
22067 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
22068 returns true, round up size and alignment to the widest integer
22069 mode for maximum size.
22070 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
22071 and pass it to m_constfn.
22072 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
22073 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
22074 initialize m_push. Initialize m_overlap_op_by_pieces with
22075 targetm.overlap_op_by_pieces_p ().
22076 (op_by_pieces_d::run): Pass the previous RTL information to
22077 pieces_addr::adjust and generate overlapping operations if
22078 m_overlap_op_by_pieces is true.
22080 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
22082 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
22084 (can_store_by_pieces): Use by_pieces_constfn on constfun.
22085 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
22086 for op_by_pieces_d change.
22087 (clear_by_pieces_1): Add a dummy argument.
22088 (clear_by_pieces): Updated for op_by_pieces_d change.
22089 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
22090 (string_cst_read_str): Add a dummy argument.
22091 * expr.h (by_pieces_constfn): Add a dummy argument.
22092 (by_pieces_prev): New.
22093 * target.def (overlap_op_by_pieces_p): New target hook.
22094 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
22095 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
22096 * doc/tm.texi: Regenerated.
22098 2021-04-29 Richard Biener <rguenther@suse.de>
22100 PR tree-optimization/100253
22101 * tree-vect-stmts.c (vectorizable_load): Do not assume
22102 element alignment when DR_MISALIGNMENT is -1.
22103 (vectorizable_store): Likewise.
22105 2021-04-29 Jakub Jelinek <jakub@redhat.com>
22108 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
22109 absu_hwi instead of abs_hwi.
22111 2021-04-29 Richard Biener <rguenther@suse.de>
22113 PR middle-end/38474
22114 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
22115 forwarding when indirect forwarding through ESCAPED
22118 2021-04-29 Tom de Vries <tdevries@suse.de>
22121 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
22122 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
22123 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
22124 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
22126 2021-04-29 Richard Biener <rguenther@suse.de>
22128 PR tree-optimization/99912
22129 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
22130 (dse_dom_walker::todo): Likewise.
22131 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
22133 (dse_dom_walker::before_dom_children): Remove trivially
22134 dead SSA defs and schedule CFG cleanup if we removed all
22136 (pass_dse::execute): Get TODO as computed by the DOM walker
22137 and return it. Wipe dominator info earlier.
22139 2021-04-29 Richard Biener <rguenther@suse.de>
22142 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
22143 Track blocks to cleanup EH in new m_need_eh_cleanup.
22144 (ipcp_modif_dom_walker::cleanup_eh): New.
22145 (ipcp_transform_function): Release dominator info before
22148 2021-04-29 Martin Sebor <msebor@redhat.com>
22150 PR middle-end/100250
22151 * attribs.c (attr_access::array_as_string): Avoid dereferencing
22152 a pointer when it's null.
22154 2021-04-29 Martin Sebor <msebor@redhat.com>
22156 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
22157 * ipa-free-lang-data.cc: New file.
22158 * tree.c: Move pass free_lang_data to file above.
22159 (build_array_type_1): Declare extern.
22160 * tree.h (build_array_type_1): Declare.
22162 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22164 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
22165 make consistent with updated RTL pattern.
22166 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
22167 Implement using ss_truncate and us_truncate rather than
22169 * config/aarch64/iterators.md: Remove redundant unspecs and
22170 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
22172 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22174 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
22175 attributes consistent with those defined in arm_neon.h.
22177 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22179 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
22180 attributes consistent with those defined in arm_neon.h.
22182 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22184 * config/aarch64/aarch64-simd-builtins.def: Add
22185 float_trunc_rodd builtin generator macros.
22186 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
22188 (aarch64_float_trunc_rodd_lo_v2sf): Define.
22189 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
22190 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
22191 (aarch64_float_trunc_rodd_hi_v4sf): Define.
22192 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
22193 instead of inline asm.
22194 (vcvtx_high_f32_f64): Likewise.
22195 (vcvtxd_f32_f64): Likewise.
22196 * config/aarch64/iterators.md: Add FCVTXN unspec.
22198 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22200 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
22202 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
22204 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
22205 instead of inline asm.
22206 (vqtbx1_u8): Likewise.
22207 (vqtbx1_p8): Likewise.
22208 (vqtbx1q_s8): Likewise.
22209 (vqtbx1q_u8): Likewise.
22210 (vqtbx1q_p8): Likewise.
22211 (vtbx2_s8): Likewise.
22212 (vtbx2_u8): Likewise.
22213 (vtbx2_p8): Likewise.
22215 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22217 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
22219 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
22220 instead of inline asm.
22221 (vqtbl1_s8): Likewise.
22222 (vqtbl1_u8): Likewise.
22223 (vqtbl1q_p8): Likewise.
22224 (vqtbl1q_s8): Likewise.
22225 (vqtbl1q_u8): Likewise.
22226 (vtbl1_s8): Likewise.
22227 (vtbl1_u8): Likewise.
22228 (vtbl1_p8): Likewise.
22229 (vtbl2_s8): Likewise.
22230 (vtbl2_u8): Likewise.
22231 (vtbl2_p8): Likewise.
22233 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22235 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
22236 ssri_n buitin generator macro.
22237 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
22238 instead of inline asm.
22239 (vsri_n_p16): Likewise.
22240 (vsri_n_p64): Likewise.
22241 (vsriq_n_p8): Likewise.
22242 (vsriq_n_p16): Likewise.
22243 (vsriq_n_p64): Likewise.
22245 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22247 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
22248 iterator for polynomial ssli_n builtin generator macro.
22249 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
22250 instead of inline asm.
22251 (vsli_n_p16): Likewise.
22252 (vsliq_n_p8): Likewise.
22253 (vsliq_n_p16): Likewise.
22254 * config/aarch64/iterators.md: Define VALLP mode iterator.
22256 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22258 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
22259 iterator to generate [su]adalp RTL builtins.
22260 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
22261 [su]adalp RTL pattern.
22262 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
22263 instead of inline asm.
22264 (vpadal_u32): Likewise.
22266 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22268 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
22269 builtin generator macros.
22270 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
22272 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
22273 instead of inline asm.
22274 (vpaddl_s16): Likewise.
22275 (vpaddl_s32): Likewise.
22276 (vpaddl_u8): Likewise.
22277 (vpaddl_u16): Likewise.
22278 (vpaddl_u32): Likewise.
22279 (vpaddlq_s8): Likewise.
22280 (vpaddlq_s16): Likewise.
22281 (vpaddlq_s32): Likewise.
22282 (vpaddlq_u8): Likewise.
22283 (vpaddlq_u16): Likewise.
22284 (vpaddlq_u32): Liwewise.
22285 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
22286 appropriate attributes.
22288 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22290 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
22291 for aarch64_addp<mode> builtin macro generator.
22292 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
22293 aarch64_addp<mode> RTL pattern.
22294 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
22295 instead of inline asm.
22296 (vpaddq_s16): Likewise.
22297 (vpaddq_s32): Likewise.
22298 (vpaddq_s64): Likewise.
22299 (vpaddq_u8): Likewise.
22300 (vpaddq_u16): Likewise.
22301 (vpaddq_u32): Likewise.
22302 (vpaddq_u64): Likewise.
22304 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22306 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
22307 builtin generator macros.
22308 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
22310 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
22311 instead of inline asm.
22312 (vqdmulh_n_s32): Likewise.
22313 (vqdmulhq_n_s16): Likewise.
22314 (vqdmulhq_n_s32): Likewise.
22315 (vqrdmulh_n_s16): Likewise.
22316 (vqrdmulh_n_s32): Likewise.
22317 (vqrdmulhq_n_s16): Likewise.
22318 (vqrdmulhq_n_s32): Likewise.
22320 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
22322 * doc/install.texi (--enable-offload-defaulted): Document.
22324 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
22326 * config/avr/avr-dimode.md: Turn existing patterns into
22327 define_insn_and_split style patterns where the splitter
22328 adds a clobber of the condition code register. Drop "cc"
22329 attribute. Add new patterns to match output of
22331 * config/avr/avr-fixed.md: Likewise.
22332 * config/avr/avr.c (cc_reg_rtx): New.
22333 (avr_parallel_insn_from_insns): Adjust insn count
22334 for removal of set of cc0.
22335 (avr_is_casesi_sequence): Likewise.
22336 (avr_casei_sequence_check_operands): Likewise.
22337 (avr_optimize_casesi): Likewise. Also insert
22338 new insns after jump_insn.
22339 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
22340 for removal of set of cc0.
22341 (avr_init_expanders): Initialize cc_reg_rtx.
22342 (avr_regno_reg_class): Handle REG_CC.
22343 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
22344 (avr_notice_update_cc): Remove function.
22345 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
22346 (compare_condition): Adjust for PARALLEL with
22348 (out_shift_with_cnt): Likewise.
22349 (ashlhi3_out): Likewise.
22350 (ashrhi3_out): Likewise.
22351 (lshrhi3_out): Likewise.
22352 (avr_class_max_nregs): Return single reg for REG_CC.
22353 (avr_compare_pattern): Check for REG_CC instead
22355 (avr_reorg_remove_redundant_compare): Likewise.
22356 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
22357 (avr_hard_regno_nregs): Return single reg for REG_CC.
22358 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
22359 (avr_md_asm_adjust): Clobber REG_CC.
22360 (TARGET_HARD_REGNO_NREGS): Define.
22361 (TARGET_CLASS_MAX_NREGS): Define.
22362 (TARGET_MD_ASM_ADJUST): Define.
22363 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
22365 (enum reg_class): Add CC_REG class.
22366 (NOTICE_UPDATE_CC): Remove.
22367 (CC_OVERFLOW_UNUSABLE): Remove.
22368 (CC_NO_CARRY): Remove.
22369 * config/avr/avr.md: Turn existing patterns into
22370 define_insn_and_split style patterns where the splitter
22371 adds a clobber of the condition code register. Drop "cc"
22372 attribute. Add new patterns to match output of
22374 (sez): Remove unused pattern.
22376 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
22379 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
22382 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
22385 * config/aarch64/constraints.md (Utq): Require the address to
22386 be valid for both the element mode and for V2DImode.
22388 2021-04-28 Jakub Jelinek <jakub@redhat.com>
22389 Tobias Burnus <tobias@codesourcery.com>
22391 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
22392 * gcc.c (process_command): New variable.
22393 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
22394 set it if -foffload is defaulted.
22395 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
22396 (compile_offload_image): If OFFLOAD_DEFAULTED and
22397 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
22398 if corresponding mkoffload can't be found.
22399 (compile_images_for_offload_targets): Likewise. Free and clear
22400 offload_names if no valid offload is found.
22401 * config.in: Regenerate.
22402 * configure: Regenerate.
22404 2021-04-28 Richard Biener <rguenther@suse.de>
22406 PR tree-optimization/100292
22407 * tree-vect-generic.c (expand_vector_condition): Do not fold
22410 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22412 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
22413 * config/rs6000/aix64.opt (m64): New.
22416 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22418 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
22419 (index_term_p): Handle ASHIFT too.
22421 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22423 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
22424 (sync_lock_test_and_set<mode>): Adjust accordingly.
22425 (sync_lock_release<mode>): Likewise.
22427 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22429 * config/vax/vax-protos.h (adjacent_operands_p): Remove
22431 * config/vax/vax.c (adjacent_operands_p): Remove.
22433 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
22435 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
22436 through to the non-conditional execution case if getting the
22437 condition for conditional execution has failed.
22439 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22441 PR middle-end/100284
22442 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
22443 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
22444 than asserting on it.
22446 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22448 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
22449 with TARGET_AIX_OS.
22451 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22454 * calls.c (precompute_register_parameters): Additionally test
22455 targetm.precompute_tls_p to pre-compute argument.
22456 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
22457 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
22458 * target.def (precompute_tls_p): New.
22459 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
22460 * doc/tm.texi: Regenerated.
22462 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22465 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
22466 back to HOST_WIDE_INT.
22468 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22471 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
22472 memory alignment for the outer mode.
22474 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
22476 PR middle-end/90773
22477 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
22478 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
22480 2021-04-27 Alex Coplan <alex.coplan@arm.com>
22483 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
22484 with negative immediates: ensure we expand cbranchsi4_scratch
22485 correctly and ensure we satisfy its constraints.
22486 * config/arm/sync.md
22487 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
22488 attempt to tie two output operands together with constraints;
22489 collapse two alternatives.
22490 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
22491 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
22493 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22496 * config/aarch64/predicates.md (aarch64_sub_immediate,
22497 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
22498 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
22499 * config/aarch64/aarch64.c (aarch64_print_operand,
22500 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
22502 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22504 PR tree-optimization/100239
22505 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
22506 permutations with all indices from the first zero element as vec_shl.
22508 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22510 PR rtl-optimization/100254
22511 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
22512 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
22514 2021-04-27 Richard Biener <rguenther@suse.de>
22516 PR tree-optimization/99912
22517 * passes.def: Add comment about new TODO_remove_unused_locals.
22518 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
22521 2021-04-27 Richard Biener <rguenther@suse.de>
22523 PR tree-optimization/99912
22524 * passes.def (pass_all_optimizations): Add pass_dse before
22525 the first pass_dce, move the first pass_dse before the
22526 pass_dce following pass_pre.
22528 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22530 PR tree-optimization/95527
22531 * generic-match-head.c: Include tm.h.
22532 * gimple-match-head.c: Include tm.h.
22533 * match.pd (CLZ == INTEGER_CST): Don't use
22534 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
22535 if clz == CFN_CLZ. Add missing val declaration.
22536 (CTZ cmp CST): New simplifications.
22538 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22540 PR tree-optimization/96696
22541 * expr.c (expand_expr_divmod): New function.
22542 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
22543 divisions. Formatting fixes.
22544 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
22547 2021-04-27 Martin Jambor <mjambor@suse.cz>
22550 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
22551 If removing a call statement LHS SSA name, release it.
22553 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
22556 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
22557 is valid before including it in the mask.
22559 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22562 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
22565 2021-04-27 Richard Biener <rguenther@suse.de>
22567 PR tree-optimization/100051
22568 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
22569 disambiguator based on access size vs. decl size.
22571 2021-04-27 Richard Biener <rguenther@suse.de>
22573 PR tree-optimization/100278
22574 * tree-ssa-pre.c (compute_avail): Give up when we cannot
22575 adjust TBAA beacuse of mismatching bases.
22577 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22580 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
22581 For any_rotate define_insn_split and following splitters, use
22582 SWI iterator instead of SWI48.
22584 2021-04-27 Richard Biener <rguenther@suse.de>
22586 PR tree-optimization/99776
22587 * match.pd (bit_field_ref (ctor)): Relax element extract
22588 type compatibility checks.
22590 2021-04-27 Cui,Lili <lili.cui@intel.com>
22592 * common/config/i386/i386-common.c (processor_names):
22593 Sync processor_names with processor_type.
22594 * config/i386/i386-options.c (processor_cost_table):
22595 Sync processor_cost_table with processor_type.
22597 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22599 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
22600 (irange::set): Call irange_set_1bit_anti_range for handling all
22601 1-bit ranges. Fall through on ~[MIN,MAX].
22603 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22605 * value-range.cc (irange::legacy_num_pairs): Remove.
22606 (irange::invert): Change gcc_assert to gcc_checking_assert.
22607 * value-range.h (irange::num_pairs): Adjust for a cached
22608 num_pairs(). Also, rename all gcc_assert's to
22609 gcc_checking_assert's.
22611 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22613 * value-range.cc (irange::operator=): Set m_kind.
22614 (irange::copy_to_legacy): Handle varying and undefined sources
22615 as a legacy copy since they can be easily copied.
22616 (irange::irange_set): Set m_kind.
22617 (irange::irange_set_anti_range): Same.
22618 (irange::set): Rename normalize_min_max to normalize_kind.
22619 (irange::verify_range): Adjust for multi-ranges having the
22621 (irange::irange_union): Set m_kind.
22622 (irange::irange_intersect): Same.
22623 (irange::invert): Same.
22624 * value-range.h (irange::kind): Always return m_kind.
22625 (irange::varying_p): Rename to...
22626 (irange::varying_comptaible_p): ...this.
22627 (irange::undefined_p): Only look at m_kind.
22628 (irange::irange): Always set VR_UNDEFINED if applicable.
22629 (irange::set_undefined): Always set VR_UNDEFINED.
22630 (irange::set_varying): Always set m_kind to VR_VARYING.
22631 (irange::normalize_min_max): Rename to...
22632 (irange::normalize_kind): ...this.
22634 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22636 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
22637 Adjust for constant_p including varying_p.
22638 * tree-vrp.c (vrp_prop::finalize): Same.
22639 (determine_value_range): Same.
22640 * vr-values.c (vr_values::range_of_expr): Same.
22641 * value-range.cc (irange::symbolic_p): Do not check varying_p.
22642 (irange::constant_p): Same.
22644 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22646 * value-range.cc (irange::legacy_lower_bound): Replace
22647 !undefined_p check with num_ranges > 0.
22648 (irange::legacy_upper_bound): Same.
22649 * value-range.h (irange::type): Same.
22650 (irange::lower_bound): Same.
22651 (irange::upper_bound): Same.
22653 2021-04-26 Richard Biener <rguenther@suse.de>
22655 PR tree-optimization/99956
22656 * gimple-loop-interchange.cc (compute_access_stride):
22657 Try instantiating the access in a shallower loop nest
22658 if instantiating failed.
22659 (compute_access_strides): Pass adjustable loop_nest
22660 to compute_access_stride.
22662 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
22664 * doc/sourcebuild.texi (arm_cmse_hw): Document.
22666 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22668 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
22670 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
22671 Nathan Sidwell <nathan@codesourcery.com>
22672 Tom de Vries <vries@codesourcery.com>
22673 Julian Brown <julian@codesourcery.com>
22674 Kwok Cheung Yeung <kcy@codesourcery.com>
22676 * omp-offload.c (oacc_validate_dims): Implement
22677 '-Wopenacc-parallelism'.
22678 * doc/invoke.texi (-Wopenacc-parallelism): Document.
22680 2021-04-26 Richard Biener <rguenther@suse.de>
22682 * tree-cfg.h (gimplify_build1): Remove.
22683 (gimplify_build2): Likewise.
22684 (gimplify_build3): Likewise.
22685 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
22686 (gimplify_build2): Likewise.
22687 (gimplify_build3): Likewise.
22688 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
22690 (gimplify_build2): Likewise.
22691 (gimplify_build3): Likewise.
22692 (tree_vec_extract): Use resimplify with following SSA edges.
22693 (expand_vector_parallel): Avoid passing NULL size/bitpos
22694 to tree_vec_extract.
22695 * expr.c (store_constructor): Deal with zero-element CTORs.
22696 * match.pd (bit_field_ref <vector CTOR>): Make sure to
22697 produce vector constants when possible.
22699 2021-04-26 Richard Biener <rguenther@suse.de>
22701 * tree-complex.c: Include gimple-fold.h.
22702 (expand_complex_addition): Use gimple_build.
22703 (expand_complex_multiplication_components): Likewise.
22704 (expand_complex_multiplication): Likewise.
22705 (expand_complex_div_straight): Likewise.
22706 (expand_complex_div_wide): Likewise.
22707 (expand_complex_division): Likewise.
22708 (expand_complex_conjugate): Likewise.
22709 (expand_complex_comparison): Likewise.
22711 2021-04-26 Richard Biener <rguenther@suse.de>
22713 * tree-ssa-phiopt.c (two_value_replacement): Remove use
22714 of legacy gimplify_buildN API.
22716 2021-04-26 Richard Biener <rguenther@suse.de>
22718 PR tree-optimization/99473
22719 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
22722 2021-04-26 Richard Biener <rguenther@suse.de>
22724 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
22725 Use replace_call_with_value.
22727 2021-04-26 Richard Biener <rguenther@suse.de>
22729 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
22730 (update_gimple_call): Likewise.
22731 (update_call_from_tree): Likewise.
22732 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
22733 (valid_gimple_call_p): Likewise.
22734 (move_ssa_defining_stmt_for_defs): Likewise.
22735 (finish_update_gimple_call): Likewise.
22736 (update_gimple_call): Likewise.
22737 (update_call_from_tree): Likewise.
22738 (propagate_tree_value_into_stmt): Use replace_call_with_value.
22739 * gimple-fold.h (update_gimple_call): Declare.
22740 * gimple-fold.c (valid_gimple_rhs_p): Move here from
22741 tree-ssa-propagate.c.
22742 (update_gimple_call): Likewise.
22743 (valid_gimple_call_p): Likewise.
22744 (finish_update_gimple_call): Likewise, and simplify.
22745 (gimplify_and_update_call_from_tree): Implement
22746 update_call_from_tree functionality, avoid excessive
22747 push/pop_gimplify_context.
22748 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
22749 (gimple_fold_call): Likewise.
22750 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
22751 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
22752 (pass_fold_builtins::execute): Likewise.
22753 (optimize_stack_restore): Use replace_call_with_value.
22754 * tree-cfg.c (fold_loop_internal_call): Likewise.
22755 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
22756 only gimplify_and_update_call_from_tree.
22757 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
22758 (handle_builtin_strchr): Likewise.
22759 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
22761 2021-04-26 Jakub Jelinek <jakub@redhat.com>
22764 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
22765 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
22768 2021-04-25 liuhongt <hongtao.liu@intel.com>
22771 * config/i386/i386-builtin.def (BDESC): Change the icode of
22772 the following builtins to CODE_FOR_nothing.
22773 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
22774 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
22775 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
22776 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
22777 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
22778 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
22779 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
22780 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
22781 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
22782 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
22783 (sse2_eq<mode>3): Ditto.
22784 (sse4_1_eqv2di3): Ditto.
22785 (sse2_gt<mode>3): Rename to ..
22786 (*sse2_gt<mode>3): .. this.
22788 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22791 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22794 * config/darwin.c (darwin_binds_local_p): Assume that any
22795 public symbol might be interposed for PIC code. Update function
22796 header comment to reflect current Darwin capability.
22798 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22801 * config/darwin.c (darwin_binds_local_p): Assume that any
22802 public symbol might be interposed for PIC code. Update function
22803 header comment to reflect current Darwin capability.
22805 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
22807 * doc/sourcebuild.texi: Document no-opts and any-opts target
22810 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
22812 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
22814 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
22816 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
22818 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
22821 * config/i386/i386-options.c (ix86_option_override_internal):
22822 Error out when -m96bit-long-double is used with 64bit targets.
22823 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
22825 2021-04-23 Martin Liska <mliska@suse.cz>
22827 * lto-wrapper.c: Remove FIXME about usage of
22828 hardware_concurrency. The function is not on par with
22831 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
22834 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
22835 Copy operand 3 to operand 4. Use sse_reg_operand
22836 as operand 3 predicate.
22837 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
22838 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
22839 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
22840 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
22841 Copy operand 1 to operand 0.
22842 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
22843 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
22844 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
22846 2021-04-23 Alex Coplan <alex.coplan@arm.com>
22848 PR rtl-optimization/100230
22849 * early-remat.c (early_remat::sort_candidates): Use delete[]
22850 instead of delete for array allocated with new[].
22852 2021-04-23 Richard Biener <rguenther@suse.de>
22854 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
22855 (capture_info::capture_info): Likewise.
22856 (capture_info::walk_match): Likewise.
22857 (expr::gen_transform): Likewise.
22858 (dt_simplify::gen_1): Likewise.
22859 * gimple-match-head.c (maybe_resimplify_conditional_op):
22860 Remove VEC_COND_EXPR special-casing.
22861 (gimple_simplify): Likewise.
22862 * gimple.c (gimple_could_trap_p_1): Adjust.
22863 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
22864 to participate in PRE.
22866 2021-04-23 Richard Biener <rguenther@suse.de>
22868 * cfganal.c (connect_infinite_loops_to_exit): First call
22869 add_noreturn_fake_exit_edges.
22870 * ipa-sra.c (process_scan_results): Do not call the now redundant
22871 add_noreturn_fake_exit_edges.
22872 * predict.c (tree_estimate_probability): Likewise.
22873 (rebuild_frequencies): Likewise.
22874 * store-motion.c (one_store_motion_pass): Likewise.
22876 2021-04-23 Richard Biener <rguenther@suse.de>
22878 PR tree-optimization/100222
22879 * predict.c (pass_profile::execute): Remove redundant call to
22880 mark_irreducible_loops.
22881 (report_predictor_hitrates): Likewise.
22883 2021-04-23 Richard Biener <rguenther@suse.de>
22885 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
22886 valid_gimple_rhs_p by instead gimplifying to one.
22888 2021-04-23 Richard Biener <rguenther@suse.de>
22890 PR tree-optimization/99971
22891 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
22892 Always use TBAA for loads.
22894 2021-04-23 liuhongt <hongtao.liu@intel.com>
22897 * config/i386/i386-options.c (ix86_option_override_internal):
22898 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
22899 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
22900 by target attribute.
22902 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
22904 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
22906 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
22908 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
22910 * config.gcc (powerpc-ibm-aix6.*): Remove.
22911 * config/rs6000/aix61.h: Delete.
22913 2021-04-22 Martin Liska <mliska@suse.cz>
22915 PR testsuite/100159
22916 PR testsuite/100192
22917 * builtins.c (expand_builtin): Fix typos and missing comments.
22918 * dwarf2out.c (gen_subprogram_die): Likewise.
22919 (gen_struct_or_union_type_die): Likewise.
22921 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
22924 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
22925 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
22927 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
22929 * config/i386/darwin.h (TARGET_64BIT): Remove definition
22930 based on TARGET_ISA_64BIT.
22931 (TARGET_64BIT_P): Remove definition based on
22932 TARGET_ISA_64BIT_P().
22934 2021-04-21 Martin Liska <mliska@suse.cz>
22937 2021-04-21 Martin Liska <mliska@suse.cz>
22939 * lto-wrapper.c (cpuset_popcount): Remove.
22940 (init_num_threads): Remove and use hardware_concurrency.
22942 2021-04-21 Martin Liska <mliska@suse.cz>
22945 * main.c (main): Call toplev::finalize in CHECKING_P mode.
22946 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
22947 when incremental LTO linking happens.
22949 2021-04-21 Martin Liska <mliska@suse.cz>
22951 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
22952 makeserver cannot be detected, then use -flto=N fallback.
22954 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
22956 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
22957 default to yes for aarch64-linux-gnu.
22958 * configure: Regenerate.
22960 2021-04-21 Martin Liska <mliska@suse.cz>
22962 * lto-wrapper.c (cpuset_popcount): Remove.
22963 (init_num_threads): Remove and use hardware_concurrency.
22965 2021-04-21 Martin Liska <mliska@suse.cz>
22967 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
22968 which remains to be '(... || 0)' and clang complains about it.
22969 * dwarf2out.c (AT_vms_delta): Declare conditionally.
22970 (add_AT_vms_delta): Likewise.
22971 * tree.c (fld_simplified_type): Use rather more common pattern
22972 for disabling of something (#if 0).
22973 (get_tree_code_name): Likewise.
22974 (verify_type_variant): Likewise.
22976 2021-04-21 Martin Liska <mliska@suse.cz>
22978 * config/i386/i386-expand.c (decide_alignment): Use newly named
22979 macro TARGET_CPU_P.
22980 * config/i386/i386.c (ix86_decompose_address): Likewise.
22981 (ix86_address_cost): Likewise.
22982 (ix86_lea_outperforms): Likewise.
22983 (ix86_avoid_lea_for_addr): Likewise.
22984 (ix86_add_stmt_cost): Likewise.
22985 * config/i386/i386.h (TARGET_*): Remove.
22986 (TARGET_CPU_P): New macro.
22987 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
22988 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
22989 (swap_top_of_ready_list): Likewise.
22990 (ix86_atom_sched_reorder): Likewise.
22991 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
22992 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
22994 2021-04-21 Martin Liska <mliska@suse.cz>
22996 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
22998 (SET_TARGET_NO_SAHF): Likewise.
22999 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
23000 (SET_TARGET_PREFETCH_SSE): Likewise.
23001 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
23002 (SET_TARGET_NO_TUNE): Likewise.
23003 (TARGET_EXPLICIT_NO_80387_P): Likewise.
23004 (SET_TARGET_NO_80387): Likewise.
23006 * config/i386/i386.h (TARGET_*): Remove.
23007 * opth-gen.awk: Generate new used macros.
23009 2021-04-21 Martin Liska <mliska@suse.cz>
23011 * config/i386/i386.h (PTA_*): Remove.
23012 (enum pta_flag): New.
23013 (DEF_PTA): Generate PTA_* values from i386-isa.def.
23014 * config/i386/i386-isa.def: New file.
23016 2021-04-21 Alex Coplan <alex.coplan@arm.com>
23019 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
23020 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
23021 jump table targets.
23023 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
23025 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
23026 x86_64-*-* targets.
23027 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
23029 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
23030 (ix86_handle_option): Handle -mmwait.
23031 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
23032 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
23033 __builtin_ia32_monitor and __builtin_ia32_mwait.
23034 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
23035 (ix86_valid_target_attribute_inner_p): Likewise.
23036 (ix86_option_override_internal): Enable mwait/monitor
23037 instructions for -msse3.
23038 * config/i386/i386.h (TARGET_MWAIT): New.
23039 (TARGET_MWAIT_P): Likewise.
23040 * config/i386/i386.opt: Add -mmwait.
23041 * config/i386/mwaitintrin.h: New file.
23042 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
23043 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
23045 (@sse3_monitor_<mode>): Likewise.
23046 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
23047 * doc/extend.texi: Document mwait target attribute.
23048 * doc/invoke.texi: Document -mmwait.
23050 2021-04-21 Martin Liska <mliska@suse.cz>
23052 * config/i386/i386-options.c (DEF_ENUM): Remove it.
23053 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
23054 * config/i386/stringop.def (DEF_ENUM): Likewise.
23056 2021-04-21 Martin Liska <mliska@suse.cz>
23058 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
23059 of print_generic_expr.
23061 2021-04-21 Jakub Jelinek <jakub@redhat.com>
23063 PR rtl-optimization/100148
23064 * cprop.c (constprop_register): Use next_nondebug_insn instead of
23067 2021-04-21 Martin Liska <mliska@suse.cz>
23070 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
23071 free_dominance_info calls.
23073 2021-04-21 Richard Biener <rguenther@suse.de>
23075 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
23076 parameter (and assume it to be false).
23077 (fold_gimple_assign): Adjust, remove all callers of
23078 maybe_fold_reference calling it with is_lhs true.
23079 (gimple_fold_call): Likewise.
23080 (fold_stmt_1): Likewise.
23082 2021-04-21 Richard Biener <rguenther@suse.de>
23084 * fold-const.c (pedantic_non_lvalue_loc): Remove.
23085 (fold_binary_loc): Adjust.
23086 (fold_ternary_loc): Likewise.
23088 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
23090 PR middle-end/100130
23091 * varasm.c (get_block_for_decl): Make sure that any use of the
23092 retain attribute matches the section's retain flag.
23093 (switch_to_section): Check for retain mismatches even when
23094 changing sections, but do not warn if the given decl is the
23095 section's named.decl.
23096 (output_object_block): Pass the first decl in the block (if any)
23097 to switch_to_section.
23099 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
23101 * config/i386/i386-c.c (ix86_target_macros_internal): Define
23102 __CRC32__ for -mcrc32.
23103 * config/i386/i386-options.c (ix86_option_override_internal):
23104 Enable crc32 instruction for -msse4.2.
23105 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
23107 (sse4_2_crc32di): Likewise.
23108 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
23111 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
23114 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
23117 2021-04-20 Martin Liska <mliska@suse.cz>
23119 * doc/invoke.texi: Fix typo.
23120 * params.opt: Likewise.
23122 2021-04-20 Martin Liska <mliska@suse.cz>
23124 * doc/invoke.texi: Document new param.
23126 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
23128 PR tree-optimization/100081
23129 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
23130 rather than gori_compute_cache.
23131 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
23132 (range_def_chain::m_logical_depth): New member.
23133 (range_def_chain::range_def_chain): Initialize m_logical_depth.
23134 (range_def_chain::get_def_chain): Don't build defchains through more
23135 than LOGICAL_LIMIT logical expressions.
23136 * params.opt (param_ranger_logical_depth): New.
23138 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
23141 * config/arm/arm.c (arm_configure_build_target): Do not strip
23142 extended FPU/SIMD feature bits from the target ISA when -mfpu
23143 is specified (partial revert of r11-8168).
23145 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
23147 * params.opt (-param=openacc-kernels=): Add.
23148 * omp-oacc-kernels-decompose.cc
23149 (pass_omp_oacc_kernels_decompose::gate): Use it.
23150 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
23151 (--param): ... here, 'openacc-kernels'.
23153 2021-04-19 Martin Liska <mliska@suse.cz>
23156 * gengtype.c (finish_root_table): Align function arguments
23157 in between declaration and definition.
23159 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
23161 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
23162 frames larger than the SEH maximum frame size.
23164 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
23166 PR rtl-optimization/99927
23167 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
23168 is dead, just drop it.
23170 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23173 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
23174 * doc/tm.texi: Regenerate.
23175 * doc/tm.texi.in (D language and ABI): Add @hook for
23176 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
23178 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23180 * config/darwin-d.c (darwin_d_handle_target_object_format): New
23182 (darwin_d_register_target_info): New function.
23183 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23184 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
23186 (dragonfly_d_register_target_info): New function.
23187 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23188 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
23190 (freebsd_d_register_target_info): New function.
23191 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23192 * config/glibc-d.c (glibc_d_handle_target_object_format): New
23194 (glibc_d_register_target_info): New function.
23195 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23196 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
23198 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
23199 as handler for objectFormat key.
23200 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
23202 (winnt_d_register_target_info): New function.
23203 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23204 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
23206 (netbsd_d_register_target_info): New function.
23207 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23208 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
23210 (openbsd_d_register_target_info): New function.
23211 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23212 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
23213 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
23214 handler for objectFormat key.
23215 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
23217 (rs6000_d_register_target_info): Add
23218 rs6000_d_handle_target_object_format as handler for objectFormat key.
23219 * config/sol2-d.c (solaris_d_handle_target_object_format): New
23221 (solaris_d_register_target_info): New function.
23222 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23224 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23227 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
23228 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
23230 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
23231 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
23232 the old and new alignment after applying MIN/MAX to it is different.
23234 2021-04-16 Tamar Christina <tamar.christina@arm.com>
23237 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
23238 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
23240 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
23242 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
23244 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
23245 this section and its subsections.
23247 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23250 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
23251 define_insn patterns.
23253 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23255 PR rtl-optimization/98689
23256 * reg-notes.def (UNTYPED_CALL): New note.
23257 * combine.c (distribute_notes): Handle it.
23258 * emit-rtl.c (try_split): Likewise.
23259 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
23260 that calls with the note implicitly set all return value registers.
23261 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
23264 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23266 PR rtl-optimization/99596
23267 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
23268 register accesses for const calls. Assume that pure functions
23269 can only read from global registers. Ignore cases in which
23270 the stack pointer has been marked global.
23272 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23275 * tree-vect-loop.c (vect_transform_loop): Don't remove just
23276 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
23277 them by their last argument.
23279 2021-04-15 Martin Liska <mliska@suse.cz>
23281 * doc/invoke.texi: Other params don't use it, remove it.
23283 2021-04-15 Richard Biener <rguenther@suse.de>
23285 * gimple-builder.h: Add deprecation note.
23287 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23290 * attribs.h (restrict_type_identity_attributes_to): Declare.
23291 * attribs.c (restrict_type_identity_attributes_to): New function.
23293 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23296 * attribs.h (affects_type_identity_attributes): Declare.
23297 * attribs.c (remove_attributes_matching): New function.
23298 (affects_type_identity_attributes): Likewise.
23300 2021-04-15 Jakub Jelinek <jakub@redhat.com>
23303 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
23304 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
23305 ZERO_EXTEND, SIGN_EXTEND or AND.
23307 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23309 PR rtl-optimization/99929
23310 * rtl.h (same_vector_encodings_p): New function.
23311 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
23312 * cselib.c (rtx_equal_for_cselib_1): Likewise.
23313 * jump.c (rtx_renumbered_equal_p): Likewise.
23314 * lra-constraints.c (operands_match_p): Likewise.
23315 * reload.c (operands_match_p): Likewise.
23316 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
23318 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23320 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
23321 more information about variable-length CONST_VECTORs.
23323 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
23325 PR rtl-optimization/100066
23326 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
23327 ordered modes when choosing splitting mode for hard reg.
23329 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23332 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
23334 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
23336 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
23338 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
23339 for mask operand types.
23340 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
23341 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
23343 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
23345 * config/s390/s390.c (s390_const_operand_ok): Check the new
23346 operand types and generate a list of valid values.
23348 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23350 * doc/tm.texi: Regenerate.
23351 * doc/tm.texi.in (D language and ABI): Add @hook for
23352 TARGET_D_REGISTER_OS_TARGET_INFO.
23354 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23356 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
23358 (aarch64_d_register_target_info): New function.
23359 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
23361 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
23363 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
23364 (arm_d_register_target_info): New function.
23365 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
23366 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23367 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
23368 (ix86_d_register_target_info): New function.
23369 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
23370 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23371 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
23372 (mips_d_register_target_info): New function.
23373 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
23374 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23375 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
23376 (pa_d_register_target_info): New function.
23377 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
23378 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23379 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
23381 (riscv_d_register_target_info): New function.
23382 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
23383 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23384 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
23386 (rs6000_d_register_target_info): New function.
23387 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
23389 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23390 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
23391 (s390_d_register_target_info): New function.
23392 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
23393 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23394 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
23396 (sparc_d_register_target_info): New function.
23397 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
23398 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23399 * doc/tm.texi: Regenerate.
23400 * doc/tm.texi.in (D language and ABI): Add @hook for
23401 TARGET_D_REGISTER_CPU_TARGET_INFO.
23403 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23405 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
23406 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
23407 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
23408 * doc/tm.texi: Regenerate.
23409 * doc/tm.texi.in (D language and ABI): Add @hook for
23410 TARGET_D_HAS_STDCALL_CONVENTION.
23412 2021-04-14 Richard Biener <rguenther@suse.de>
23414 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
23415 VEC_COND_EXPRs have a gimple_val condition.
23416 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
23417 can no longer have a GENERIC condition.
23419 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
23422 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
23423 from the isa_delta when -mfpu has been used.
23424 (arm_options_perform_arch_sanity_checks): It's the architecture that
23425 lacks an FPU not the processor.
23427 2021-04-13 Richard Biener <rguenther@suse.de>
23429 PR tree-optimization/100053
23430 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
23431 not use optimistic dominance queries for backedges to validate
23433 (dominated_by_p_w_unex): Add parameter to ignore executable
23434 state on backedges.
23435 (rpo_elim::eliminate_avail): Adjust.
23437 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23440 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
23441 *aarch64_bfxilsi_extrdi): New define_insn patterns.
23443 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23446 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
23447 outermode, return NULL if the result doesn't encode back to the
23448 original byte sequence.
23449 (simplify_gen_subreg): Don't create SUBREGs from constants to
23450 MODE_COMPOSITE_P outermode.
23452 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23454 PR rtl-optimization/99905
23455 * combine.c (expand_compound_operation): If pos + len > modewidth,
23456 perform the right shift by pos in inner_mode and then convert to mode,
23457 instead of trying to simplify a shift of rtx with inner_mode by pos
23458 as if it was a shift in mode.
23460 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23463 * combine.c (simplify_and_const_int_1): Don't optimize varop
23464 away if it has side-effects.
23466 2021-04-12 Martin Liska <mliska@suse.cz>
23468 * doc/extend.texi: Escape @smallexample content.
23470 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
23472 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
23473 alternative in order to load a DFP zero.
23475 2021-04-12 Martin Liska <mliska@suse.cz>
23477 * doc/extend.texi: Be more precise in documentation
23478 of symver attribute.
23480 2021-04-12 Martin Liska <mliska@suse.cz>
23483 * gimplify.c (gimplify_expr): Right now, we unpoison all
23484 variables before a goto <dest>. We should not do it if we are
23487 2021-04-12 Cui,Lili <lili.cui@intel.com>
23489 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
23491 * common/config/i386/i386-common.c (processor_names): Add
23493 (processor_alias_table): Add rocketlake.
23494 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
23495 INTEL_COREI7_ROCKETLAKE.
23496 * config.gcc: Add -march=rocketlake.
23497 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23499 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
23500 (processor_cost_table): Add rocketlake cost.
23501 * config/i386/i386.h (ix86_size_cost) : Define
23503 (processor_type) : Add PROCESSOR_ROCKETLAKE.
23504 (PTA_ROCKETLAKE): Ditto.
23505 * doc/extend.texi: Add rocketlake.
23506 * doc/invoke.texi: Add rocketlake.
23508 2021-04-12 Cui,Lili <lili.cui@intel.com>
23510 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
23511 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
23512 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
23513 * doc/invoke.texi: Change alderlake ISA list.
23515 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
23517 PR middle-end/98088
23518 * omp-expand.c (expand_oacc_collapse_init): Update condition in
23521 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
23524 * config/i386/serializeintrin.h (_serialize): Defined as macro.
23526 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23529 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
23530 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
23532 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23534 PR middle-end/99989
23535 * gimple-ssa-warn-alloca.c
23536 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
23537 0 with integer precision unconditionally.
23539 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23541 PR rtl-optimization/98601
23542 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
23543 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
23544 unaligned_mems handle VOIDmode like BLKmode.
23546 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
23549 * tree.c (free_lang_data_in_decl): Do not release body of
23550 declare_variant_alt.
23552 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23554 * config/aarch64/aarch64.c (aarch64_option_restore): If the
23555 architecture was specified explicitly and the tuning wasn't,
23556 tune for the architecture rather than the configured default CPU.
23558 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23560 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
23561 as the temporary register.
23563 2021-04-09 Martin Liska <mliska@suse.cz>
23565 * doc/extend.texi: Move non-target attributes on the top level.
23567 2021-04-09 Martin Liska <mliska@suse.cz>
23569 * doc/invoke.texi: Document minimum and maximum value of the
23570 argument for both supported compression algorithms.
23572 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
23574 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
23575 TLS BSS before TLS data.
23576 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
23578 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23580 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
23582 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23584 * match.pd: Extend vec_cond folds to handle shifts.
23586 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
23588 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
23591 2021-04-08 Alex Coplan <alex.coplan@arm.com>
23594 * config/arm/iterators.md (MVE_vecs): New.
23595 (V_elem): Also handle V2DF.
23596 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
23597 (*mve_vdup<mode>): ... this. Remove second alternative since
23598 vec_duplicate of const_int is not canonical RTL, and we don't
23599 want to match symbol_refs.
23600 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
23602 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
23604 * fold-const.c (fold_single_bit_test): Fix typo.
23605 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
23608 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
23610 PR tree-optimization/97513
23611 * tree-vect-slp.c (vect_add_slp_permutation): New function,
23613 (vectorizable_slp_permutation): ...here. Detect cases in which
23614 all VEC_PERM_EXPRs are guaranteed to have the same stepped
23615 permute vector and only generate one permute vector for that case.
23616 Extend that case to handle variable-length vectors.
23618 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
23620 PR tree-optimization/99873
23621 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
23622 (vect_build_slp_instance): Don't split store groups that could
23623 use IFN_STORE_LANES.
23625 2021-04-07 Jakub Jelinek <jakub@redhat.com>
23628 * varasm.c (output_constant_pool_contents): Don't strip name encoding
23629 from XSTR (desc->sym, 0) or from label before passing those to
23632 2021-04-07 Richard Biener <rguenther@suse.de>
23634 PR tree-optimization/99954
23635 * tree-loop-distribution.c: Include tree-affine.h.
23636 (generate_memcpy_builtin): Try using tree-affine to prove
23638 (loop_distribution::classify_builtin_ldst): Always classify
23641 2021-04-07 Richard Biener <rguenther@suse.de>
23643 PR tree-optimization/99947
23644 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
23645 steps vector to avoid pushing elements from the reallocated
23648 2021-04-07 Richard Biener <rguenther@suse.de>
23650 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
23651 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
23653 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
23655 (debug_vn_reference_ops): New.
23657 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
23659 PR tree-optimization/98736
23660 * tree-loop-distribution.c
23661 * (loop_distribution::bb_top_order_init):
23662 Compute RPO with programing order preserved by calling function
23663 rev_post_order_and_mark_dfs_back_seme.
23665 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
23668 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
23669 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
23671 (process_bb_lives): Don't update biggest mode of hard reg for
23672 implicit in multi-register group. Use the new functions for
23673 updating dead_set and unused_set by register notes.
23675 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
23677 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
23680 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
23682 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
23683 (skylake_memset): Likewise.
23684 (skylake_cost): Change CLEAR_RATIO to 17.
23685 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
23686 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
23687 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
23689 2021-04-06 Richard Biener <rguenther@suse.de>
23691 PR tree-optimization/99880
23692 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
23693 set vectorized defs of relevant PHIs.
23695 2021-04-06 Richard Biener <rguenther@suse.de>
23697 PR tree-optimization/99924
23698 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
23699 nodes w/o scalar stmts as visited.
23701 2021-04-06 Alex Coplan <alex.coplan@arm.com>
23704 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
23705 PCS for [su]fix_optab.
23707 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
23709 * config/darwin.c (machopic_legitimize_pic_address): Check
23710 that the current pic register is one of the hard reg set
23711 before setting liveness.
23713 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
23715 * config/darwin.c (machopic_legitimize_pic_address): Fix
23716 whitespace, remove unused code.
23718 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23720 PR tree-optimization/99882
23721 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
23724 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23726 PR rtl-optimization/99863
23727 * dse.c (replace_read): Drop regs_live argument. Instead of
23728 regs_live, use store_insn->fixed_regs_live if non-NULL,
23729 otherwise punt if insns sequence clobbers or sets any hard
23732 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23735 * targhooks.h (default_print_patchable_function_entry_1): Declare.
23736 * targhooks.c (default_print_patchable_function_entry_1): New function,
23737 copied from default_print_patchable_function_entry with an added flags
23739 (default_print_patchable_function_entry): Rewritten into a small
23740 wrapper around default_print_patchable_function_entry_1.
23741 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23743 (rs6000_print_patchable_function_entry): New function.
23745 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
23747 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
23749 2021-04-01 Jason Merrill <jason@redhat.com>
23752 * common.opt: Document v15 and v16.
23754 2021-04-01 Richard Biener <rguenther@suse.de>
23756 PR tree-optimization/99863
23757 * gimplify.c (gimplify_init_constructor): Recompute vector
23760 2021-04-01 Jakub Jelinek <jakub@redhat.com>
23762 * doc/extend.texi (symver attribute): Fix up syntax errors
23765 2021-04-01 Jakub Jelinek <jakub@redhat.com>
23767 PR tree-optimization/96573
23768 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
23769 also pointer types.
23771 2021-04-01 Richard Biener <rguenther@suse.de>
23773 PR tree-optimization/99856
23774 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
23775 precision to vector element precision.
23777 2021-04-01 Martin Jambor <mjambor@suse.cz>
23779 PR tree-optimization/97009
23780 * tree-sra.c (access_or_its_child_written): New function.
23781 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
23784 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
23787 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
23789 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
23792 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
23793 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
23794 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
23795 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
23796 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
23797 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
23799 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
23800 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
23801 (define_attr "prefixed"): Update initializer.
23803 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23806 * dwarf2out.c (debug_ranges_dwo_section): New variable.
23807 (DW_RANGES_IDX_SKELETON): Define.
23808 (struct dw_ranges): Add begin_entry and end_entry members.
23809 (DEBUG_DWO_RNGLISTS_SECTION): Define.
23810 (add_ranges_num): Adjust r initializer for addition of *_entry
23812 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
23813 set idx to DW_RANGES_IDX_SKELETON.
23814 (use_distinct_base_address_for_range): New function.
23815 (index_rnglists): Don't set r->idx if it is equal to
23816 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
23817 r->end_entry for -gsplit-dwarf if those will be needed by
23819 (output_rnglists): Add DWO argument. If true, switch to
23820 debug_ranges_dwo_section rather than debug_ranges_section.
23821 Adjust l1/l2 label indexes. Only output the offset table when
23822 dwo is true and don't include in there the skeleton range
23823 entry if present. For -gsplit-dwarf, skip ranges that belong
23824 to the other rnglists section. Change return type from void
23825 to bool and return true if there are any range entries for
23826 the other section. For dwarf_split_debug_info use
23827 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
23828 entries instead of DW_RLE_start_end, DW_RLE_start_length and
23829 DW_RLE_base_address. Use use_distinct_base_address_for_range.
23830 (init_sections_and_labels): Initialize debug_ranges_dwo_section
23831 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
23832 and range_base_label indexes.
23833 (dwarf2out_finish): Call index_rnglists earlier before finalizing
23834 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
23835 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
23836 with different dwo arguments.
23837 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
23839 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23841 PR tree-optimization/98268
23842 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
23843 recompute_tree_invariant_for_addr_expr after successfully
23844 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
23846 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23848 PR tree-optimization/99726
23849 * tree-data-ref.c (create_intersect_range_checks_index): Bail
23850 out if there is more than one access function SCEV for the loop
23853 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23855 PR rtl-optimization/97141
23856 PR rtl-optimization/98726
23857 * emit-rtl.c (valid_for_const_vector_p): Return true for
23859 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
23860 poly_wide_int instead of a wide_int.
23861 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
23863 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
23864 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
23865 false for CONST_VECTORs that cannot be forced to memory.
23866 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
23867 is too complex to force to memory, build it up from individual
23870 2021-03-31 Jan Hubicka <jh@suse.cz>
23873 * cgraph.c (cgraph_node::release_body): Fix overactive check.
23875 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
23878 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
23881 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
23883 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
23884 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
23886 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
23887 "rep movsb/stosb" only for known sizes.
23888 * config/i386/i386-options.c (processor_cost_table): Use Ice
23889 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
23890 Rapids and Alder Lake.
23891 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
23892 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
23893 (icelake_memset): Likewise.
23894 (icelake_cost): Likewise.
23895 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
23898 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23901 * config/aarch64/aarch64.c
23902 (aarch64_vectorize_preferred_vector_alignment): Query the size
23903 of the provided SVE vector; do not assume that all SVE vectors
23904 have the same size.
23906 2021-03-31 Jan Hubicka <jh@suse.cz>
23909 * cgraph.c (cgraph_node::release_body): Remove all callers and
23911 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
23912 * cgraphunit.c (cgraph_node::expand): And here.
23914 2021-03-31 Martin Liska <mliska@suse.cz>
23916 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
23917 and one negated condition.
23919 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23920 Richard Sandiford <richard.sandiford@arm.com>
23923 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
23924 constraints on operands[2] and similarly 0 and rk constraints
23925 on operands[1] corresponding to that.
23927 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23930 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
23931 linker doesn't support DWARF sections new in DWARF5.
23932 * config/i386/i386-options.c (ix86_option_override_internal): Default
23933 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
23935 * config.in: Regenerated.
23936 * configure: Regenerated.
23938 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23941 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
23942 available issue_info before using it.
23944 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23947 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
23950 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
23953 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
23954 (altivec_lvsl_reg_<mode>): ... this.
23955 (altivec_lvsr_reg): Change to ...
23956 (altivec_lvsr_reg_<mode>): ... this.
23957 * config/rs6000/predicates.md (vec_set_index_operand): New.
23958 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23959 Enable 32bit variable vec_insert for all TARGET_VSX.
23960 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
23961 Enable 32bit variable vec_insert for p9 and above.
23962 (rs6000_expand_vector_set_var_p8): Rename to ...
23963 (rs6000_expand_vector_set_var_p7): ... this.
23964 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
23966 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
23967 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
23968 gen_altivec_lvsr_reg_di.
23970 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
23973 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
23974 (__rdtscp): Likewise.
23976 2021-03-30 Tamar Christina <tamar.christina@arm.com>
23978 PR tree-optimization/99825
23979 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
23980 Reject non-mult 2 lanes.
23982 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
23985 * config/arm/arm.c (arm_file_start): Fix emission of
23986 Tag_ABI_VFP_args attribute.
23988 2021-03-30 Richard Biener <rguenther@suse.de>
23990 PR tree-optimization/99824
23991 * stor-layout.c (set_min_and_max_values_for_integral_type):
23992 Assert the precision is within the bounds of
23993 WIDE_INT_MAX_PRECISION.
23994 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
23995 the outermost component ref only to lower the access size
23996 and initialize that from the access type.
23998 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
24001 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
24002 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
24004 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
24006 * config/aarch64/aarch64.md
24007 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
24008 attribute to disambiguate between SIMD and FP variants of the
24011 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
24013 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
24014 (analyze_ssa_name_flags): Fix typo in comment.
24016 2021-03-29 Alex Coplan <alex.coplan@arm.com>
24019 * config/aarch64/aarch64-sve-builtins.cc
24020 (function_builder::add_function): Add placeholder_p argument, use
24021 placeholder decls if this is set.
24022 (function_builder::add_unique_function): Instead of conditionally adding
24023 direct overloads, unconditionally add either a direct overload or a
24025 (function_builder::add_overloaded_function): Set placeholder_p if we're
24026 using C++ overloads. Use the obstack for string storage instead
24027 of relying on the tree nodes.
24028 (function_builder::add_overloaded_functions): Don't return early for
24029 m_direct_overloads: we need to add placeholders.
24030 * config/aarch64/aarch64-sve-builtins.h
24031 (function_builder::add_function): Add placeholder_p argument.
24033 2021-03-29 Richard Biener <rguenther@suse.de>
24035 PR tree-optimization/99807
24036 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
24037 assert below VEC_PERM handling.
24039 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24042 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
24043 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
24044 matching const_int 0.
24045 (move_lo_quad_internal_be_<mode>): Likewise.
24046 (move_lo_quad_<mode>): Update for the above.
24047 * config/aarch64/iterators.md (VQ_2E): Delete.
24049 2021-03-29 Jakub Jelinek <jakub@redhat.com>
24051 PR tree-optimization/99777
24052 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
24053 types other than scalar integral types.
24055 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
24057 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
24058 XCOFF TLS reloc decorations.
24060 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
24062 * doc/analyzer.texi (Analyzer Internals): Update link to
24063 "A Memory Model for Static Analysis of C Programs".
24065 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
24067 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
24068 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
24070 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
24071 (rs6000_special_round_type_align): Recursively check innermost first
24074 2021-03-26 Jakub Jelinek <jakub@redhat.com>
24077 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
24078 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
24079 assignment with drap_reg active, queue reg save for hfp with offset 0
24080 and flush queued reg saves. When handling a push with rule18,
24081 defer queueing reg save for hfp and just assert the offset is 0.
24082 (scan_trace): Assert that fde->rule18 is false.
24084 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
24087 * ira-costs.c (record_reg_classes): Put case with
24088 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
24089 * ira.c (ira_setup_alts): Ditto.
24090 * lra-constraints.c (process_alt_operands): Ditto.
24091 * recog.c (asm_operand_ok): Ditto.
24092 * reload.c (find_reloads): Ditto.
24094 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24096 * config/aarch64/aarch64-protos.h
24097 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
24098 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
24099 * config/aarch64/aarch64.c (generic_addrcost_table): Update
24100 accordingly, using the same costs as for post_modify.
24101 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
24102 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
24103 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
24104 (a64fx_addrcost_table): Likewise.
24105 (neoversev1_addrcost_table): New.
24106 (neoversev1_tunings): Use neoversev1_addrcost_table.
24107 (aarch64_address_cost): Use the new post_modify costs for CImode
24110 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24112 * config/aarch64/aarch64.opt
24113 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
24114 * doc/invoke.texi: Document it.
24115 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
24116 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
24117 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
24118 (aarch64_vec_issue_info): New structures.
24119 (cpu_vector_cost): Write comments above the variables rather
24121 (cpu_vector_cost::issue_info): New member variable.
24122 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
24123 and tree-ssa-loop-niter.h.
24124 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
24125 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
24126 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
24127 (thunderx3t110_vector_cost): Initialize issue_info to null.
24128 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
24129 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
24130 (neoversev1_vector_cost): Use them.
24131 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
24132 (aarch64_vector_costs::saw_sve_only_op): New member variable.
24133 (aarch64_vector_costs::num_vector_iterations): Likewise.
24134 (aarch64_vector_costs::scalar_ops): Likewise.
24135 (aarch64_vector_costs::advsimd_ops): Likewise.
24136 (aarch64_vector_costs::sve_ops): Likewise.
24137 (aarch64_vector_costs::seen_loads): Likewise.
24138 (aarch64_simd_vec_costs_for_flags): New function.
24139 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
24140 Count the number of predicate operations required by SVE WHILE
24142 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
24143 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
24144 (aarch64_count_ops): Likewise.
24145 (aarch64_add_stmt_cost): Record whether see an SVE operation
24146 that cannot currently be implementing using Advanced SIMD.
24147 Record issue information about the scalar, Advanced SIMD
24148 and (where relevant) SVE versions of a loop.
24149 (aarch64_vec_op_count::dump): New function.
24150 (aarch64_sve_op_count::dump): Likewise.
24151 (aarch64_estimate_min_cycles_per_iter): Likewise.
24152 (aarch64_adjust_body_cost): If issue information is available,
24153 try to compare the issue rates of the various loop implementations
24154 and increase or decrease the vector body cost accordingly.
24156 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24158 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
24159 Assume a zero cost for induction phis.
24161 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24163 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
24165 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
24166 vector comparisons.
24168 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24170 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
24172 (aarch64_add_stmt_cost): Call it.
24174 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24176 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
24177 New tuning parameter.
24178 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
24179 (aarch64_estimated_sve_vq): New function.
24180 (aarch64_vector_costs::analyzed_vinfo): New member variable.
24181 (aarch64_vector_costs::is_loop): Likewise.
24182 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
24183 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
24184 (aarch64_record_potential_advsimd_unrolling): New function.
24185 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
24186 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
24187 aarch64_analyze_bb_vinfo on the first use of a costs structure.
24188 Detect whether we're vectorizing a loop for SVE that might be
24189 completely unrolled if it used Advanced SIMD instead.
24190 (aarch64_adjust_body_cost_for_latency): New function.
24191 (aarch64_finish_cost): Call it.
24193 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24195 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
24196 (aarch64_init_cost): New function.
24197 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
24198 the default unsigned[3].
24199 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
24200 (TARGET_VECTORIZE_INIT_COST): Override.
24201 (TARGET_VECTORIZE_FINISH_COST): Likewise.
24202 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
24204 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24206 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
24207 (neoversev1_sve_vector_cost): New cost structures.
24208 (neoversev1_vector_cost): Likewise.
24209 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
24211 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24213 * config/aarch64/aarch64-protos.h
24214 (sve_vec_cost::scatter_store_elt_cost): New member variable.
24215 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24216 accordingly, taking the cost from the cost of a scalar_store.
24217 (a64fx_sve_vector_cost): Likewise.
24218 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
24220 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24222 * config/aarch64/aarch64-protos.h
24223 (simd_vec_cost::store_elt_extra_cost): New member variable.
24224 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24225 accordingly, using the vec_to_scalar cost for the new field.
24226 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24227 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24228 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24229 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24230 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24231 (thunderx3t110_advsimd_vector_cost): Likewise.
24232 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
24234 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24236 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
24237 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
24238 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
24239 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24240 accordingly, using zero for the new costs.
24241 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24242 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24243 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24244 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24245 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24246 (thunderx3t110_advsimd_vector_cost): Likewise.
24247 (aarch64_ld234_st234_vectors): New function.
24248 (aarch64_adjust_stmt_cost): Likewise.
24249 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
24250 the new vector costs.
24252 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24254 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
24255 derived class of simd_vec_cost. Add information about CLAST[AB]
24256 and FADDA instructions.
24257 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24258 accordingly, using the vec_to_scalar costs for the new fields.
24259 (a64fx_sve_vector_cost): Likewise.
24260 (aarch64_reduc_type): New function.
24261 (aarch64_sve_in_loop_reduction_latency): Likewise.
24262 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
24263 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
24264 that occur in the loop body.
24265 (aarch64_add_stmt_cost): Update call accordingly.
24267 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24269 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
24271 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
24272 above the fields rather than to the right.
24273 (simd_vec_cost::reduc_i8_cost): New member variable.
24274 (simd_vec_cost::reduc_i16_cost): Likewise.
24275 (simd_vec_cost::reduc_i32_cost): Likewise.
24276 (simd_vec_cost::reduc_i64_cost): Likewise.
24277 (simd_vec_cost::reduc_f16_cost): Likewise.
24278 (simd_vec_cost::reduc_f32_cost): Likewise.
24279 (simd_vec_cost::reduc_f64_cost): Likewise.
24280 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24281 accordingly, using the vec_to_scalar_cost for the new fields.
24282 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24283 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24284 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24285 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24286 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24287 (thunderx3t110_advsimd_vector_cost): Likewise.
24288 (aarch64_use_new_vector_costs_p): New function.
24289 (aarch64_simd_vec_costs): New function, split out from...
24290 (aarch64_builtin_vectorization_cost): ...here.
24291 (aarch64_is_reduction): New function.
24292 (aarch64_detect_vector_stmt_subtype): Likewise.
24293 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
24294 using the new vector costs.
24296 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24299 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
24300 TLS declarations as public.
24302 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24304 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
24305 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
24306 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
24307 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
24308 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
24309 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
24310 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
24311 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
24312 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
24314 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24317 * config.gcc (*-*-cygwin*): Add winnt-d.o
24318 (*-*-mingw*): Likewise.
24319 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
24320 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
24321 * config/i386/t-cygming: Add winnt-d.o.
24322 * config/i386/winnt-d.c: New file.
24324 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24326 * config/freebsd-d.c: Include memmodel.h.
24328 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24331 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
24332 * config/t-openbsd: Add openbsd-d.o.
24333 * config/openbsd-d.c: New file.
24335 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
24337 PR tree-optimization/96974
24338 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
24339 with graceful exit.
24341 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24344 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24348 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24349 always_inline in system headers.
24351 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
24353 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
24355 2021-03-25 Jakub Jelinek <jakub@redhat.com>
24358 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
24359 * fold-const.c (operand_compare::operand_equal_p): Don't compare
24360 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
24362 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24366 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24367 always_inline in system headers.
24369 2021-03-25 Richard Biener <rguenther@suse.de>
24371 PR tree-optimization/99746
24372 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
24373 the scalar stmt as patterned. Instead set up required things
24376 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
24378 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
24381 2021-03-24 Martin Liska <mliska@suse.cz>
24384 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
24386 * config/i386/i386-options.c (ix86_option_override_internal):
24387 Add run-time assert.
24389 2021-03-24 Martin Jambor <mjambor@suse.cz>
24392 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
24393 parameters with unknown type.
24394 (ipacp_value_safe_for_type): New function.
24395 (propagate_vals_across_arith_jfunc): Verify that the constant type
24396 can be used for a type of the formal parameter.
24397 (propagate_vals_across_ancestor): Likewise.
24398 (propagate_scalar_across_jump_function): Likewise. Pass the type
24399 also to propagate_vals_across_ancestor.
24401 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
24404 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
24406 (movmisalign<mode>_mve_load): Likewise.
24408 2021-03-24 Jakub Jelinek <jakub@redhat.com>
24411 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
24412 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
24414 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24416 * doc/sourcebuild.texi (sysconf): New effective target.
24418 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24420 * config/i386/predicates.md (reg_or_const_vec_operand): New.
24421 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
24422 the now *-prefixed insn_and_split, turn the splitter const vec
24423 into an input for the insn, making it an ignored immediate for
24424 non-split cases, and loaded into the scratch register
24427 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
24430 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
24431 Use define_relaxed_memory_constraint for them.
24433 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
24436 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
24437 colon to the diagnostic message.
24439 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
24441 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
24443 (try_fwprop_subst_note): Use set_info instead of insn_info.
24444 (try_fwprop_subst_pattern): Likewise.
24445 (try_fwprop_subst_notes): Likewise.
24446 (try_fwprop_subst): Likewise.
24447 (forward_propagate_subreg): Likewise.
24448 (forward_propagate_and_simplify): Likewise.
24449 (forward_propagate_into): Likewise.
24450 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
24452 (set_info::single_nondebug_insn_use): Likewise.
24453 (set_info::single_phi_use): Likewise.
24454 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
24456 (set_info::single_nondebug_insn_use): Likewise.
24457 (set_info::single_phi_use): Likewise.
24459 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
24461 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
24463 2021-03-23 Jakub Jelinek <jakub@redhat.com>
24466 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
24467 expand_mult to perform an unsigned rather than a signed
24470 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
24473 * config/i386/cpuid.h (__cpuid): Add __volatile__.
24474 (__cpuid_count): Likewise.
24476 2021-03-23 Richard Biener <rguenther@suse.de>
24478 PR tree-optimization/99721
24479 * tree-vect-slp.c (vect_slp_analyze_node_operations):
24480 Make sure we can schedule the node.
24482 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24484 * config/riscv/riscv.c (riscv_subword): Take endianness into
24485 account when calculating the byte offset.
24487 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24489 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
24490 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
24491 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
24492 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
24493 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
24494 new predicate "subreg_lowpart_operator"
24496 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24498 * config/riscv/riscv.c (riscv_swap_instruction): New function
24499 to byteswap an SImode rtx containing an instruction.
24500 (riscv_trampoline_init): Byteswap the generated instructions
24503 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24505 * common/config/riscv/riscv-common.c
24506 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
24507 * config.gcc (riscv32be-*, riscv64be-*): Set
24508 TARGET_BIG_ENDIAN_DEFAULT to 1.
24509 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
24510 depending on default endianness.
24511 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24512 * config/riscv/linux.h (LINK_SPEC): Likewise.
24513 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
24514 default endianness.
24515 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
24517 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24519 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
24520 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24521 * config/riscv/linux.h (LINK_SPEC): Likewise.
24522 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
24524 (BYTES_BIG_ENDIAN): Handle big endian.
24525 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
24526 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
24528 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
24530 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
24532 * regcprop.c (find_oldest_value_reg): Ask target whether
24533 different mode is fine for replacement register.
24535 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
24537 PR tree-optimization/99296
24538 * value-range.cc (irange::irange_set_1bit_anti_range): New.
24539 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
24540 * value-range.h (irange::irange_set_1bit_anti_range): New.
24542 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
24545 * config/aarch64/constraints.md (UtQ): Use
24546 define_relaxed_memory_constraint for it.
24547 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
24548 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24549 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
24550 (have_relaxed_memory_constraints): New static var.
24551 (relaxed_memory_start, relaxed_memory_end): Ditto.
24552 (add_constraint): Add arg is_relaxed_memory. Check name for
24553 relaxed memory. Set up is_relaxed_memory in constraint_data and
24554 have_relaxed_memory_constraints. Adjust calls.
24555 (choose_enum_order): Process relaxed memory.
24556 (write_tm_preds_h): Ditto.
24557 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24558 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24559 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
24560 * ira-lives.c (single_reg_class): Use
24561 insn_extra_relaxed_memory_constraint.
24562 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
24563 * lra-constraints.c (valid_address_p): Use
24564 insn_extra_relaxed_memory_constraint instead of other memory
24566 (process_alt_operands): Process CT_RELAXED_MEMORY.
24567 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
24568 * recog.c (asm_operand_ok, preprocess_constraints): Process
24570 * reload.c (find_reloads): Ditto.
24571 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
24572 * stmt.c (parse_input_constraint): Use
24573 insn_extra_relaxed_memory_constraint.
24575 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
24578 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
24581 2021-03-22 Alex Coplan <alex.coplan@arm.com>
24584 * config/arm/arm-protos.h (neon_make_constant): Add generate
24585 argument to guard emitting insns, default to true.
24586 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
24587 CONST_VECTORs which neon_make_constant can't handle.
24588 (neon_vdup_constant): Add generate argument, avoid emitting
24589 insns if it's not set.
24590 (neon_make_constant): Plumb new generate argument through.
24591 * config/arm/constraints.md (Ui): New. Use it...
24592 * config/arm/mve.md (*mve_mov<mode>): ... here.
24593 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
24594 synthesize constants.
24596 2021-03-22 Richard Biener <rguenther@suse.de>
24598 * debug.h: Add deprecation warning.
24600 2021-03-22 Richard Biener <rguenther@suse.de>
24602 PR tree-optimization/99694
24603 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
24606 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
24609 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
24610 after type checking.
24612 2021-03-22 Jakub Jelinek <jakub@redhat.com>
24616 * dwarf2out.c (get_full_len): Use get_precision rather than
24618 (add_const_value_attribute): Make sure add_AT_wide argument has
24619 precision prec rather than some very wide one.
24621 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
24623 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
24624 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
24625 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
24626 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
24627 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
24628 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
24629 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
24630 *fix_trunc<mode>si2_internal): Fix empty split condition.
24631 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
24632 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
24633 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
24634 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
24636 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
24639 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
24640 Convert idx to DImode.
24641 (rs6000_expand_vector_set_var_p8): Likewise.
24643 2021-03-21 Jakub Jelinek <jakub@redhat.com>
24646 * dwarf2out.c (insert_float): Change return type from void to
24647 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
24648 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
24651 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
24654 * config/i386/i386.c (construct_container): Check cfun != NULL
24655 before accessing silent_p.
24657 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
24659 * asan.c: Fix typos in comments.
24661 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
24663 PR rtl-optimization/99680
24664 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
24665 (process_address_1): Check empty constraint before using
24668 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
24670 * config/rs6000/rs6000.c (power10_cost): New.
24671 (rs6000_option_override_internal): Set Power10 costs.
24672 (rs6000_issue_rate): Set Power10 issue rate.
24673 * config/rs6000/power10.md: Rewrite for Power10.
24675 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
24678 * lra-constraints.c (process_address_1): Don't use unknown
24679 constraint for address constraint.
24681 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
24684 * config.gcc (powerpc-*-darwin8): Delete the reference to
24685 the now removed darwin8.h.
24687 2021-03-19 Olivier Hainque <hainque@adacore.com>
24690 * config/vxworksae.h (VX_CPU_PREFIX): Define.
24692 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
24694 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
24696 2021-03-19 Tamar Christina <tamar.christina@arm.com>
24698 PR tree-optimization/99656
24699 * tree-vect-slp-patterns.c (linear_loads_p,
24700 complex_add_pattern::matches, is_eq_or_top,
24701 vect_validate_multiplication, complex_mul_pattern::matches,
24702 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
24703 * tree-vectorizer.h: (complex_load_perm_t): Removed.
24704 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
24705 complex_load_perm_t.
24707 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
24710 * config/i386/i386-options.c (ix86_init_machine_status): Set
24712 * config/i386/i386.c (init_cumulative_args): Set silent_p to
24714 (construct_container): Return early for return and argument
24715 errors if silent_p is true.
24716 * config/i386/i386.h (machine_function): Add silent_p.
24718 2021-03-19 Jakub Jelinek <jakub@redhat.com>
24721 * config/arm/constraints.md (Ds): New constraint.
24722 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
24723 constraint instead of w,Dm.
24725 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
24727 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
24730 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
24732 PR middle-end/99641
24733 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
24734 array type, do the computation of the current position in sizetype.
24736 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
24739 * lra-constraints.c (process_address_1): Use lookup_constraint
24740 only for a single constraint.
24742 2021-03-18 Martin Sebor <msebor@redhat.com>
24744 PR middle-end/99502
24745 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
24746 (inbounds_memaccess_p): ...to this. Check the ending offset of
24747 the accessed member.
24749 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
24751 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
24752 %> quote markers to error messages.
24753 (gcn_goacc_validate_dims): Likewise.
24754 (gcn_conditional_register_usage): Remove exclaimation mark from error
24756 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
24758 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
24760 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
24763 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
24764 Kito Cheng <kito.cheng@sifive.com>
24766 * config/riscv/riscv.c (riscv_block_move_straight): Change type
24767 to unsigned HOST_WIDE_INT for parameter and local variable with
24768 HOST_WIDE_INT type.
24769 (riscv_adjust_block_mem): Ditto.
24770 (riscv_block_move_loop): Ditto.
24771 (riscv_expand_block_move): Ditto.
24773 2021-03-18 Nick Clifton <nickc@redhat.com>
24775 * config/v850/v850.c (construct_restore_jr): Increase static
24777 (construct_save_jarl): Likewise.
24778 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
24780 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24782 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
24783 (aarch64_override_options_internal): Use it.
24784 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
24787 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
24789 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
24790 error message format issues.
24791 (nios2_option_override): Likewise.
24792 (nios2_expand_fpu_builtin): Likewise.
24793 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
24794 truncation warning.
24795 (nios2_expand_custom_builtin): More error message format fixes.
24796 (nios2_expand_rdwrctl_builtin): Likewise.
24797 (nios2_expand_rdprs_builtin): Likewise.
24798 (nios2_expand_eni_builtin): Likewise.
24799 (nios2_expand_builtin): Likewise.
24800 (nios2_register_custom_code): Likewise.
24801 (nios2_valid_target_attribute_rec): Likewise.
24802 (nios2_add_insn_asm): Fix uninitialized variable warning.
24804 2021-03-17 Jan Hubicka <jh@suse.cz>
24806 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
24807 of gather to match reality.
24808 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
24810 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24812 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
24813 to compare against CC_REG rather than NE.
24815 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
24818 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
24819 inline assembly statements.
24820 (ix86_print_operand): Update 'P' handling for -fno-plt.
24822 2021-03-17 Tamar Christina <tamar.christina@arm.com>
24825 * config/aarch64/aarch64.c
24826 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
24828 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
24831 * config/rs6000/predicates.md (branch_comparison_operator): Allow
24832 ordered and unordered for CCFPmode, if flag_finite_math_only.
24834 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24837 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
24838 rather than ASHIFT.
24839 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
24841 2021-03-16 Martin Liska <mliska@suse.cz>
24844 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
24845 cl_optimization_compare function.
24847 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
24849 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
24852 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24855 * config/i386/i386.h (struct machine_function): Add
24856 has_explicit_vzeroupper bitfield.
24857 * config/i386/i386-expand.c (ix86_expand_builtin): Set
24858 cfun->machine->has_explicit_vzeroupper when expanding
24859 IX86_BUILTIN_VZEROUPPER.
24860 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
24861 Do the mode switching only when TARGET_VZEROUPPER, expensive
24862 optimizations turned on and not optimizing for size.
24863 (pass_insert_vzeroupper::gate): Enable even when
24864 cfun->machine->has_explicit_vzeroupper is set.
24866 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24869 * config/aarch64/aarch64.c
24870 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
24871 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
24872 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
24874 2021-03-15 Richard Biener <rguenther@suse.de>
24876 PR tree-optimization/98834
24877 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
24878 subsetting by truncating the access size.
24880 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
24882 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
24883 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
24886 2021-03-15 Martin Liska <mliska@suse.cz>
24888 * spellcheck.c: Add missing comma in initialization.
24890 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
24892 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
24893 alternative 2 and alternative 1 with alternative 3 using
24894 YW register constraint.
24895 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
24896 using YW register constraint.
24897 (*vec_extractv16qi_zext): Ditto.
24898 (*vec_extractv4si): Merge alternatives 4 and 5
24899 using Yw register constraint.
24900 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
24902 2021-03-13 Martin Sebor <msebor@redhat.com>
24904 PR tree-optimization/99489
24905 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
24906 is not a call statement.
24908 2021-03-13 Jakub Jelinek <jakub@redhat.com>
24910 PR tree-optimization/99544
24911 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
24912 if for vector types multiplication can't be done in type's mode.
24914 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
24917 * config/sparc/constraints.md (w): Rename to...
24918 (W): ... this and ditch previous implementation.
24919 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
24920 (*movdf_insn_sp64): Likewise.
24921 (*mov<VM64:mode>_insn_sp64): Likewise.
24922 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
24924 (atomic_compare_and_swap_leon3_1): Likewise.
24925 (*atomic_compare_and_swapdi_v8plus): Likewise.
24926 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
24927 architecture and add missing address validity check during LRA.
24929 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
24932 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
24933 occuring for assumed-size arrays in use_device_{ptr,addr}.
24935 2021-03-12 Jakub Jelinek <jakub@redhat.com>
24938 * config/i386/constraints.md (YW): New internal constraint.
24939 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
24940 (*<sse2_avx2>_<insn><mode>3<mask_name>,
24941 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
24942 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
24944 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
24945 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
24946 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
24947 into one, use Yw instead of former x,v.
24948 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
24949 the last alternative.
24950 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
24951 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
24952 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
24953 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
24954 into one, use <v_Yw> instead of former x,v.
24955 (avx2_interleave_highv32qi<mask_name>,
24956 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
24957 constraints. Add && <mask_avx512bw_condition> to condition.
24958 (avx2_interleave_lowv32qi<mask_name>,
24959 vec_interleave_lowv16qi<mask_name>,
24960 avx2_interleave_highv16hi<mask_name>,
24961 vec_interleave_highv8hi<mask_name>,
24962 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
24963 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
24964 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
24965 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
24966 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
24967 Yw instead of v in constraints.
24968 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
24969 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
24970 instead of Yv in constraints.
24971 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
24972 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
24973 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
24974 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
24975 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
24977 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
24978 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
24979 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
24980 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
24981 two, one with just x, another isa avx512vl with v.
24983 2021-03-12 Martin Liska <mliska@suse.cz>
24985 * doc/invoke.texi: Add missing param documentation.
24987 2021-03-11 David Malcolm <dmalcolm@redhat.com>
24990 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
24991 analyzer/trimmed-graph.o.
24992 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
24993 feasibility checking to reflect new implementation.
24994 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
24996 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
24998 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25000 * digraph.cc (selftest::test_shortest_paths): Update
25001 shortest_paths init for new param. Add test of
25002 SPS_TO_GIVEN_TARGET.
25003 * shortest-paths.h (enum shortest_path_sense): New.
25004 (shortest_paths::shortest_paths): Add "sense" param.
25005 Update for renamings. Generalize to use "sense" param.
25006 (shortest_paths::get_shortest_path): Rename param.
25007 (shortest_paths::m_sense): New field.
25008 (shortest_paths::m_prev): Rename...
25009 (shortest_paths::m_best_edge): ...to this.
25010 (shortest_paths::get_shortest_path): Update for renamings.
25011 Conditionalize flipping of path on sense of traversal.
25013 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25015 * digraph.cc (selftest::test_shortest_paths): Add test coverage
25016 for paths from B and C.
25017 * shortest-paths.h (shortest_paths::shortest_paths): Handle
25018 unreachable nodes, rather than asserting.
25020 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
25023 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
25024 xcoff_tbss_section_name.
25025 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
25026 * xcoffout.c (xcoff_tbss_section_name): Delete.
25027 * xcoffout.h (xcoff_tbss_section_name): Delete.
25029 2021-03-11 Richard Biener <rguenther@suse.de>
25031 PR tree-optimization/99523
25032 * tree-cfg.c (dump_function_to_file): Dump SSA names
25033 w/o identifier to the decls section as well, not only those
25034 without a VAR_DECL.
25036 2021-03-11 Jakub Jelinek <jakub@redhat.com>
25039 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
25040 function calls with lhs fail if the lhs don't have compatible types.
25042 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
25044 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
25045 Change FRAME_POINTER_REGNUM to correspond to a new faked
25046 register faked_fp, part of GENNONACR_REGS like faked_ap.
25047 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
25048 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
25049 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
25050 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
25051 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
25053 (CRIS_REAL_FP_REGNUM): New constant.
25054 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
25055 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
25056 (cris_initial_elimination_offset): Handle elimination changes
25057 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
25058 and add one from FRAME_POINTER_REGNUM to
25059 HARD_FRAME_POINTER_REGNUM.
25060 (cris_expand_prologue, cris_expand_epilogue): Emit code for
25061 hard_frame_pointer_rtx instead of frame_pointer_rtx.
25063 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
25066 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
25067 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
25069 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
25072 * lra-constraints.c (process_address_1): Don't check unknown
25073 constraint, use X for empty constraint.
25075 2021-03-10 Alex Coplan <alex.coplan@arm.com>
25077 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
25078 Fix typo in comment describing "is_ha" argument.
25080 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
25082 * doc/sourcebuild.texi: Document LRA target selector.
25084 2021-03-10 David Malcolm <dmalcolm@redhat.com>
25086 * doc/ux.texi: Add subsection contrasting interactive versus
25087 batch usage of GCC.
25089 2021-03-10 Joel Hutton <joel.hutton@arm.com>
25092 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
25094 (vectorizable_load): Fix gather load mask check condition.
25096 2021-03-10 Richard Biener <rguenther@suse.de>
25098 PR tree-optimization/99510
25099 * tree.c (check_aligned_type): Check that the candidate
25100 has TYPE_USER_ALIGN set instead of matching with the
25103 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
25105 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
25106 float and vector integer modes only if the mode is not larger.
25108 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
25110 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
25112 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25114 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
25116 * ira-lives.c (single_reg_class): Ditto.
25118 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
25120 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
25121 the architecture-specific rtems.h.
25122 (aarch64-*-rtems*): Likewise.
25123 (arm*-*-rtems*): Likewise.
25124 (epiphany-*-rtems*): Likewise.
25125 (riscv*-*-rtems*): Likewise.
25127 2021-03-09 Jakub Jelinek <jakub@redhat.com>
25129 PR tree-optimization/99305
25130 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
25131 before integer_all_onesp instead of vice versa.
25133 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
25135 * common/config/arm/arm-common.c (arm_config_default): Change type
25136 of 'i' to unsigned.
25138 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25141 * lra-constraints.c (process_address_1): Process constraint 'g'
25142 separately and digital constraints containing more one digit.
25144 2021-03-09 Nick Clifton <nickc@redhat.com>
25146 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
25147 (DWARF"_DEBUGGING_INFO): Define.
25149 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
25152 * calls.c (initialize_argument_information): When the argument
25153 is passed by reference, do not make a copy in a thunk only if
25154 the argument is already in memory. Remove redundant test for
25155 the case of callee copy.
25157 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25160 * lra-constraints.c (process_address_1): Process 0..9 constraints
25161 in process_address_1.
25163 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
25165 * config/s390/s390.c (struct s390_processor processor_table):
25166 Binutils name string must not be empty.
25168 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
25170 * config/arc/arc.c (arc_attr_type): Remove function.
25172 2021-03-09 Martin Liska <mliska@suse.cz>
25175 * config/i386/i386-options.c (ix86_option_override_internal):
25176 Set isa_flags for OPTS argument and not for the global
25179 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25181 * config/rs6000/predicates.md (ds_form_mem_operand): Check
25184 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25187 * config/rs6000/predicates.md (ds_form_mem_operand) New
25189 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
25190 ds_form_mem_operand in ld/lwa patterns.
25191 * config/rs6000/fusion.md: Regenerate file.
25193 2021-03-08 Martin Sebor <msebor@redhat.com>
25195 PR middle-end/98266
25196 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
25197 (array_bounds_checker::check_array_bounds): Call it.
25199 2021-03-08 Martin Sebor <msebor@redhat.com>
25201 PR middle-end/97631
25202 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
25203 (handle_builtin_stxncpy_strncat): Rename locals. Determine
25204 destination size from allocation calls. Issue a more appropriate
25206 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
25207 (handle_builtin_memset): Same.
25209 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
25212 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
25213 to ensure we do not have an Altivec style address.
25214 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
25215 an Altivec style address.
25216 (*vsx_le_perm_store_<mode>): Likewise.
25217 (splitters after *vsx_le_perm_store_<mode>): Likewise.
25218 (vsx_load_<mode>): Disable special expander if passed an Altivec
25220 (vsx_store_<mode>): Likewise.
25222 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25225 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
25226 (aarch64_simd_shift_imm_vec_hi): Likewise.
25227 (aarch64_simd_shift_imm_vec_si): Likewise.
25228 (aarch64_simd_shift_imm_vec_di): Likewise.
25229 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
25230 predicate from above.
25231 (aarch64_shrn<mode>_insn_be): Likewise.
25232 (aarch64_rshrn<mode>_insn_le): Likewise.
25233 (aarch64_rshrn<mode>_insn_be): Likewise.
25234 (aarch64_shrn2<mode>_insn_le): Likewise.
25235 (aarch64_shrn2<mode>_insn_be): Likewise.
25236 (aarch64_rshrn2<mode>_insn_le): Likewise.
25237 (aarch64_rshrn2<mode>_insn_be): Likewise.
25239 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
25242 * lra-constraints.c (skip_contraint_modifiers): New function.
25243 (process_address_1): Use it before lookup_constraint call.
25245 2021-03-08 Martin Liska <mliska@suse.cz>
25248 * config/i386/i386-options.c (ix86_option_override_internal):
25249 Enable UINTR and HRESET for -march that supports it.
25251 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
25253 * config/s390/s390.c (f_constraint_p): New function.
25254 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
25255 (TARGET_MD_ASM_ADJUST): Likewise.
25257 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
25260 * tree-nested.c (convert_local_reference_stmt): Avoid calling
25261 lookup_field_for_decl for Fortran module (= namespace context).
25263 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
25265 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
25266 comparison with arithmetic right shift.
25267 (s390_expand_vcond): No need for a force_reg anymore.
25268 s390_vec_compare will do it.
25269 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
25270 immediate operands.
25272 2021-03-07 Jakub Jelinek <jakub@redhat.com>
25275 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
25276 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
25278 * config/i386/sse.md (v_Yw): New define_mode_attr.
25279 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
25280 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
25282 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
25283 xYw in constraints.
25285 2021-03-06 Julian Brown <julian@codesourcery.com>
25287 * tree-pretty-print.c (dump_generic_node): Emit non-generic
25288 address space info for aggregates.
25290 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
25292 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
25294 2021-03-05 Jakub Jelinek <jakub@redhat.com>
25296 PR middle-end/99322
25297 * tree-cfg.c (bb_to_omp_idx): New variable.
25298 (execute_build_cfg): Release the bb_to_omp_idx vector after
25299 cleanup_tree_cfg returns.
25300 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
25301 for bb_to_omp_idx being a vec<int> instead of pointer to array
25303 (make_edges): Remove bb_to_omp_idx local variable, don't pass
25304 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
25305 vec<int> instead of pointer to array of ints and don't free/release
25307 (remove_bb): When removing a bb and placing forced label somewhere
25308 else, ensure it is put into the same OpenMP region during cfg
25309 pass if possible or to entry successor as fallback. Unregister
25310 bb from bb_to_omp_idx.
25312 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
25315 * lra-constraints.c (process_address_1): Skip decomposing address
25316 for asm insn operand with unknown constraint.
25318 2021-03-05 Martin Jambor <mjambor@suse.cz>
25321 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
25322 corresponding speculative edges if we are about to resolve
25323 sepculation. Make edge direct (and so resolve speculations) before
25324 removing it from call_site_hash.
25325 (cgraph_edge::make_direct): Relax the initial assert to allow calling
25326 the function on speculative direct edges.
25328 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
25330 PR rtl-optimization/99376
25331 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
25332 of low-order zero bits is too large, set the result to 0 directly.
25334 2021-03-04 Jakub Jelinek <jakub@redhat.com>
25336 PR middle-end/93235
25337 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
25338 SUBREG and a SUBREG to op_mode can't be created.
25340 2021-03-04 Alex Coplan <alex.coplan@arm.com>
25343 * config/aarch64/aarch64-sve-builtins.cc
25344 (function_resolver::require_vector_type): Handle error_mark_node.
25346 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
25348 * cfgexpand.c (expand_asm_loc): Pass new parameter.
25349 (expand_asm_stmt): Likewise.
25350 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
25352 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25353 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
25354 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
25355 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
25356 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
25357 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
25358 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
25359 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
25360 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
25361 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
25362 * doc/tm.texi (md_asm_adjust): Likewise.
25363 * target.def (md_asm_adjust): Likewise.
25365 2021-03-04 Richard Biener <rguenther@suse.de>
25367 PR middle-end/97855
25368 * tree-pretty-print.c: Poison pp_printf.
25369 (dump_decl_name): Avoid use of pp_printf.
25370 (dump_block_node): Likewise.
25371 (dump_generic_node): Likewise.
25373 2021-03-04 Martin Sebor <msebor@redhat.com>
25375 PR middle-end/96963
25376 PR middle-end/94655
25377 * builtins.c (handle_array_ref): New helper.
25378 (handle_mem_ref): New helper.
25379 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
25380 into new helper functions. Correct a workaround for vectorized
25383 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
25385 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
25386 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
25387 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
25388 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
25389 attribute for Power10.
25390 * config/rs6000/mma.md (*movoo): Likewise.
25391 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
25392 (define_mode_attr bits): Add DD/TD modes.
25393 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
25394 store_conditionalpti): Update size attribute for Power10.
25396 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25399 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
25400 -Wuninitialized, -Wmaybe-uninitialized.
25401 (wide-int.o-warn): Likewise.
25403 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
25405 * common/config/arm/arm-common.c: Include configargs.h.
25406 (arm_config_default): New function.
25407 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
25408 processors that do not support Thumb. Take into account the
25409 --with-mode configuration setting for selecting the default.
25410 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
25411 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
25413 2021-03-03 Martin Liska <mliska@suse.cz>
25415 PR gcov-profile/97461
25416 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
25418 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
25421 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25422 point back the hard frame pointer to its default location when the
25423 frame is larger than SEH_MAX_FRAME_SIZE.
25425 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25428 * config/i386/predicates.md (logic_operator): New define_predicate.
25429 * config/i386/i386.md (mov + mem using comm arith peephole2):
25430 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
25431 and the inner mode is [QH]Imode.
25433 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25436 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
25437 (new_loc_list): Clear end_entry.
25438 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
25439 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
25441 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
25442 initialize also end_entry.
25444 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25447 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
25448 partitions, if in non-layout mode after reorder_blocks also move
25449 affected blocks to ensure a single partition transition.
25451 2021-03-03 Jason Merrill <jason@redhat.com>
25454 * cgraphunit.c (process_function_and_variable_attributes): Don't
25455 warn about flatten on an alias if the target also has it.
25456 * cgraph.h (symtab_node::get_alias_target_tree): New.
25458 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
25460 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
25461 period to symbol name.
25462 (tls_get_addr_internal<mode>): Same.
25464 2021-03-02 David Malcolm <dmalcolm@redhat.com>
25467 * diagnostic-show-locus.c
25468 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
25471 2021-03-02 Martin Sebor <msebor@redhat.com>
25473 PR middle-end/99276
25474 * builtins.c (warn_for_access): Remove stray warning text.
25476 2021-03-02 Martin Sebor <msebor@redhat.com>
25478 PR middle-end/99295
25479 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
25482 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25485 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
25486 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
25487 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
25488 DW_MACRO_define_strx and DW_MACRO_undef_strx.
25489 (save_macinfo_strings): Use DW_MACRO_*_str* even with
25490 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
25491 DW_MACRO_undef_strx.
25493 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25495 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
25497 (BT_FN_V8HI_V8HI_UINT): Likewise.
25498 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
25499 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
25500 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
25501 New builtin definitions.
25502 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
25503 vector extension version.
25504 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
25505 available with current -march level.
25506 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
25507 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
25508 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
25509 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
25510 (vec_extend_to_fp32_lo): Likewise.
25511 (vec_round_from_fp32): Likewise.
25512 (vec_convert_to_fp16): Likewise.
25513 (vec_convert_from_fp16): Likewise.
25514 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
25515 (vclfnls_v8hi): Likewise.
25516 (vcrnfs_v8hi): Likewise.
25517 (vcfn_v8hi): Likewise.
25518 (vcnf_v8hi): Likewise.
25520 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25522 * common/config/s390/s390-common.c (processor_flags_table): New entry.
25523 * config.gcc: Enable arch14 for --with-arch and --with-tune.
25524 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
25525 arch14 for unknown CPU models.
25526 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
25527 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
25528 (s390_get_sched_attrmask): Likewise.
25529 (s390_get_unit_mask): Likewise.
25530 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
25531 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
25532 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
25533 (TARGET_NNPA_P): New macro definitions.
25534 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
25535 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
25537 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25539 PR middle-end/95757
25540 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
25541 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
25542 != 1 comparisons if name is lhs of a comparison.
25544 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
25548 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
25549 * config/darwin.c (darwin_should_restore_cfa_state): New.
25550 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
25551 * doc/tm.texi: Regenerated.
25552 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
25553 * dwarf2cfi.c (connect_traces): If the target requests, restore
25554 the CFA expression after a DW_CFA_restore.
25555 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
25557 2021-03-01 Martin Liska <mliska@suse.cz>
25560 * optc-save-gen.awk: Add 4 more exceptions.
25562 2021-03-01 Nathan Sidwell <nathan@acm.org>
25565 * tree.h (TYPE_ALIGN_RAW): New accessor.
25566 (TYPE_ALIGN): Use it.
25568 2021-03-01 Jan Hubicka <jh@suse.cz>
25571 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
25573 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
25576 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25577 point the hard frame pointer to the SSE register save area instead
25578 of the general register save area. Perform only minimal adjustment
25579 for small frames if it is initially not correctly aligned.
25580 (ix86_expand_prologue): Remove early saves for a SEH target.
25581 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
25583 2021-02-28 Jakub Jelinek <jakub@redhat.com>
25586 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
25587 typo - referneced -> referenced.
25588 * tree.c (component_ref_size): Fix comment typo -
25589 refernce -> reference.
25590 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
25591 traling -> trailing.
25592 (aliasing_component_refs_p): Fix comment typos -
25593 refernce -> reference and refernece -> reference and
25594 traling -> trailing.
25595 (nonoverlapping_refs_since_match_p): Fix comment typo -
25596 referneces -> references.
25597 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
25598 referneces -> references.
25600 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
25602 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
25603 diagnostic message to avoid use of a contraction and format
25606 2021-02-27 Jakub Jelinek <jakub@redhat.com>
25609 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
25610 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
25611 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
25612 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
25615 2021-02-26 Richard Biener <rguenther@suse.de>
25617 PR middle-end/99281
25618 * expr.c (store_field): For calls with return-slot optimization
25619 and addressable return type expand the store directly.
25621 2021-02-26 Richard Biener <rguenther@suse.de>
25624 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
25626 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
25629 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
25632 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
25634 * config.gcc: Add rs6000-pcrel-opt.o.
25635 * config/rs6000/rs6000-pcrel-opt.c: New file.
25636 * config/rs6000/pcrel-opt.md: New file.
25637 * config/rs6000/predicates.md: Add d_form_memory predicate.
25638 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
25639 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
25640 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
25641 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
25642 and make_pass_pcrel_opt().
25643 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
25644 (rs6000_option_override_internal): Add pcrel-opt.
25645 (rs6000_delegitimize_address): Support pcrel-opt.
25646 (rs6000_opt_masks): Add pcrel-opt.
25647 (pcrel_opt_valid_mem_p): New function.
25648 (reg_to_non_prefixed): Make global.
25649 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
25650 (output_pcrel_opt_reloc): New function.
25651 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
25652 (pcrel_extern_addr): Set loads_extern_addr.
25653 Add include for pcrel-opt.md.
25654 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
25655 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
25658 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
25661 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
25662 If TARGET_64BIT and dest is SUBREG, we check the width, if it
25663 equal to SImode, we use SImode operation, just like what we are
25666 2021-02-26 Marek Polacek <polacek@redhat.com>
25668 * builtins.c (warn_for_access): Fix typos.
25670 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
25672 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
25673 mark in front of the immediate quantity.
25674 (<optab>_rolsi3_uxtw): Likewise.
25676 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
25679 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
25680 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
25681 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
25682 address and disable when the FPCXT is not available.
25683 (nonsecure_call_value_reg_thumb2): Likewise.
25685 2021-02-25 Nathan Sidwell <nathan@acm.org>
25688 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
25690 2021-02-25 Tamar Christina <tamar.christina@arm.com>
25692 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
25694 2021-02-25 Richard Biener <rguenther@suse.de>
25696 PR tree-optimization/99253
25697 * tree-vect-loop.c (check_reduction_path): First compute
25698 code, then verify out-of-loop uses.
25700 2021-02-25 Jakub Jelinek <jakub@redhat.com>
25703 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
25705 2021-02-25 Jakub Jelinek <jakub@redhat.com>
25707 PR tree-optimization/80635
25708 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
25709 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
25710 has mode precision.
25712 2021-02-25 Richard Biener <rguenther@suse.de>
25714 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
25715 load_map population.
25716 (vect_match_slp_patterns_2): Revert part of last change.
25717 (vect_analyze_slp): Do not interleave optimize_load_redistribution
25718 with pattern detection but do it afterwards. Dump the
25719 whole SLP graph after pattern recognition and load
25720 redistribution optimization finished.
25722 2021-02-24 Jakub Jelinek <jakub@redhat.com>
25725 * omp-low.c (struct omp_context): Add teams_nested_p and
25726 nonteams_nested_p members.
25727 (scan_omp_target): Diagnose teams nested inside of target with other
25728 directives strictly nested inside of the same target.
25729 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
25730 ctx->nonteams_nested_p as needed.
25732 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
25734 PR inline-asm/99123
25735 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
25737 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
25739 * config/cris/cris.c (cris_expand_prologue): Set
25740 current_function_static_stack_size, if flag_stack_usage_info.
25742 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
25744 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
25745 (rs6000_final_prescan_insn): Adjust.
25746 (rs6000_asm_output_opcode): Likewise.
25748 2021-02-24 Martin Sebor <msebor@redhat.com>
25750 PR middle-end/97172
25751 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
25752 from function arguments.
25754 2021-02-24 Tamar Christina <tamar.christina@arm.com>
25756 PR tree-optimization/99220
25757 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
25758 node from cache when it's about to be deleted.
25760 2021-02-24 Jakub Jelinek <jakub@redhat.com>
25762 PR tree-optimization/99225
25763 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
25764 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
25765 build_int_cst (..., 1). Formatting fixes.
25767 2021-02-24 Tamar Christina <tamar.christina@arm.com>
25769 PR tree-optimization/99149
25770 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
25772 (vect_slp_reset_pattern): Remove.
25773 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
25774 (complex_mul_pattern::build, complex_fma_pattern::build,
25775 complex_fms_pattern::build): Fix ref counts.
25776 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
25777 when node is being deleted.
25778 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
25779 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
25781 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
25783 2021-02-24 Matthias Klose <doko@ubuntu.com>
25786 2020-12-07 Matthias Klose <doko@ubuntu.com>
25788 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
25789 and ENABLE_RTL_FLAG_CHECKING.
25791 2021-02-24 Richard Biener <rguenther@suse.de>
25794 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
25796 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
25798 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
25799 (vsx_assemble_pair): ...to this.
25800 (*mma_assemble_pair): Rename from this...
25801 (*vsx_assemble_pair): ...to this.
25802 (mma_disassemble_pair): Rename from this...
25803 (vsx_disassemble_pair): ...to this.
25804 (*mma_disassemble_pair): Rename from this...
25805 (*vsx_disassemble_pair): ...to this.
25806 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
25807 BU_COMPAT): New macros.
25808 (mma_assemble_pair): Rename from this...
25809 (vsx_assemble_pair): ...to this.
25810 (mma_disassemble_pair): Rename from this...
25811 (vsx_disassemble_pair): ...to this.
25812 (mma_assemble_pair): New compatibility built-in.
25813 (mma_disassemble_pair): Likewise.
25814 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
25815 (RS6000_BUILTIN_COMPAT): Define.
25816 (bdesc_compat): New.
25817 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
25818 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
25819 and VSX_BUILTIN_ASSEMBLE_PAIR.
25820 (rs6000_init_builtins): Register compatibility built-ins.
25821 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
25822 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
25823 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
25824 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
25825 (__builtin_vsx_assemble_pair): ...to this.
25826 (__builtin_mma_disassemble_pair): Rename from this...
25827 (__builtin_vsx_disassemble_pair): ...to this.
25829 2021-02-23 Martin Liska <mliska@suse.cz>
25832 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
25833 with different alignment. That leads to an invalid red zone
25834 size allocated in runtime.
25836 2021-02-23 Jakub Jelinek <jakub@redhat.com>
25838 PR tree-optimization/99204
25839 * fold-const.c (fold_read_from_constant_string): Check that
25840 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
25842 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
25843 Kewen Lin <linkw@gcc.gnu.org>
25845 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
25846 (rotl<mode>3_insert_3): ...this.
25847 (plus_ior_xor): New code_iterator.
25848 (define_split for GPR rl*imi): New splitter.
25849 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
25850 for integer merging.
25852 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25854 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
25856 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
25857 into a register when the above is enabled.
25858 * config/aarch64/aarch64.c (neoversev1_tunings):
25859 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
25860 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
25862 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
25864 * config/cris/cris.c (cris_print_operand) <'T'>: Change
25865 valid operand from is now an addi mult-value to shift-value.
25866 * config/cris/cris.md (*addi): Change expression of scaled
25867 operand from mult to ashift.
25868 * config/cris/cris.md (*addi_reload): New insn_and_split.
25870 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
25873 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
25874 hook_bool_const_tree_hwi_hwi_const_tree_true.
25875 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
25877 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
25879 PR rtl-optimization/98791
25880 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
25881 for unordered modes.
25883 2021-02-22 Martin Liska <mliska@suse.cz>
25885 * tree-inline.c (inline_forbidden_p): Set
25886 inline_forbidden_reason.
25888 2021-02-22 Richard Biener <rguenther@suse.de>
25890 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
25893 2021-02-22 Richard Biener <rguenther@suse.de>
25895 PR tree-optimization/99165
25896 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
25897 Accumulate changed to ret.
25899 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
25902 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
25904 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
25906 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
25909 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
25911 (trunctf<DFP_ALL:mode>2): Likewise.
25912 (trunctdtf2_vr): Likewise.
25913 (trunctdtf2): Likewise.
25914 (extend<DFP_ALL:mode>tf2_vr): Likewise.
25915 (extend<DFP_ALL:mode>tf2): Likewise.
25916 (extendtftd2_vr): Likewise.
25917 (extendtftd2): Likewise.
25919 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
25921 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
25922 add memory alternative.
25923 (tf_to_fprx2): New pattern.
25925 2021-02-19 Martin Sebor <msebor@redhat.com>
25928 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
25929 (attr_access::free_lang_data): Remove a spurious test.
25931 2021-02-19 Nathan Sidwell <nathan@acm.org>
25933 * doc/invoke.texi (flang-info-module-read): Document.
25935 2021-02-19 Martin Liska <mliska@suse.cz>
25937 PR translation/99167
25938 * params.opt: Fix typo.
25940 2021-02-19 Richard Biener <rguenther@suse.de>
25942 PR middle-end/99122
25943 * tree-inline.c (inline_forbidden_p): Do not inline functions
25944 with VLA arguments or return value.
25946 2021-02-19 Jakub Jelinek <jakub@redhat.com>
25949 * config/arm/arm.md (*stack_protect_combined_set_insn,
25950 *stack_protect_combined_test_insn): If force_const_mem result
25951 is not valid general operand, force its address into the destination
25954 2021-02-19 Jakub Jelinek <jakub@redhat.com>
25957 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
25958 pad or non-local label, put FORCED_LABELs from bb b after that label
25959 rather than before it.
25961 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
25964 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
25965 expand_vector_broadcast' to emit the vec_duplicate operand.
25967 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
25969 PR rtl-optimization/96264
25970 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
25973 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
25976 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
25977 looking up the retain attribute.
25978 (resolve_unique_section): Likewise.
25979 (get_variable_section): Likewise.
25980 (switch_to_section): Likewise. Warn when a symbol without the
25981 retain attribute and a symbol with the retain attribute are
25982 placed in the section with the same name, instead of the used
25984 * doc/extend.texi: Document the "retain" attribute.
25986 2021-02-18 Nathan Sidwell <nathan@acm.org>
25989 * doc/invoke.texi (flang-info-include-translate): Document header
25992 2021-02-18 Richard Biener <rguenther@suse.de>
25994 PR middle-end/99122
25995 * ipa-fnsummary.c (analyze_function_body): Set
25996 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
25997 * tree-inline.c (insert_init_debug_bind): Pass NULL for
25998 error_mark_node values.
25999 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
26001 (setup_one_parameter): Delay force_value_to_type until when
26004 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
26006 PR tree-optimization/99142
26007 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
26009 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26011 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
26012 wide_int_bitmask::wide_int_bitmask (uint64_t),
26013 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
26014 wide_int_bitmask::operator ~ () const,
26015 wide_int_bitmask::operator | (wide_int_bitmask) const,
26016 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
26018 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
26019 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
26020 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
26021 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
26022 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
26023 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
26024 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
26025 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
26026 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
26027 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
26028 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
26029 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
26030 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
26031 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
26032 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
26033 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
26034 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
26035 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
26036 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
26037 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
26038 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
26039 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
26040 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
26041 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
26043 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26045 PR middle-end/99109
26046 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
26047 (build_printable_array_type): ... this. Add nelts argument. For
26048 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
26049 nelts, call build_array_type_nelts.
26050 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
26051 instead of build_zero_elt_array_type and build_array_type_nelts.
26053 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26056 * config/i386/i386.c (distance_non_agu_define): Don't call
26057 extract_insn_cached here.
26058 (ix86_lea_outperforms): Save and restore recog_data around call
26059 to distance_non_agu_define and distance_agu_use.
26060 (ix86_ok_to_clobber_flags): Remove.
26061 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
26062 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
26063 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
26064 into define_insn. Move the splitting to define_peephole2 and
26065 check there using peep2_regno_dead_p if FLAGS_REG is dead.
26067 2021-02-17 Julian Brown <julian@codesourcery.com>
26069 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
26072 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
26075 * config/mips/mips.c (mips_symbol_insns): Do not use
26076 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
26078 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
26080 PR inline-asm/98096
26081 * stmt.c (resolve_operand_name_1): Take inout operands into account
26082 for access to labels by names.
26083 * doc/extend.texi: Describe counting operands for accessing labels.
26085 2021-02-16 Richard Biener <rguenther@suse.de>
26087 PR tree-optimization/38474
26088 * tree-ssa-structalias.c (variable_info::address_taken): New.
26089 (new_var_info): Initialize address_taken.
26090 (process_constraint): Set address_taken.
26091 (solve_constraints): Use the new address_taken flag rather
26092 than is_reg_var for sorting variables.
26093 (dump_constraint): Dump the variable number if the name
26096 2021-02-16 Jakub Jelinek <jakub@redhat.com>
26099 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
26100 multiply by 4096 and for inbranch by 8192.
26101 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
26102 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
26104 2021-02-15 Maya Rashish <coypu@sdf.org>
26106 * config/aarch64/aarch64.c (aarch64_init_builtins):
26107 Call SUBTARGET_INIT_BUILTINS.
26109 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
26111 PR rtl-optimization/98872
26112 * init-regs.c (initialize_uninitialized_regs): Skip initialization
26113 if CONST0_RTX is NULL.
26115 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
26117 PR rtl-optimization/98863
26118 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
26119 (function_info::build_info): Turn into a declaration, moving the
26120 definition to internals.h.
26121 (function_info::bb_walker): Declare.
26122 (function_info::create_reg_use): Likewise.
26123 (function_info::calculate_potential_phi_regs): Take a build_info
26125 (function_info::place_phis, function_info::create_ebbs): Declare.
26126 (function_info::calculate_ebb_live_in_for_debug): Likewise.
26127 (function_info::populate_backedge_phis): Delete.
26128 (function_info::start_block, function_info::end_block): Declare.
26129 (function_info::populate_phi_inputs): Delete.
26130 (function_info::m_potential_phi_regs): Move information to build_info.
26131 * rtl-ssa/internals.h: New file.
26132 (function_info::bb_phi_info): New class.
26133 (function_info::build_info): Moved from functions.h.
26134 Add a constructor and destructor.
26135 (function_info::build_info::ebb_use): Delete.
26136 (function_info::build_info::ebb_def): Likewise.
26137 (function_info::build_info::bb_live_out): Likewise.
26138 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
26139 (function_info::build_info::potential_phi_regs): Likewise.
26140 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
26141 (function_info::build_info::ebb_def_regs): Likewise.
26142 (function_info::build_info::bb_phis): Likewise.
26143 (function_info::build_info::bb_mem_live_out): Likewise.
26144 (function_info::build_info::bb_to_rpo): Likewise.
26145 (function_info::build_info::def_stack): Likewise.
26146 (function_info::build_info::old_def_stack_limit): Likewise.
26147 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
26148 Remove the regno argument. Push the previous definition onto the
26149 definition stack where necessary.
26150 * rtl-ssa/accesses.cc: Include internals.h.
26151 * rtl-ssa/changes.cc: Likewise.
26152 * rtl-ssa/blocks.cc: Likewise.
26153 (function_info::build_info::build_info): Define.
26154 (function_info::build_info::~build_info): Likewise.
26155 (function_info::bb_walker): New class.
26156 (function_info::bb_walker::bb_walker): Define.
26157 (function_info::add_live_out_use): Convert a logarithmic-complexity
26158 test into a linear one. Allow the same definition to be passed
26160 (function_info::calculate_potential_phi_regs): Moved from
26161 functions.cc. Take a build_info parameter and store the
26162 information there instead.
26163 (function_info::place_phis): New function.
26164 (function_info::add_entry_block_defs): Update call to record_reg_def.
26165 (function_info::calculate_ebb_live_in_for_debug): New function.
26166 (function_info::add_phi_nodes): Use bb_phis to decide which
26167 registers need phi nodes and initialize ebb_def_regs accordingly.
26168 Do not add degenerate phis here.
26169 (function_info::add_artificial_accesses): Use create_reg_use.
26170 Assert that all definitions are listed in the DF LR sets.
26171 Update call to record_reg_def.
26172 (function_info::record_block_live_out): Record live-out register
26173 values in the phis of successor blocks. Use the live-out set
26174 when processing the last block in an EBB, instead of always
26175 using the live-in sets of successor blocks. AND the live sets
26176 with the set of registers that have been defined in the EBB,
26177 rather than with all potential phi registers. Cope correctly
26178 with branches back to the start of the current EBB.
26179 (function_info::start_block): New function.
26180 (function_info::end_block): Likewise.
26181 (function_info::populate_phi_inputs): Likewise.
26182 (function_info::create_ebbs): Likewise.
26183 (function_info::process_all_blocks): Rewrite into a multi-phase
26185 * rtl-ssa/functions.cc: Include internals.h.
26186 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
26187 (function_info::init_function_data): Remove caller.
26188 * rtl-ssa/insns.cc: Include internals.h
26189 (function_info::create_reg_use): New function. Lazily any
26190 degenerate phis needed by the linear RPO view.
26191 (function_info::record_use): Use create_reg_use. When processing
26192 debug uses, use potential_phi_regs and test it before checking
26193 whether the register is live on entry to the current EBB. Lazily
26194 calculate ebb_live_in_for_debug.
26195 (function_info::record_call_clobbers): Update call to record_reg_def.
26196 (function_info::record_def): Likewise.
26198 2021-02-15 Martin Liska <mliska@suse.cz>
26200 * toplev.c (init_asm_output): Free output of
26201 gen_command_line_string function.
26202 (process_options): Likewise.
26204 2021-02-15 Martin Liska <mliska@suse.cz>
26206 * params.opt: Add 2 missing Param keywords.
26208 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
26210 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
26212 2021-02-15 Jakub Jelinek <jakub@redhat.com>
26214 PR tree-optimization/99079
26215 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
26216 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
26217 require both type and TREE_TYPE (@1) to be integral types and either
26218 type having smaller or equal precision, or TREE_TYPE (@1) being
26219 unsigned type, or type being signed type. If TREE_TYPE (@1)
26220 doesn't have wrapping overflow, perform the subtraction of one in
26223 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
26224 Richard Biener <rguether@suse.de>
26227 * ipa-reference.c (ipa_init): Only conditinally initialize
26228 reference_vars_to_consider.
26229 (propagate): Conditionally deninitialize reference_vars_to_consider.
26230 (ipa_reference_write_optimization_summary): Sanity check that
26231 reference_vars_to_consider is not allocated.
26233 2021-02-13 Levy Hsu <admin@levyhsu.com>
26236 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
26237 extend parameter to get_si_mem_base_reg declaration.
26238 (get_si_mem_base_reg): Add extend parameter. Set it.
26239 (analyze): Pass extend arg to get_si_mem_base_reg.
26240 (transform): Likewise. Use it when rewriting mems.
26241 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
26242 loads and emit sign/zero extending load followed by subreg move.
26244 2021-02-13 Jim Wilson <jimw@sifive.com>
26247 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
26248 exit when !reload_completed. Only perform check for compressed reg
26249 if reload_completed.
26250 (riscv_rtx_costs): In MEM case, when optimizing for size and
26251 shorten memrefs, if not compressible, then increase cost.
26253 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26255 PR rtl-optimization/98439
26256 * recog.c (pass_split_before_regstack::gate): Enable even when
26257 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
26260 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26263 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
26264 swap of V2SImode elements in memory into DImode memory rotate by 32.
26266 2021-02-12 Martin Sebor <msebor@redhat.com>
26268 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
26270 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26272 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
26273 m_temp_obstack rather than m_obstack to allocate the temporary use.
26275 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26277 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
26278 as read-modify operations.
26280 2021-02-12 Richard Biener <rguenther@suse.de>
26282 PR middle-end/38474
26283 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
26284 fbi->aa_walk_budget is bigger than zero. Update
26285 fbi->aa_walk_budget.
26286 (param_change_prob): Likewise.
26287 * ipa-prop.c (detect_type_change_from_memory_writes):
26288 Properly account walk_aliased_vdefs.
26289 (parm_preserved_before_stmt_p): Canonicalize updates.
26290 (parm_ref_data_preserved_p): Likewise.
26291 (parm_ref_data_pass_through_p): Likewise.
26292 (determine_known_aggregate_parts): Account own alias queries.
26294 2021-02-12 Martin Liska <mliska@suse.cz>
26296 * opts-common.c (decode_cmdline_option): Release werror_arg.
26297 * opts.c (gen_producer_string): Release output of
26298 gen_command_line_string.
26300 2021-02-12 Richard Biener <rguenther@suse.de>
26302 PR tree-optimization/38474
26303 * params.opt (-param=max-store-chains-to-track=): New param.
26304 (-param=max-stores-to-track=): Likewise.
26305 * doc/invoke.texi (max-store-chains-to-track): Document.
26306 (max-stores-to-track): Likewise.
26307 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
26309 (pass_store_merging::m_n_stores): Likewise.
26310 (pass_store_merging::terminate_and_process_chain): Update
26311 m_n_stores and m_n_chains.
26312 (pass_store_merging::process_store): Likewise. Terminate
26313 oldest chains if the number of stores or chains get too large.
26314 (imm_store_chain_info::terminate_and_process_chain): Dump
26317 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
26319 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
26320 the cold section, emit a nop before the directive if the previous
26321 active instruction can throw.
26323 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
26326 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
26327 memory addresses that are legal for quad word accesses.
26329 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
26332 * config/arm/thumb2.md (*doloop_end_internal): Generate
26333 alternative sequence to handle long range branches.
26335 2021-02-11 Joel Hutton <joel.hutton@arm.com>
26337 PR tree-optimization/98772
26338 * optabs-tree.c (supportable_half_widening_operation): New function
26339 to check for supportable V8QI->V8HI widening patterns.
26340 * optabs-tree.h (supportable_half_widening_operation): New function.
26341 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
26342 to create promotion stmts for V8QI->V8HI widening patterns.
26343 (vectorizable_conversion): Add case for V8QI->V8HI.
26345 2021-02-11 Richard Biener <rguenther@suse.de>
26347 * sparseset.h (SPARSESET_ELT_BITS): Remove.
26348 (SPARSESET_ELT_TYPE): Use unsigned int.
26349 * fwprop.c: Do not include sparseset.h.
26351 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26354 * varasm.c (declare_weak): For -fsyntax-only, allow even
26355 TREE_ASM_WRITTEN function decls.
26357 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26360 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
26361 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
26362 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
26363 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
26364 calling simplify_gen_subreg on it.
26366 2021-02-10 Martin Liska <mliska@suse.cz>
26368 * config/nvptx/nvptx.c (nvptx_option_override): Use
26369 flag_patchable_function_entry instead of the removed
26370 function_entry_patch_area_size.
26372 2021-02-10 Martin Liska <mliska@suse.cz>
26374 PR tree-optimization/99002
26375 PR tree-optimization/99026
26376 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
26377 leak when adjacent cases are merged.
26378 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
26380 (make_pass_lower_switch): Remove trailing whitespace.
26381 * tree-switch-conversion.h (release_clusters): New.
26383 2021-02-10 Richard Biener <rguenther@suse.de>
26385 PR rtl-optimization/99054
26386 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
26387 (fixup_partitions): Adjust.
26388 (rtl_verify_edges): Likewise.
26390 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26392 PR middle-end/99007
26393 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
26394 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
26397 2021-02-10 Richard Biener <rguenther@suse.de>
26400 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
26403 2021-02-10 Richard Biener <rguenther@suse.de>
26405 PR tree-optimization/99024
26406 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
26407 clear loop->aux if it is associated with the destroyed loop_vinfo.
26409 2021-02-10 Martin Liska <mliska@suse.cz>
26411 PR tree-optimization/99002
26412 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
26415 2021-02-10 Martin Liska <mliska@suse.cz>
26418 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
26419 a reference exists.
26421 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26424 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
26425 at class scope for DWARF5+.
26427 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
26429 PR rtl-optimization/96015
26430 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
26431 (relax_delay_slots): When deleting a jump to the next active
26432 instruction over a barrier, first delete the barrier if the
26433 jump is the only way to reach the target label.
26435 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
26437 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
26438 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
26439 vector multiplies and vect.alu for SSRA.
26440 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
26441 vect.mul cost field.
26442 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
26443 * config/arm/arm.c: Likewise.
26445 2021-02-09 Richard Biener <rguenther@suse.de>
26447 PR tree-optimization/98863
26448 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
26449 * tree-ssa-sccvn.c (last_pushed_avail): New global.
26450 (rpo_elim::eliminate_push_avail): Chain pushed avails.
26451 (unwind_state::avail_top): Add.
26452 (do_unwind): Rewrite unwinding of avail entries.
26453 (do_rpo_vn): Initialize last_pushed_avail and
26454 avail_top of the undo state.
26456 2021-02-09 Jakub Jelinek <jakub@redhat.com>
26458 PR middle-end/99004
26459 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
26460 const char * to char * and free those pointers after use.
26462 2021-02-09 Richard Biener <rguenther@suse.de>
26464 PR tree-optimization/99017
26465 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
26466 zero vector cost entries.
26468 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
26470 PR middle-end/98974
26471 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
26472 parameter in vectorizable_condition.
26474 2021-02-08 Richard Biener <rguenther@suse.de>
26477 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
26479 2021-02-08 Martin Liska <mliska@suse.cz>
26482 * cfgexpand.c (pass_expand::execute): Parse per-function option
26483 flag_patchable_function_entry and use it.
26484 * common.opt: Remove function_entry_patch_area_size and
26485 function_entry_patch_area_start global variables.
26486 * opts.c (parse_and_check_patch_area): New function.
26487 (common_handle_option): Use it.
26488 * opts.h (parse_and_check_patch_area): New function.
26489 * toplev.c (process_options): Parse and use
26490 function_entry_patch_area_size.
26492 2021-02-08 Martin Sebor <msebor@redhat.com>
26494 * doc/extend.texi (attribute malloc): Correct typos.
26496 2021-02-05 Nathan Sidwell <nathan@acm.org>
26499 * gcc.c (driver::maybe_run_linker): Check for input file
26500 accessibility if not linking.
26502 2021-02-05 Richard Biener <rguenther@suse.de>
26504 PR tree-optimization/98855
26505 * tree-vectorizer.h (add_stmt_cost): New overload.
26506 * tree-vect-slp.c (li_cost_vec_cmp): New.
26507 (vect_bb_slp_scalar_cost): Cost individual loop regions
26508 separately. Account for the scalar instance root stmt.
26510 2021-02-05 Tom de Vries <tdevries@suse.de>
26513 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
26515 (bit_test_cluster::emit): Reuse location_t for newly created
26517 (switch_decision_tree::try_switch_expansion): Preserve
26519 * tree-switch-conversion.h: Change function signatures.
26521 2021-02-05 Jakub Jelinek <jakub@redhat.com>
26524 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
26525 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
26526 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
26527 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
26529 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26531 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
26532 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
26533 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
26534 (vget_high_f16): Reimplement using new builtin.
26535 (vget_high_f32): Likewise.
26536 (vget_high_f64): Likewise.
26537 (vget_high_p8): Likewise.
26538 (vget_high_p16): Likewise.
26539 (vget_high_p64): Likewise.
26540 (vget_high_s8): Likewise.
26541 (vget_high_s16): Likewise.
26542 (vget_high_s32): Likewise.
26543 (vget_high_s64): Likewise.
26544 (vget_high_u8): Likewise.
26545 (vget_high_u16): Likewise.
26546 (vget_high_u32): Likewise.
26547 (vget_high_u64): Likewise.
26549 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26551 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
26552 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
26553 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
26554 (vget_low_f16): Reimplement using new builtin.
26555 (vget_low_f32): Likewise.
26556 (vget_low_f64): Likewise.
26557 (vget_low_p8): Likewise.
26558 (vget_low_p16): Likewise.
26559 (vget_low_p64): Likewise.
26560 (vget_low_s8): Likewise.
26561 (vget_low_s16): Likewise.
26562 (vget_low_s32): Likewise.
26563 (vget_low_s64): Likewise.
26564 (vget_low_u8): Likewise.
26565 (vget_low_u16): Likewise.
26566 (vget_low_u32): Likewise.
26567 (vget_low_u64): Likewise.
26569 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
26571 * gcc.c (print_multilib_info): Check all required argument is provided
26574 2021-02-05 liuhongt <hongtao.liu@intel.com>
26577 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
26578 generate integer mask comparison for 128/256-bits vector when
26579 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
26580 delete redundant !maskcmp condition.
26581 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
26583 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
26584 condition directly to if (maskcmp), add extra check for
26585 cmpmode, it should be MODE_INT.
26586 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
26587 parameters op_true/op_false.
26588 (ix86_use_mask_cmp_p): New.
26590 2021-02-05 liuhongt <hongtao.liu@intel.com>
26593 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
26594 Remove m_GENERIC from ~list.
26595 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
26597 2021-02-04 David Malcolm <dmalcolm@redhat.com>
26600 * diagnostic-show-locus.c (compatible_locations_p): Require
26601 locations in the same macro map to be either both from the
26602 macro definition, or both from the macro arguments.
26604 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
26606 * config/aarch64/aarch64-simd-builtins.def: Add
26607 [su]mull_hi_lane[q] builtin generator macros.
26608 * config/aarch64/aarch64-simd.md
26609 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
26610 (aarch64_<su>mull_hi_lane<mode>): Define.
26611 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
26612 (aarch64_<su>mull_hi_laneq<mode>): Define.
26613 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
26614 builtin instead of inline asm.
26615 (vmull_high_lane_s32): Likewise.
26616 (vmull_high_lane_u16): Likewise.
26617 (vmull_high_lane_u32): Likewise.
26618 (vmull_high_laneq_s16): Likewise.
26619 (vmull_high_laneq_s32): Likewise.
26620 (vmull_high_laneq_u16): Likewise.
26621 (vmull_high_laneq_u32): Liekwise.
26623 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
26625 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
26626 builtin generator macros.
26627 * config/aarch64/aarch64-simd.md
26628 (aarch64_<su>mull_hi_n<mode>_insn): Define.
26629 (aarch64_<su>mull_hi_n<mode>): Define.
26630 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
26631 instead of inline asm.
26632 (vmull_high_n_s32): Likewise.
26633 (vmull_high_n_u16): Likewise.
26634 (vmull_high_n_u32): Likewise.
26636 2021-02-04 Richard Biener <rguenther@suse.de>
26638 PR tree-optimization/98855
26639 * tree-vect-loop.c (vectorizable_phi): Do not cost
26640 single-argument PHIs.
26641 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
26642 * tree-vect-stmts.c (vectorizable_bswap): Also perform
26643 costing for SLP operation.
26645 2021-02-04 Martin Liska <mliska@suse.cz>
26647 * doc/extend.texi: Mention -mprefer-vector-width in target
26650 2021-02-03 Martin Sebor <msebor@redhat.com>
26652 PR tree-optimization/98937
26653 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
26654 Flush pointer_query cache.
26656 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
26658 * config/rs6000/genfusion.pl (gen_2logical): Add missing
26659 fixes based on patch review.
26660 * config/rs6000/fusion.md: Regenerate file.
26662 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
26664 * config/rs6000/t-rs6000: Comment out auto generation of
26667 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
26669 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
26670 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
26671 (output_file_start): Add gfx908.
26672 * config/gcn/gcn.opt (gpu_type): Add gfx908.
26673 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
26674 (MULTILIB_DIRNAMES): Add gfx908.
26675 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
26676 (main): Recognize gfx908.
26677 * config/gcn/t-omp-device: Add gfx908.
26679 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26681 * config/aarch64/aarch64-simd-builtins.def: Add
26682 [su]mlsl_hi_lane[q] builtin macro generators.
26683 * config/aarch64/aarch64-simd.md
26684 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
26685 (aarch64_<su>mlsl_hi_lane<mode>): Define.
26686 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
26687 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
26688 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
26689 builtin instead of inline asm.
26690 (vmlsl_high_lane_s32): Likewise.
26691 (vmlsl_high_lane_u16): Likewise.
26692 (vmlsl_high_lane_u32): Likewise.
26693 (vmlsl_high_laneq_s16): Likewise.
26694 (vmlsl_high_laneq_s32): Likewise.
26695 (vmlsl_high_laneq_u16): Likewise.
26696 (vmlsl_high_laneq_u32): Likewise.
26697 (vmlal_high_laneq_u32): Likewise.
26699 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26701 * config/aarch64/aarch64-simd-builtins.def: Add
26702 [su]mlal_hi_lane[q] builtin generator macros.
26703 * config/aarch64/aarch64-simd.md
26704 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
26705 (aarch64_<su>mlal_hi_lane<mode>): Define.
26706 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
26707 (aarch64_<su>mlal_hi_laneq<mode>): Define.
26708 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
26709 builtin instead of inline asm.
26710 (vmlal_high_lane_s32): Likewise.
26711 (vmlal_high_lane_u16): Likewise.
26712 (vmlal_high_lane_u32): Likewise.
26713 (vmlal_high_laneq_s16): Likewise.
26714 (vmlal_high_laneq_s32): Likewise.
26715 (vmlal_high_laneq_u16): Likewise.
26716 (vmlal_high_laneq_u32): Likewise.
26718 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26720 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
26721 builtin generator macros.
26722 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
26724 (aarch64_<su>mlsl_hi_n<mode>): Define.
26725 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
26726 instead of inline asm.
26727 (vmlsl_high_n_s32): Likewise.
26728 (vmlsl_high_n_u16): Likewise.
26729 (vmlsl_high_n_u32): Likewise.
26731 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26733 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
26734 builtin generator macros.
26735 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
26737 (aarch64_<su>mlal_hi_n<mode>): Define.
26738 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
26739 instead of inline asm.
26740 (vmlal_high_n_s32): Likewise.
26741 (vmlal_high_n_u16): Likewise.
26742 (vmlal_high_n_u32): Likewise.
26744 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26746 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
26748 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
26750 (aarch64_<su>mlal_hi<mode>_insn): This.
26751 (aarch64_<su>mlal_hi<mode>): Define.
26752 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
26753 instead of inline asm.
26754 (vmlal_high_s16): Likewise.
26755 (vmlal_high_s32): Likewise.
26756 (vmlal_high_u8): Likewise.
26757 (vmlal_high_u16): Likewise.
26758 (vmlal_high_u32): Likewise.
26760 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
26762 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
26763 after calling alter_subreg() on a (mem).
26765 2021-02-03 Martin Liska <mliska@suse.cz>
26768 * lto-streamer-out.c (produce_lto_section): Fill up missing
26770 * lto-streamer.h (struct lto_section): Add _padding field.
26772 2021-02-03 Richard Biener <rguenther@suse.de>
26774 * lto-streamer.c (lto_get_section_name): Free temporary
26776 * tree-loop-distribution.c
26777 (loop_distribution::merge_dep_scc_partitions): Free edge data.
26779 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26781 PR middle-end/97487
26782 * ifcvt.c (noce_can_force_operand): New function.
26783 (noce_emit_move_insn): Use it.
26784 (noce_try_sign_mask): Likewise. Formatting fix.
26786 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26788 PR middle-end/97971
26789 * lra-constraints.c (process_alt_operands): For inline asm, don't call
26790 fatal_insn, but instead return false.
26792 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26794 PR tree-optimization/98287
26795 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
26798 2021-02-03 Tamar Christina <tamar.christina@arm.com>
26800 PR tree-optimization/98928
26801 * tree-vect-loop.c (vect_analyze_loop_2): Change
26802 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
26803 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
26804 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
26805 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
26807 2021-02-02 Richard Biener <rguenther@suse.de>
26809 * gimple-loop-interchange.cc (prepare_data_references):
26811 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
26812 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
26813 * tree-vect-stmts.c (vectorizable_condition): Do not
26815 (vectorizable_comparison): Likewise.
26817 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26819 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
26820 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
26821 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
26822 (vrsqrteq_u32): Likewise.
26824 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26826 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
26827 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
26828 (aarch64_sqxtun2<mode>_be): Likewise.
26829 (aarch64_sqxtun2<mode>): Likewise.
26830 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
26831 (vqmovun_high_s32): Likewise.
26832 (vqmovun_high_s64): Likewise.
26833 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
26835 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26837 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
26839 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
26841 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26843 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
26844 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
26845 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
26846 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
26847 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
26848 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
26849 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
26850 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
26852 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26854 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
26855 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
26856 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
26858 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26860 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
26861 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
26863 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26865 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
26866 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
26867 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
26868 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
26869 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
26870 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
26871 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
26872 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
26873 NONE builtin flags.
26875 2021-02-02 Jakub Jelinek <jakub@redhat.com>
26877 PR tree-optimization/98848
26878 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
26879 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
26881 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
26884 * expr.c: Check mode before calling store_expr.
26886 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
26888 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
26890 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
26891 instruction using expression ior.
26892 (mve_vornq_u<mode>): New expander.
26893 (mve_vornq_f<mode>): Use ior code instead of unspec.
26894 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
26896 2021-02-02 Alexandre Oliva <oliva@adacore.com>
26898 * tree-nested.c (convert_nonlocal_reference_op): Move
26899 current_function_decl restore after re-gimplification.
26900 (convert_local_reference_op): Likewise.
26902 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26904 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
26906 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
26908 (aarch64_rshrn<mode>_insn_be): Likewise.
26909 (aarch64_rshrn<mode>): Likewise.
26910 (aarch64_rshrn2<mode>_insn_le): Likewise.
26911 (aarch64_rshrn2<mode>_insn_be): Likewise.
26912 (aarch64_rshrn2<mode>): Likewise.
26913 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
26914 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
26916 (vrshrn_high_n_s32): Likewise.
26917 (vrshrn_high_n_s64): Likewise.
26918 (vrshrn_high_n_u16): Likewise.
26919 (vrshrn_high_n_u32): Likewise.
26920 (vrshrn_high_n_u64): Likewise.
26921 (vrshrn_n_s16): Likewise.
26922 (vrshrn_n_s32): Likewise.
26923 (vrshrn_n_s64): Likewise.
26924 (vrshrn_n_u16): Likewise.
26925 (vrshrn_n_u32): Likewise.
26926 (vrshrn_n_u64): Likewise.
26928 2021-02-01 Sergei Trofimovich <siarheit@google.com>
26930 PR tree-optimization/98499
26931 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
26932 conservatively and assume all possible side-effects.
26934 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26936 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
26937 vec_unpacku_hi_): Define builtins.
26938 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
26940 (vmovl_high_s16): Likewise.
26941 (vmovl_high_s32): Likewise.
26942 (vmovl_high_u8): Likewise.
26943 (vmovl_high_u16): Likewise.
26944 (vmovl_high_u32): Likewise.
26946 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26948 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
26950 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
26952 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
26954 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
26956 (vabdl_s16): Likewise.
26957 (vabdl_s32): Likewise.
26958 (vabdl_u8): Likewise.
26959 (vabdl_u16): Likewise.
26960 (vabdl_u32): Likewise.
26961 * config/aarch64/iterators.md (ABDL): New int iterator.
26962 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
26964 2021-02-01 Martin Sebor <msebor@redhat.com>
26966 * tree.h (BLOCK_VARS): Add comment.
26967 (BLOCK_SUBBLOCKS): Same.
26968 (BLOCK_SUPERCONTEXT): Same.
26969 (BLOCK_ABSTRACT_ORIGIN): Same.
26970 (inlined_function_outer_scope_p): Same.
26972 2021-02-01 Martin Sebor <msebor@redhat.com>
26974 PR middle-end/97172
26975 * attribs.c (attr_access::free_lang_data): Define new function.
26976 * attribs.h (attr_access::free_lang_data): Declare new function.
26978 2021-02-01 Richard Biener <rguenther@suse.de>
26980 * vec.h (auto_vec::auto_vec): Add memory stat parameters
26982 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
26984 2021-02-01 Tamar Christina <tamar.christina@arm.com>
26986 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
26987 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
26989 2021-02-01 Richard Biener <rguenther@suse.de>
26991 PR rtl-optimization/98863
26992 * config/i386/i386-features.c (convert_scalars_to_vector):
26993 Set DF_RD_PRUNE_DEAD_DEFS.
26995 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
26997 * system.h (SIZE_MAX): Define if not already defined.
26999 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
27001 * config/rs6000/genfusion.pl (gen_2logical): New function to
27002 generate patterns for logical-logical fusion.
27003 * config/rs6000/fusion.md: Regenerated patterns.
27004 * config/rs6000/rs6000-cpus.def: Add
27005 OPTION_MASK_P10_FUSION_2LOGICAL.
27006 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27007 Enable logical-logical fusion for p10.
27008 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
27010 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
27012 * config/rs6000/rs6000.opt: Add periods to new AIX options.
27014 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
27016 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
27017 (mabi=vec-default): New.
27018 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
27019 __EXTABI__ for AIX Vector extended ABI.
27020 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
27022 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
27024 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
27026 2021-01-30 Jakub Jelinek <jakub@redhat.com>
27028 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
27029 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
27031 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
27034 * lra-constraints.c (in_class_p): Don't narrow class only for REG
27037 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
27039 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
27040 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
27041 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
27043 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
27045 PR tree-optimization/98866
27046 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
27047 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
27048 (gori_map::m_maybe_invariant): Rename from all_outgoing.
27049 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
27050 (gori_map::is_export_p): Ditto.
27051 (gori_map::calculate_gori): Ditto.
27052 (gori_compute::set_range_invariant): New.
27053 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
27054 invariant for pointers evaluating to [1, +INF].
27056 2021-01-29 Richard Biener <rguenther@suse.de>
27058 PR rtl-optimization/98863
27059 * config/i386/i386-features.c (remove_partial_avx_dependency):
27060 Do not perform DF analysis.
27061 (pass_data_remove_partial_avx_dependency): Remove
27064 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27066 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
27067 builtin generator macros.
27068 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
27070 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
27071 instead of inline asm.
27072 (vmull_n_s32): Likewise.
27073 (vmull_n_u16): Likewise.
27074 (vmull_n_u32): Likewise.
27076 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27078 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
27080 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
27082 (aarch64_<sur>abdl2<mode>): ... This.
27083 (<sur>sadv16qi): Adjust use of above.
27084 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
27086 (vabdl_high_s16): Likewise.
27087 (vabdl_high_s32): Likewise.
27088 (vabdl_high_u8): Likewise.
27089 (vabdl_high_u16): Likewise.
27090 (vabdl_high_u32): Likewise.
27092 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27094 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
27096 (uabal2): Likewise.
27097 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
27099 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
27101 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
27103 (vabal_high_s16): Likewise.
27104 (vabal_high_s32): Likewise.
27105 (vabal_high_u8): Likewise.
27106 (vabal_high_u16): Likewise.
27107 (vabal_high_u32): Likewise.
27108 * config/aarch64/iterators.md (ABAL2): New mode iterator.
27109 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
27111 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27113 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
27116 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
27118 (aarch64_<sur>abal<mode>): ... This
27119 (<sur>sadv16qi): Adust use of the above.
27120 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
27122 (vabal_s16): Likewise.
27123 (vabal_s32): Likewise.
27124 (vabal_u8): Likewise.
27125 (vabal_u16): Likewise.
27126 (vabal_u32): Likewise.
27128 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27130 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
27132 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
27134 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
27136 (vaddlv_s16): Likewise.
27137 (vaddlv_u8): Likewise.
27138 (vaddlv_u16): Likewise.
27139 (vaddlvq_s8): Likewise.
27140 (vaddlvq_s16): Likewise.
27141 (vaddlvq_s32): Likewise.
27142 (vaddlvq_u8): Likewise.
27143 (vaddlvq_u16): Likewise.
27144 (vaddlvq_u32): Likewise.
27145 (vaddlv_s32): Likewise.
27146 (vaddlv_u32): Likewise.
27147 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
27148 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
27149 (Vwstype): New mode attribute.
27151 (VWIDE_S): Likewise.
27152 (USADDLV): New int iterator.
27153 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
27155 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27157 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
27158 builtin generator macros.
27159 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
27161 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
27162 instead of inline asm.
27163 (vmlsl_lane_s32): Likewise.
27164 (vmlsl_lane_u16): Likewise.
27165 (vmlsl_lane_u32): Likewise.
27166 (vmlsl_laneq_s16): Likewise.
27167 (vmlsl_laneq_s32): Likewise.
27168 (vmlsl_laneq_u16): Likewise.
27169 (vmlsl_laneq_u32): Likewise.
27171 2021-01-29 Richard Biener <rguenther@suse.de>
27173 * doc/invoke.texi (--param max-gcse-memory): Document unit
27175 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
27176 * params.opt (--param max-gcse-memory): Adjust default and
27177 document unit of size.
27179 2021-01-29 Richard Biener <rguenther@suse.de>
27181 PR rtl-optimization/98863
27182 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
27183 HOST_WIDE_INT for the memory estimate.
27185 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
27186 Richard Biener <rguenther@suse.de>
27188 PR tree-optimization/97627
27189 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
27190 Do not analyze fake edges.
27192 2021-01-29 Richard Biener <rguenther@suse.de>
27194 PR rtl-optimization/98144
27195 * df.h (df_mir_bb_info): Add con_visited member.
27196 * df-problems.c (df_mir_alloc): Initialize con_visited,
27197 do not fully populate IN and OUT.
27198 (df_mir_reset): Likewise.
27199 (df_mir_confluence_0): Set con_visited.
27200 (df_mir_confluence_n): Properly handle implicitely
27201 fully populated IN and OUT as designated by con_visited
27202 and update con_visited accordingly.
27204 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27207 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
27208 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
27209 && !TARGET_REALLY_IWMMXT to conditions.
27211 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27214 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
27217 2021-01-28 Marek Polacek <polacek@redhat.com>
27220 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
27221 the main variant, maybe reset it in its variants too.
27222 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
27223 (check_aligned_type): Check if TYPE_USER_ALIGN match.
27225 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
27228 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
27229 of constant zero for comparisons.
27231 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
27233 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
27234 support for mapping built-in function names for long double
27235 built-in functions if long double is IEEE 128-bit.
27237 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27239 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
27240 builtin generator macros.
27241 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
27243 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
27244 instead of inline asm.
27245 (vmlsl_n_s32): Likewise.
27246 (vmlsl_n_u16): Likewise.
27247 (vmlsl_n_u32): Likewise.
27249 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27251 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
27252 builtin generator macros.
27253 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
27255 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
27256 instead of inline asm.
27257 (vmlal_n_s32): Likewise.
27258 (vmlal_n_u16): Likewise.
27259 (vmlal_n_u32): Likewise.
27261 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27263 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
27265 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
27267 (aarch64_shrn2<mode>_insn_be): Likewise.
27268 (aarch64_shrn2<mode>): Likewise.
27269 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
27271 (vshrn_high_n_s32): Likewise.
27272 (vshrn_high_n_s64): Likewise.
27273 (vshrn_high_n_u16): Likewise.
27274 (vshrn_high_n_u32): Likewise.
27275 (vshrn_high_n_u64): Likewise.
27277 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27279 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
27281 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
27283 (aarch64_shrn<mode>_insn_be): Likewise.
27284 (aarch64_shrn<mode>): Likewise.
27285 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
27287 (vshrn_n_s32): Likewise.
27288 (vshrn_n_s64): Likewise.
27289 (vshrn_n_u16): Likewise.
27290 (vshrn_n_u32): Likewise.
27291 (vshrn_n_u64): Likewise.
27292 * config/aarch64/iterators.md (vn_mode): New mode attribute.
27294 2021-01-28 Richard Biener <rguenther@suse.de>
27296 PR rtl-optimization/80960
27297 * dse.c (check_mem_read_rtx): Call get_addr on the
27300 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
27301 David Edelsohn <dje.gcc@gmail.com>
27304 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27305 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
27307 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27309 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
27310 wrapper call rs6000_expand_vector_set_var for cleanup. Call
27311 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
27313 (rs6000_expand_vector_set_var): Delete.
27314 (rs6000_expand_vector_set_var_p9): Make static.
27315 (rs6000_expand_vector_set_var_p8): Make static.
27317 2021-01-28 Xing GUO <higuoxing@gmail.com>
27319 * common/config/riscv/riscv-common.c
27320 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
27321 when `p` extension exists.
27323 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
27325 PR rtl-optimization/97684
27326 * ira.c (ira): Call ira_set_pseudo_classes before
27327 update_equiv_regs when it is necessary.
27329 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27332 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
27333 %w0, %w1 and %2 instead of %0, %1 and %2.
27335 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
27337 * config/rs6000/genfusion.pl: New script to generate
27338 define_insn_and_split patterns so combine can arrange fused
27339 instructions next to each other.
27340 * config/rs6000/fusion.md: New file, generated fused instruction
27341 patterns for combine.
27342 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
27343 (non_update_memory_operand): New predicate.
27344 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
27345 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
27347 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
27349 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27350 Automatically set OPTION_MASK_P10_FUSION and
27351 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
27352 (rs600_opt_masks): Allow -mpower10-fusion
27353 in function attributes.
27354 (address_is_non_pfx_d_or_x): New function.
27355 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
27356 * config/rs6000/rs6000.md: Include fusion.md.
27357 * config/rs6000/rs6000.opt: Add -mpower10-fusion
27358 and -mpower10-fusion-ld-cmpi.
27359 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
27361 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27363 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
27364 builtin generator macros.
27365 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
27367 (aarch64_<su>mlal<mode>): This.
27368 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
27369 instead of inline asm.
27370 (vmlal_s16): Likewise.
27371 (vmlal_s32): Likewise.
27372 (vmlal_u8): Likewise.
27373 (vmlal_u16): Likewise.
27374 (vmlal_u32): Likewise.
27376 2021-01-27 Richard Biener <rguenther@suse.de>
27378 PR tree-optimization/98854
27379 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
27380 PHIs from scalars when the number of CTORs matches the
27381 number of children.
27383 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27385 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
27387 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
27389 (aarch64_mls_n<mode>): This.
27390 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
27392 (vmls_n_s32): Likewise.
27393 (vmls_n_u16): Likewise.
27394 (vmls_n_u32): Likewise.
27395 (vmlsq_n_s16): Likewise.
27396 (vmlsq_n_s32): Likewise.
27397 (vmlsq_n_u16): Likewise.
27398 (vmlsq_n_u32): Likewise.
27400 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27402 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
27404 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
27406 (vmls_s16): Likewise.
27407 (vmls_s32): Likewise.
27408 (vmls_u8): Likewise.
27409 (vmls_u16): Likewise.
27410 (vmls_u32): Likewise.
27411 (vmlsq_s8): Likewise.
27412 (vmlsq_s16): Likewise.
27413 (vmlsq_s32): Likewise.
27414 (vmlsq_u8): Likewise.
27415 (vmlsq_u16): Likewise.
27416 (vmlsq_u32): Likewise.
27418 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27420 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
27422 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
27424 (aarch64_mla_n<mode>): This.
27425 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
27427 (vmla_n_s32): Likewise.
27428 (vmla_n_u16): Likewise.
27429 (vmla_n_u32): Likewise.
27430 (vmlaq_n_s16): Likewise.
27431 (vmlaq_n_s32): Likewise.
27432 (vmlaq_n_u16): Likewise.
27433 (vmlaq_n_u32): Likewise.
27435 2021-01-27 liuhongt <hongtao.liu@intel.com>
27438 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
27439 (*sse2_eq<mode>3): Ditto.
27441 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27443 * tree-pass.h (PROP_trees): Rename to ...
27444 (PROP_gimple): ... this.
27445 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
27446 * passes.c (execute_function_dump, execute_function_todo,
27447 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
27448 * varpool.c (ctor_for_folding): Likewise.
27450 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27452 PR tree-optimization/97260
27453 * varpool.c: Include tree-pass.h.
27454 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
27455 non-TREE_SIDE_EFFECTS automatic variables.
27457 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
27459 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
27461 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
27462 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
27463 (gen_compile_unit_die): Recognise C++23.
27465 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27468 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
27471 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27474 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
27475 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
27476 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
27478 2021-01-26 Richard Biener <rguenther@suse.de>
27480 * gimple-pretty-print.c (dump_binary_rhs): Handle
27481 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
27483 2021-01-26 Richard Biener <rguenther@suse.de>
27485 PR middle-end/98726
27486 * tree.h (vector_cst_int_elt): Remove.
27487 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
27490 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
27492 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
27493 for V64DFmode min/max reductions.
27495 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27497 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
27498 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
27499 two halves, one with x and the other with const0_rtx, ordered
27500 depending on endianity.
27502 2021-01-26 Alexandre Oliva <oliva@adacore.com>
27504 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
27505 temporaries not seen in binding block, and not about to be
27506 added as gimple variables.
27508 2021-01-25 Martin Sebor <msebor@redhat.com>
27511 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
27513 2021-01-25 Martin Liska <mliska@suse.cz>
27515 * value-prof.c (get_nth_most_common_value): Use %s instead
27518 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27521 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
27522 readelf -wi is able to read the emitted .debug_info back.
27523 * configure: Regenerated.
27525 2021-01-25 Martin Liska <mliska@suse.cz>
27527 PR gcov-profile/98739
27528 * common.opt: Add missing sign symbol.
27529 * value-prof.c (get_nth_most_common_value): Restore handling
27530 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
27531 PROFILE_REPRODUCIBILITY_MULTITHREADED.
27533 2021-01-25 Richard Biener <rguenther@suse.de>
27535 PR middle-end/98807
27536 * tree.c (vector_element_bits): Always use precision of
27537 the element type for boolean vectors.
27539 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27541 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
27542 (ENDFILE_SPEC): Evaluate qnolinkcmds.
27544 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27546 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
27547 nostartfiles handling since this is already done by
27548 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
27549 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
27550 is already done by LINK_COMMAND_SPEC.
27551 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
27552 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
27555 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27558 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
27559 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
27560 value rather than host size_t.
27561 (fold_const_call): Change type of s2 from size_t to
27562 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
27563 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
27564 instead of s2 as last argument.
27566 2021-01-25 Tamar Christina <tamar.christina@arm.com>
27568 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
27569 VCMLA_OP, VCMUL_OP): New.
27570 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
27571 * config/arm/neon.md (cmul<conj_op><mode>3): New.
27572 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
27573 UNSPEC_VCMUL_CONJ): New.
27574 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
27575 cml<fcmac1><conj_op><mode>4): New.
27577 2021-01-23 Jakub Jelinek <jakub@redhat.com>
27580 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
27582 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
27584 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
27586 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
27588 (vmla_s16): Likewise.
27589 (vmla_s32): Likewise.
27590 (vmla_u8): Likewise.
27591 (vmla_u16): Likewise.
27592 (vmla_u32): Likewise.
27593 (vmlaq_s8): Likewise.
27594 (vmlaq_s16): Likewise.
27595 (vmlaq_s32): Likewise.
27596 (vmlaq_u8): Likewise.
27597 (vmlaq_u16): Likewise.
27598 (vmlaq_u32): Likewise.
27600 2021-01-22 David Malcolm <dmalcolm@redhat.com>
27602 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
27605 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27608 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
27609 filenames to emit, still emit the required 0 index directory and
27610 filename entries that match DW_AT_comp_dir and DW_AT_name of the
27613 2021-01-22 Marek Polacek <polacek@redhat.com>
27616 * doc/invoke.texi: Update C++ ABI Version 15 description.
27618 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27620 PR tree-optimization/98766
27621 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
27622 comparing against type size with param_avoid_fma_max_bits.
27624 2021-01-22 Richard Biener <rguenther@suse.de>
27626 PR middle-end/98793
27627 * tree.c (vector_element_bits): Key single-bit bool vector on
27628 integer mode rather than not vector mode.
27630 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
27633 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27634 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
27636 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
27637 to call different path for P8 and P9.
27638 (rs6000_expand_vector_set_var_p9): New function.
27639 (rs6000_expand_vector_set_var_p8): New function.
27641 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
27645 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27646 Ajdust variable index vec_insert from address dereference to
27647 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
27648 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27650 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
27652 2021-01-22 Martin Liska <mliska@suse.cz>
27654 PR gcov-profile/98739
27655 * profile.c (compute_value_histograms): Drop time profile for
27656 -fprofile-reproducible=multithreaded.
27658 2021-01-22 Nathan Sidwell <nathan@acm.org>
27660 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
27663 2021-01-22 Richard Biener <rguenther@suse.de>
27665 PR middle-end/98773
27666 * tree-data-ref.c (initalize_matrix_A): Revert previous
27667 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
27669 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27671 PR tree-optimization/90248
27672 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
27673 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
27675 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
27676 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
27678 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27680 PR tree-optimization/98255
27681 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
27682 extend index - low_bound from sizetype's precision rather than index
27684 (get_addr_base_and_unit_offset_1): Likewise.
27685 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
27686 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
27688 2021-01-22 Richard Biener <rguenther@suse.de>
27690 PR tree-optimization/98786
27691 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
27692 adding new uses of abnormals. Verify we deal with a conditional
27695 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27698 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
27700 2021-01-22 liuhongt <hongtao.liu@intel.com>
27704 * config/i386/sse.md (VI_128_256): New mode iterator.
27705 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
27706 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
27707 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
27708 define_insn_and_split to lower avx512 vector comparison to avx
27709 version when dest is vector.
27710 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
27711 define_insn_and_split for negating the comparison result.
27712 * config/i386/predicates.md (float_vector_all_ones_operand):
27714 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
27715 general NOT operator without UNSPEC_MASKOP.
27717 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
27719 PR rtl-optimization/98777
27720 * lra-int.h (lra_pmode_pseudo): New extern.
27721 * lra.c (lra_pmode_pseudo): New global.
27723 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
27725 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
27727 * fwprop.c (fwprop_propagation::classify_result): Allow
27728 (subreg (mem)) simplifications.
27730 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27732 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
27734 (aarch64_sqdmlal<mode>): ... This...
27735 (aarch64_sqdmlsl<mode>): ... And this.
27736 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
27737 (aarch64_sqdmlal_lane<mode>): ... This...
27738 (aarch64_sqdmlsl_lane<mode>): ... And this.
27739 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
27740 (aarch64_sqdmlsl_laneq<mode>): ... This...
27741 (aarch64_sqdmlal_laneq<mode>): ... And this.
27742 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
27743 (aarch64_sqdmlsl_n<mode>): ... This...
27744 (aarch64_sqdmlal_n<mode>): ... And this.
27745 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
27746 (aarch64_sqdmlal2<mode>_internal): ... This...
27747 (aarch64_sqdmlsl2<mode>_internal): ... And this.
27749 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
27751 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
27753 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
27756 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
27758 2021-01-21 liuhongt <hongtao.liu@intel.com>
27760 PR rtl-optimization/98694
27761 * regcprop.c (copy_value): If SRC had been assigned a mode
27762 narrower than the copy, we can't link DEST into the chain even
27763 they have same hard_regno_nregs(i.e. HImode/SImode in i386
27766 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27768 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
27769 Convert to define_insn_and_split. Split into simple move when moving
27772 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
27774 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
27775 Adjust comment. Simplify code.
27777 2021-01-20 Jakub Jelinek <jakub@redhat.com>
27780 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
27781 with DW_FORM_line_strp form.
27782 (prune_unused_types_update_strings): Don't add into debug_str_hash
27783 indirect strings with DW_FORM_line_strp form.
27784 (adjust_name_comp_dir): New function.
27785 (dwarf2out_finish): Call it on CU DIEs after resetting
27786 debug_line_str_hash.
27788 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
27790 PR rtl-optimization/98722
27791 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
27792 has no 3-op add insn to transform insns containing two pluses.
27794 2021-01-20 Richard Biener <rguenther@suse.de>
27796 * hwint.h (add_hwi): New function.
27797 (mul_hwi): Likewise.
27798 * tree-data-ref.c (initialize_matrix_A): Properly translate
27799 tree constants and avoid HOST_WIDE_INT_MIN.
27800 (lambda_matrix_row_add): Avoid undefined integer overflow
27801 and return true on such overflow.
27802 (lambda_matrix_right_hermite): Handle overflow from
27803 lambda_matrix_row_add gracefully. Simplify previous fix.
27804 (analyze_subscript_affine_affine): Likewise.
27806 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
27808 PR tree-optimization/96674
27809 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
27810 x >= y && y != XXX_MIN --> x > y - 1
27812 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
27814 PR tree-optimization/98535
27815 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
27816 If the high and low permutes are the same, remove the high permutes
27817 from the working set and only continue with the low ones.
27819 2021-01-20 Jakub Jelinek <jakub@redhat.com>
27821 PR tree-optimization/98721
27822 * builtins.c (access_ref::inform_access): Don't assume
27823 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
27824 object whenever allocfn is NULL, rather than only when DECL_P
27825 is true. Use %qE instead of %qD for that. Formatting fixes.
27827 2021-01-20 Richard Biener <rguenther@suse.de>
27829 PR tree-optimization/98758
27830 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
27831 (lambda_matrix_right_hermite): Avoid undefinedness with
27832 signed integer abs and multiplication.
27833 (analyze_subscript_affine_affine): Use lambda_int.
27835 2021-01-20 David Malcolm <dmalcolm@redhat.com>
27838 * dwarf2out.c (output_line_info): Rename static variable
27839 "generation", moving it out of the function to...
27840 (output_line_info_generation): New.
27841 (init_sections_and_labels): Likewise, renaming the variable to...
27842 (init_sections_and_labels_generation): New.
27843 (dwarf2out_c_finalize): Reset the new variables.
27845 2021-01-19 Martin Sebor <msebor@redhat.com>
27847 PR middle-end/98664
27848 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
27849 all functions, even if they're not declared artificial or inline.
27850 * tree.c (tree_inlined_location): Use macro expansion location
27851 only if scope traversal fails to expose one.
27853 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
27855 PR rtl-optimization/92294
27856 * alias.c (compare_base_symbol_refs): Take an extra parameter
27857 and add the distance between two symbols to it. Enshrine in
27858 comments that -1 means "either 0 or 1, but we can't tell
27859 which at compile time".
27860 (memrefs_conflict_p): Update call accordingly.
27861 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
27864 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27866 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
27867 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
27868 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
27869 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
27870 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
27871 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
27872 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
27873 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
27874 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
27875 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
27876 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
27877 sqmovun): Use NONE flags.
27879 2021-01-19 Richard Biener <rguenther@suse.de>
27882 * ipa-modref.c (analyze_stmt): Only record a summary for a
27885 2021-01-19 Richard Biener <rguenther@suse.de>
27887 PR middle-end/98638
27888 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
27890 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
27892 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
27893 built-in define __FIX_LEON3FT_TN0018.
27895 2021-01-19 Richard Biener <rguenther@suse.de>
27898 * tree-inline.c (tree_function_versioning): Set input_location
27899 to UNKNOWN_LOCATION throughout the function.
27901 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
27904 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
27906 2021-01-19 Martin Jambor <mjambor@suse.cz>
27909 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
27910 whether non-call exceptions allow removal of a statement.
27911 (isra_analyze_call): Pass the appropriate function to
27912 ssa_name_only_returned_p.
27914 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
27916 * config/riscv/arch-canonicalize (longext_sort): New function for
27917 sorting 'multi-letter'.
27918 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
27919 'alts'. The 'arch' may not be the first of 'alts'.
27920 (_expand_combination): Add underline for the 'ext' without '*'.
27921 This is because, a single-letter extension can always be treated well
27922 with a '_' prefix, but it cannot be separated out if it is appended
27925 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
27928 * ira.c (ira): Skip abnormal critical edge splitting.
27930 2021-01-18 Jakub Jelinek <jakub@redhat.com>
27932 PR tree-optimization/98727
27933 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
27934 second .MUL_OVERFLOW operand for signed multiplication with overflow
27935 checking if the second operand of multiplication is not constant.
27937 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
27939 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
27940 defaults to version 4.
27942 2021-01-18 David Malcolm <dmalcolm@redhat.com>
27944 * attribs.h (fndecl_dealloc_argno): New decl.
27945 * builtins.c (call_dealloc_argno): Split out second half of
27947 (fndecl_dealloc_argno): New.
27948 * doc/extend.texi (Common Function Attributes): Document the
27949 interaction between the analyzer and the malloc attribute.
27950 * doc/invoke.texi (Static Analyzer Options): Likewise.
27952 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
27954 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
27955 dwarf_version to 4.
27956 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
27958 2021-01-17 Martin Jambor <mjambor@suse.cz>
27961 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
27964 2021-01-17 Mark Wielaard <mark@klomp.org>
27966 * common.opt (gdwarf-): Init(5).
27967 * doc/invoke.texi (-gdwarf): Document default to 5.
27969 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
27971 * builtin-types.def
27972 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
27974 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
27975 ...this. Add extra argument.
27976 * gimplify.c (omp_default_clause): Ensure that event handle is
27977 firstprivate in a task region.
27978 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
27979 (gimplify_adjust_omp_clauses): Likewise.
27980 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
27981 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
27982 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
27983 if detach clause specified. Add detach argument when generating
27985 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
27987 (finish_taskreg_scan): Move field for variable containing the event
27988 handle to the front of the struct.
27989 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
27991 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
27992 OMP_CLAUSE_DETACH clause.
27993 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
27994 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
27995 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
27997 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
27999 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
28001 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28003 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
28004 multilib directory names. Use MULTILIB_REQUIRED instead of
28005 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
28006 -mcustom-fpu-cfg=fph2 multilib.
28008 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28010 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
28011 (nios2_init_fpu_configs): Provide register values for new
28012 -mcustom-fpu-cfg=fph2 option variant.
28013 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
28016 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28018 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
28019 custom instruction warnings.
28021 2021-01-16 Jakub Jelinek <jakub@redhat.com>
28023 PR tree-optimization/96669
28024 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
28026 2021-01-16 Jakub Jelinek <jakub@redhat.com>
28028 PR tree-optimization/96271
28029 * passes.def: Pass false argument to first two pass_cd_dce
28030 instances and true to last instance. Add comment that
28031 last instance rewrites no longer addressed locals.
28032 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
28034 (pass_cd_dce::set_pass_param): New method.
28035 (pass_cd_dce::execute): Return TODO_update_address_taken from
28036 last cd_dce instance.
28038 2021-01-15 Carl Love <cel@us.ibm.com>
28040 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
28042 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
28043 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
28044 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
28045 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
28046 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
28047 Add builtin define.
28048 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
28049 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
28050 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
28051 New overloaded definitions.
28052 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
28053 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
28054 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
28055 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
28056 P10V_BUILTIN_MULHU_V4SI]: Add case
28057 statement for builtins.
28058 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
28059 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
28060 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
28061 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
28062 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
28063 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
28064 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
28065 Add define_insn, mode is VIlong.
28066 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
28067 Add builtin descriptions.
28069 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
28071 * final.c (final_start_function_1): Reset force_source_line.
28073 2021-01-15 Jakub Jelinek <jakub@redhat.com>
28075 PR tree-optimization/96669
28076 * match.pd (((1 << A) & 1) != 0 -> A == 0,
28077 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
28078 possibly different power of two constants and to right shift too.
28080 2021-01-15 Jakub Jelinek <jakub@redhat.com>
28082 PR tree-optimization/96681
28083 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
28084 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
28085 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
28086 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
28088 2021-01-15 Alexandre Oliva <oliva@adacore.com>
28090 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
28092 2021-01-15 Tamar Christina <tamar.christina@arm.com>
28094 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
28095 cmul<conj_op><mode>3): New.
28096 * config/aarch64/iterators.md (UNSPEC_FCMUL,
28097 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
28098 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
28099 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
28100 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
28101 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
28102 (rot_op): Renamed to conj_op.
28103 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
28104 cmul<conj_op><mode>3): New.
28105 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
28106 cmul<conj_op><mode>3): New.
28108 2021-01-15 David Malcolm <dmalcolm@redhat.com>
28112 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28113 Escape the tempfile name when constructing the expected output.
28115 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28117 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
28119 (aarch64_<su>mlsl_hi<mode>): ... This.
28120 (aarch64_<su>mlsl_hi<mode>): Define.
28121 (*aarch64_<su>mlsl<mode): Rename to...
28122 (aarch64_<su>mlsl<mode): ... This.
28123 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
28124 smlsl_hi, umlsl_hi): Define builtins.
28125 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
28126 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
28127 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
28128 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
28130 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
28132 * config/i386/i386-c.c (ix86_target_macros):
28133 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
28135 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28138 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
28139 * Makefile.in (RTL_SSA_H): New variable.
28140 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
28141 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
28142 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
28144 * config/aarch64/aarch64-cc-fusion.cc: New file.
28146 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28148 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
28149 calling cancel_changes for changes that no longer exist.
28151 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28153 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
28154 (function_info::reg_defs): ...this.
28155 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
28156 (function_info::reg_defs): ...this.
28158 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28161 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28163 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28166 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28169 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28171 2021-01-15 Richard Biener <rguenther@suse.de>
28173 PR tree-optimization/96376
28174 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
28175 for VMAT_INVARIANT.
28177 2021-01-15 Martin Liska <mliska@suse.cz>
28179 * doc/install.texi: Document that some tests need pytest module.
28180 * doc/sourcebuild.texi: Likewise.
28182 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28185 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28187 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28189 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
28190 (mve_vshrq_n_u<mode>_imm): Likewise.
28191 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
28192 * config/arm/vec-common.md: ... here.
28194 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28196 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
28198 * config/arm/neon.md (vashl<mode>3): Delete.
28199 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
28200 (vasl<mode>3): New expander.
28202 2021-01-15 Richard Biener <rguenther@suse.de>
28204 PR tree-optimization/98685
28205 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
28206 of vector extern defs.
28208 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28211 * diagnostic.c (diagnostic_kind_text): Break out this array
28213 (diagnostic_build_prefix): ...here.
28214 (fancy_abort): Detect when diagnostic_initialize has not yet been
28215 called and fall back to a minimal implementation of printing the
28216 ICE, rather than segfaulting in internal_error.
28218 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28220 * diagnostic.c (diagnostic_initialize): Eliminate
28221 parseable_fixits_p in favor of initializing extra_output_kind from
28222 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28223 (convert_column_unit): New function, split out from...
28224 (diagnostic_converted_column): ...this.
28225 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
28226 Use them to call convert_column_unit on the column values.
28227 (diagnostic_report_diagnostic): Eliminate conditional on
28228 parseable_fixits_p in favor of a switch statement on
28229 extra_output_kind, passing the appropriate values to the new
28230 params of print_parseable_fixits.
28231 (selftest::test_print_parseable_fixits_none): Update for new
28232 params of print_parseable_fixits.
28233 (selftest::test_print_parseable_fixits_insert): Likewise.
28234 (selftest::test_print_parseable_fixits_remove): Likewise.
28235 (selftest::test_print_parseable_fixits_replace): Likewise.
28236 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28238 (selftest::diagnostic_c_tests): Call it.
28239 * diagnostic.h (enum diagnostics_extra_output_kind): New.
28240 (diagnostic_context::parseable_fixits_p): Delete field in favor
28242 (diagnostic_context::extra_output_kind): ...this new field.
28243 * doc/invoke.texi (Environment Variables): Add
28244 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28245 * opts.c (common_handle_option): Update handling of
28246 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
28249 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28251 * tree-vect-slp-patterns.c (class complex_operations_pattern,
28252 complex_operations_pattern::matches,
28253 complex_operations_pattern::recognize,
28254 complex_operations_pattern::build): New.
28255 (slp_patterns): Use it.
28257 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28259 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
28260 * optabs.def (cmls_optab, cmls_conj_optab): New.
28261 * doc/md.texi: Document them.
28262 * tree-vect-slp-patterns.c (class complex_fms_pattern,
28263 complex_fms_pattern::matches, complex_fms_pattern::recognize,
28264 complex_fms_pattern::build): New.
28266 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28268 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
28269 * optabs.def (cmla_optab, cmla_conj_optab): New.
28270 * doc/md.texi: Document them.
28271 * tree-vect-slp-patterns.c (vect_match_call_p,
28272 class complex_fma_pattern, vect_slp_reset_pattern,
28273 complex_fma_pattern::matches, complex_fma_pattern::recognize,
28274 complex_fma_pattern::build): New.
28276 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28278 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
28279 * optabs.def (cmul_optab, cmul_conj_optab): New.
28280 * doc/md.texi: Document them.
28281 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
28282 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
28283 vect_build_combine_node, class complex_mul_pattern,
28284 complex_mul_pattern::matches, complex_mul_pattern::recognize,
28285 complex_mul_pattern::build): New.
28287 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28289 * tree-vect-slp.c (optimize_load_redistribution_1): New.
28290 (optimize_load_redistribution, vect_is_slp_load_node): New.
28291 (vect_match_slp_patterns): Use it.
28293 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28295 * tree-vect-slp-patterns.c (complex_add_pattern::build):
28298 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
28300 * config/gcn/mkoffload.c (main): Create an offload image only in
28301 64-bit configurations.
28303 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
28306 * config/i386/i386-options.c (ix86_option_override_internal):
28307 Issue an error for -fcf-protection with CF_BRANCH when compiling
28308 for 32-bit non-TARGET_CMOV targets.
28310 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28313 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
28314 Remove declaration and initialization of shadow variable "ret".
28315 (ix86_option_override_internal): Remove delcaration of
28316 shadow variable "i". Redeclare shadowed variable to unsigned.
28317 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
28318 * config/i386/i386-builtins.c (get_builtin_code_for_version):
28319 Update for redeclaration.
28320 * config/i386/i386.h (pta_size): Ditto.
28322 2021-01-14 Richard Biener <rguenther@suse.de>
28324 PR tree-optimization/98674
28325 * tree-data-ref.c (base_supports_access_fn_components_p): New.
28326 (initialize_data_dependence_relation): For two bases without
28327 possible access fns resort to type size equality when determining
28328 shape compatibility.
28330 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28333 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
28334 <=, >= operators in vcle and vcge intrinsics respectively.
28335 * config/arm/arm_neon_builtins.def: Remove entry for
28338 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28341 * config/i386/i386-options.c (ix86_function_specific_save):
28342 Remove redundant assignment to opts->x_ix86_branch_cost.
28343 * config/i386/i386.c (ix86_prefetch_sse):
28344 Rename from x86_prefetch_sse. Update all uses.
28345 * config/i386/i386.h: Update for rename.
28346 * config/i386/i386-options.h: Ditto.
28348 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28351 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
28352 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
28353 Use Bm instead of m for non-avx. Add isa attribute.
28355 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28357 PR tree-optimization/96688
28358 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
28359 ~X can be simplified.
28361 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
28363 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
28364 IFN_LOAD_LANES results.
28366 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28368 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
28370 (aarch64_xtn<mode>): Likewise.
28371 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
28374 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28376 (vmovl_s16): Likewise.
28377 (vmovl_s32): Likewise.
28378 (vmovl_u8): Likewise.
28379 (vmovl_u16): Likewise.
28380 (vmovl_u32): Likewise.
28381 (vmovn_s16): Likewise.
28382 (vmovn_s32): Likewise.
28383 (vmovn_s64): Likewise.
28384 (vmovn_u16): Likewise.
28385 (vmovn_u32): Likewise.
28386 (vmovn_u64): Likewise.
28388 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28390 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
28392 (aarch64_<su>qxtn2<mode>_be): Likewise.
28393 (aarch64_<su>qxtn2<mode>): Likewise.
28394 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
28396 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
28397 (su): Handle ss_truncate and us_truncate.
28398 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
28400 (vqmovn_high_s32): Likewise.
28401 (vqmovn_high_s64): Likewise.
28402 (vqmovn_high_u16): Likewise.
28403 (vqmovn_high_u32): Likewise.
28404 (vqmovn_high_u64): Likewise.
28406 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28408 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
28410 (aarch64_xtn2<mode>_be): Likewise.
28411 (aarch64_xtn2<mode>): Likewise.
28412 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
28414 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
28416 (vmovn_high_s32): Likewise.
28417 (vmovn_high_s64): Likewise.
28418 (vmovn_high_u16): Likewise.
28419 (vmovn_high_u32): Likewise.
28420 (vmovn_high_u64): Likewise.
28422 2021-01-13 Stafford Horne <shorne@gmail.com>
28424 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
28426 2021-01-13 Stafford Horne <shorne@gmail.com>
28428 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
28430 2021-01-13 Stafford Horne <shorne@gmail.com>
28432 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
28433 define for __or1k_hard_float__.
28435 2021-01-13 Stafford Horne <shorne@gmail.com>
28437 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
28438 (PROFILE_HOOK): Define to call _mcount.
28439 (FUNCTION_PROFILER): Change from abort to no-op.
28441 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28443 PR tree-optimization/96691
28444 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
28445 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
28446 (~D ^ C) or (D ^ C) can be simplified.
28448 2021-01-13 Richard Biener <rguenther@suse.de>
28450 PR tree-optimization/92645
28451 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
28452 until after vector lowering.
28454 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28456 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
28458 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
28459 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
28461 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28463 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
28465 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
28466 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
28468 2021-01-13 Richard Biener <rguenther@suse.de>
28470 PR tree-optimization/92645
28471 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
28472 BIT_FIELD_REF argument.
28473 (vect_build_slp_tree_2): Record the desired vector type
28474 on the external vector def.
28475 (vectorizable_slp_permutation): Handle required punning
28476 of existing vector defs.
28478 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28480 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
28482 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28484 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
28486 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
28488 * config.gcc [$target == *-*-gnu*]: Enable
28489 'default_gnu_indirect_function'.
28491 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28494 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
28495 registers before calling targetm.vectorize.vec_perm_const, only after
28497 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
28498 two argument permutation when one operand is zero vector and only
28499 after that force operands into registers.
28500 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
28501 define_insn_and_split pattern.
28502 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
28503 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
28504 (*avx2_zero_extendv8hiv8si2_1): Likewise.
28505 (*avx512f_zero_extendv8siv8di2_1): Likewise.
28506 (*avx2_zero_extendv4siv4di2_1): Likewise.
28507 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
28509 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
28510 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
28511 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
28512 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
28513 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
28514 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
28516 2021-01-13 Martin Liska <mliska@suse.cz>
28518 PR tree-optimization/98455
28519 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
28520 Record also virtual PHIs.
28521 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
28524 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
28526 * doc/invoke.texi (C++ Modules): Fix typos.
28528 2021-01-13 Richard Biener <rguenther@suse.de>
28530 PR tree-optimization/98640
28531 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
28532 handle plus or minus from a truncated operand to be
28535 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28538 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
28539 define_insn_and_split patterns.
28540 (splitter after *btr<mode>_2): New splitter.
28542 2021-01-13 Martin Liska <mliska@suse.cz>
28545 * cgraphunit.c (analyze_functions): Remove dead code.
28547 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
28549 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
28550 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
28551 (a64fx_regmove_cost, a64fx_vector_cost): New.
28552 (a64fx_tunings): Use the new added cost tables.
28554 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28557 * config/i386/predicates.md (pmovzx_parallel): New predicate.
28558 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
28559 define_insn_and_split pattern.
28560 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
28561 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
28563 2021-01-13 Julian Brown <julian@codesourcery.com>
28565 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
28566 to fix v0 register.
28568 2021-01-13 Julian Brown <julian@codesourcery.com>
28570 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
28573 2021-01-13 Julian Brown <julian@codesourcery.com>
28575 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
28576 for reciprocal-approximation instructions.
28577 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
28578 refinement and division result.
28579 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
28581 2021-01-13 Julian Brown <julian@codesourcery.com>
28583 * config/gcn/gcn-valu.md (subdf): Rename to...
28586 2021-01-12 Martin Liska <mliska@suse.cz>
28588 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
28590 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
28592 * function-abi.h: Fix typo.
28594 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
28598 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
28599 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
28600 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
28601 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
28602 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
28603 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
28604 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
28605 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
28606 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
28607 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
28608 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
28609 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
28610 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
28611 (ARM_HAVE_V2DI_LDST): Likewise.
28612 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
28613 (*movmisalign<mode>_mve_load): New pattern.
28614 * config/arm/neon.md (movmisalign<mode>): Move to ...
28615 * config/arm/vec-common.md: ... here.
28617 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
28620 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
28621 of pattern 'plus (plus (hard reg, const), pseudo)'.
28623 2021-01-12 Richard Biener <rguenther@suse.de>
28625 PR tree-optimization/98550
28626 * tree-vect-slp.c (vect_record_max_nunits): Check whether
28627 the group size is a multiple of the vector element count.
28628 (vect_build_slp_tree_1): When we need to fail because
28629 the vector type choosen causes unrolling do so lazily
28630 without affecting matches only at the end to guide group splitting.
28632 2021-01-12 Martin Liska <mliska@suse.cz>
28635 * optc-save-gen.awk: Compare also n_target_save vars with
28638 2021-01-12 Martin Liska <mliska@suse.cz>
28640 * gcov.c (source_info::debug): New.
28641 (print_usage): Add --debug (-D) option.
28642 (process_args): Likewise.
28643 (generate_results): Call src->debug after
28644 accumulate_line_counts.
28645 (read_graph_file): Properly assign id for EXIT_BLOCK.
28646 * profile.c (branch_prob): Dump function body before it is
28649 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28651 PR tree-optimization/98629
28652 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
28653 unless returning non-zero.
28655 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28657 PR tree-optimization/95731
28658 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
28659 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
28660 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
28661 only after optimize_range_tests_var_bound.
28663 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28665 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
28666 * configure: Regenerated.
28668 2021-01-12 liuhongt <hongtao.liu@intel.com>
28671 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
28673 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
28676 2021-01-12 Alexandre Oliva <oliva@adacore.com>
28678 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
28680 (auto_end_imm_use_stmt_traverse): New struct.
28681 (FOR_EACH_IMM_USE_STMT): Use it.
28682 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
28684 * gimple-ssa-strength-reduction.c: ... here, ...
28685 * graphite-scop-detection.c: ... here, ...
28686 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
28687 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
28688 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
28689 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
28690 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
28691 * tree-vect-slp.c: ... and here, ...
28692 * doc/tree-ssa.texi: ... and the example here.
28694 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28696 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
28697 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
28698 (*sdiv_pow2<mode>3): New pattern.
28699 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
28700 Wrap the ASRD in an UNSPEC_PRED_X.
28701 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
28702 predicate with a constant PTRUE, if it isn't already.
28703 (*cond_<sve_int_op><mode>_z): Replace with...
28704 (*cond_<sve_int_op><mode>_any): ...this new pattern.
28706 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28708 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
28709 SVE_FULL_I to SVE_I.
28710 (*cond_bic<mode>_any): Likewise.
28712 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28714 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
28715 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
28718 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28720 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
28721 SVE_FULL_I to SVE_I.
28722 (*aarch64_cond_<su>abd<mode>_2): Likewise.
28723 (*aarch64_cond_<su>abd<mode>_any): Likewise.
28724 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
28725 for the max and min but not for the minus.
28726 (*aarch64_cond_<su>abd<mode>_3): New pattern.
28728 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28730 * config/aarch64/iterators.md (SVE_24I): New iterator.
28731 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
28732 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
28734 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28736 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
28737 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
28739 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
28740 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
28741 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
28742 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
28744 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28746 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
28747 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
28748 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
28751 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28753 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
28754 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
28755 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
28757 2021-01-11 Martin Liska <mliska@suse.cz>
28760 * symtab-clones.h (clone_info::release): Release
28761 symtab::m_clones with ggc_delete as it's a GGC memory.
28763 2021-01-11 Matthias Klose <doko@ubuntu.com>
28765 * Makefile.in (LINK_PROGRESS): Show the link target.
28767 2021-01-11 Richard Biener <rguenther@suse.de>
28769 PR tree-optimization/91403
28770 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
28771 single-element interleaving group size at 4096 elements.
28773 2021-01-11 Richard Biener <rguenther@suse.de>
28775 PR tree-optimization/98526
28776 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
28777 of the actual reduction op for the regular case.
28778 (vectorizable_reduction): Cost the stmts
28779 vect_transform_reduction produces here.
28781 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
28783 * tree-ssa-forwprop.c (simplify_vector_constructor): For
28784 big-endian, use UNPACK[_FLOAT]_HI.
28786 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28788 * tree-vect-slp-patterns.c (class complex_pattern,
28789 class complex_add_pattern): Add parameters to matches.
28790 (complex_add_pattern::build): Free memory.
28791 (complex_add_pattern::matches): Move validation end of match.
28792 (complex_add_pattern::recognize): Likewise.
28794 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28796 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
28798 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28800 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
28802 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28804 PR tree-optimization/95867
28805 * tree-ssa-math-opts.h: New header.
28806 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
28807 (powi_as_mults): No longer static. Use build_one_cst instead of
28808 build_real. Formatting fix.
28809 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
28810 (attempt_builtin_powi): Handle multiplication reassociation without
28811 powi_fndecl using powi_as_mults.
28812 (reassociate_bb): For integral types don't require
28813 -funsafe-math-optimizations to call attempt_builtin_powi.
28815 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28817 PR tree-optimization/95852
28818 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
28819 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
28820 allow in the bb any of the stmts in that vector, div_stmt and
28821 up to 3 cast stmts.
28822 (arith_cast_equal_p): New function.
28823 (arith_overflow_check_p): Add cast_stmt argument, handle signed
28824 multiply overflow checks.
28825 (match_arith_overflow): Adjust caller. Handle signed multiply
28828 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28830 PR tree-optimization/95852
28831 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
28832 (uaddsub_overflow_check_p): Renamed to ...
28833 (arith_overflow_check_p): ... this. Handle also multiplication
28834 with overflow check.
28835 (match_uaddsub_overflow): Renamed to ...
28836 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
28837 also multiplication with overflow check. Adjust function comment.
28838 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
28839 match_arith_overflow also for MULT_EXPR.
28841 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28843 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28844 __builtin_convertvector.
28845 (vmovl_s16): Likewise.
28846 (vmovl_s32): Likewise.
28847 (vmovl_u8): Likewise.
28848 (vmovl_u16): Likewise.
28849 (vmovl_u32): Likewise.
28850 (vmovn_s16): Likewise.
28851 (vmovn_s32): Likewise.
28852 (vmovn_s64): Likewise.
28853 (vmovn_u16): Likewise.
28854 (vmovn_u32): Likewise.
28855 (vmovn_u64): Likewise.
28857 2021-01-11 Martin Liska <mliska@suse.cz>
28859 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
28860 (if_chain::is_beneficial): Delete clusters
28861 (find_conditions): Make second argument of conditions_in_bbs a
28862 pointer so that we control over it's lifetime.
28863 (pass_if_to_switch::execute): Delete them.
28865 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
28867 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
28870 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28872 * config/vax/vax.md (cc): Remove mode attribute.
28873 (subst_<cc>, subst_f<cc>): Rename to...
28874 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
28875 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
28876 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
28877 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
28879 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28881 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
28882 `const_double_zero'.
28884 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28886 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
28887 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
28890 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28892 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
28894 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
28895 with `const_double_zero'.
28896 * doc/rtl.texi (Constant Expression Types): Document it.
28898 2021-01-09 Jakub Jelinek <jakub@redhat.com>
28901 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
28902 POINTER_DIFF_EXPR to be any integral type.
28904 2021-01-09 Jakub Jelinek <jakub@redhat.com>
28906 PR rtl-optimization/98603
28907 * function.c (instantiate_virtual_regs_in_insn): For asm goto
28908 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
28909 if any, set ASM_OPERANDS mode to VOIDmode and change
28910 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
28912 2021-01-09 Alexandre Oliva <oliva@gnu.org>
28915 * final.c (notice_source_line): Narrow down the condition to
28916 skip a line-0 marker.
28918 2021-01-08 Sergei Trofimovich <siarheit@google.com>
28920 * ipa-modref.c (merge_call_side_effects): Fix
28921 linebreak split by reordering two print calls.
28923 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
28925 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
28926 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
28927 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
28928 consistency, fix constraint.
28930 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
28932 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
28933 callables instead of mask values.
28934 (struct target_flag_set_p): New predicate.
28935 (s390_cpu_cpp_builtins_internal): Define or undefine
28936 __LONG_DOUBLE_VX__ macro.
28938 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
28941 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
28942 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
28945 2021-01-08 Richard Biener <rguenther@suse.de>
28947 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
28949 2021-01-08 Richard Biener <rguenther@suse.de>
28951 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
28952 (vect_build_slp_tree): On cache hit release the matched
28953 scalar stmts vector.
28954 * tree-vect-stmts.c (vectorizable_store): Properly free
28955 vec_oprnds before possibly gathering them again.
28957 2021-01-08 Richard Biener <rguenther@suse.de>
28959 PR tree-optimization/98544
28960 * tree-vect-slp.c (vect_optimize_slp): Always materialize
28961 permutes at a permute node.
28963 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
28966 * config/i386/i386.c (x86_function_profiler): Use R10 to call
28967 mcount in large model. Sorry for large model with PIC.
28969 2021-01-08 Jakub Jelinek <jakub@redhat.com>
28972 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
28973 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
28974 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
28975 TargetSave and initialize for variables with enum types.
28976 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
28977 mstack-protector-guard-symbol=): Add Save.
28978 * config/i386/i386-options.c (ix86_function_specific_save,
28979 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
28980 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
28981 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
28982 x_ix86_veclibabi_type.
28984 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
28986 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
28987 SVE_FULL_I to SVE_I.
28988 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
28990 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
28992 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
28993 SVE_FULL_I to SVE_I.
28994 (*cond_uxt<mode>_any): Likewise.
28996 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28998 * config/aarch64/iterators.md (Vwhalf): New iterator.
28999 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
29001 (aarch64_<sur>adalp<mode>): ... This. Make more
29003 (<sur>sadv16qi): Adjust callsite of the above.
29004 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
29006 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
29008 (vpadal_s16): Likewise.
29009 (vpadal_u8): Likewise.
29010 (vpadal_u16): Likewise.
29011 (vpadalq_s8): Likewise.
29012 (vpadalq_s16): Likewise.
29013 (vpadalq_s32): Likewise.
29014 (vpadalq_u8): Likewise.
29015 (vpadalq_u16): Likewise.
29016 (vpadalq_u32): Likewise.
29018 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29020 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
29022 (aarch64_<su>abd<mode>): ... This.
29023 (<sur>sadv16qi): Adjust callsite of the above.
29024 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
29026 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
29028 (vabd_s16): Likewise.
29029 (vabd_s32): Likewise.
29030 (vabd_u8): Likewise.
29031 (vabd_u16): Likewise.
29032 (vabd_u32): Likewise.
29033 (vabdq_s8): Likewise.
29034 (vabdq_s16): Likewise.
29035 (vabdq_s32): Likewise.
29036 (vabdq_u8): Likewise.
29037 (vabdq_u16): Likewise.
29038 (vabdq_u32): Likewise.
29040 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29042 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
29044 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
29045 (vaba_s16): Likewise.
29046 (vaba_s32): Likewise.
29047 (vaba_u8): Likewise.
29048 (vaba_u16): Likewise.
29049 (vaba_u32): Likewise.
29050 (vabaq_s8): Likewise.
29051 (vabaq_s16): Likewise.
29052 (vabaq_s32): Likewise.
29053 (vabaq_u8): Likewise.
29054 (vabaq_u16): Likewise.
29055 (vabaq_u32): Likewise.
29057 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29059 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
29060 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
29061 Change RTL pattern to match.
29063 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
29065 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
29066 * config/riscv/riscv-c.c (riscv-subset.h): New.
29067 (INCLUDE_STRING): Define.
29068 (riscv_cpu_cpp_builtins): Add new style architecture extension
29070 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
29071 (riscv_subset_list::end): New.
29072 (riscv_current_subset_list): New.
29074 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
29076 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
29077 Move to riscv-subset.h.
29078 (struct riscv_subset_t): Ditto.
29079 (class riscv_subset_list): Ditto.
29080 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
29081 from riscv-common.c.
29082 (struct riscv_subset_t): Ditto.
29083 (class riscv_subset_list): Ditto.
29084 * config/riscv/t-riscv ($(common_out_file)): Add file
29087 2021-01-07 Jakub Jelinek <jakub@redhat.com>
29090 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
29091 New define_insn patterns.
29093 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29095 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
29096 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
29097 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
29099 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29101 PR tree-optimization/98560
29102 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
29103 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
29105 (vec_cond_direct): Likewise argument 2.
29106 (vec_condu_direct, vec_condeq_direct): Delete.
29107 (expand_vect_cond_optab_fn): Rename to...
29108 (expand_vec_cond_optab_fn): ...this, replacing old macro.
29109 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
29110 (expand_vect_cond_mask_optab_fn): Rename to...
29111 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
29112 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
29114 (direct_vec_cond_optab_supported_p): Likewise.
29115 (direct_vec_condu_optab_supported_p): Delete.
29116 (direct_vec_condeq_optab_supported_p): Delete.
29117 * gimple-isel.cc: Include internal-fn.h.
29118 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
29121 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29123 PR tree-optimization/98560
29124 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
29125 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
29127 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
29129 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
29130 rotate_insn and optab code attributes.
29131 Update all uses to merged code attribute.
29132 * config/i386/sse.md: Update all uses to merged code attribute.
29133 * config/i386/mmx.md: Update all uses to merged code attribute.
29135 2021-01-07 Jakub Jelinek <jakub@redhat.com>
29137 PR tree-optimization/98568
29138 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
29139 (bswap_replace): Use it.
29141 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
29143 PR rtl-optimization/97978
29144 * lra-int.h (lra_hard_reg_split_p): New external.
29145 * lra.c (lra_hard_reg_split_p): New global.
29146 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
29147 * lra-assigns.c (lra_assign): Don't check allocation correctness
29148 after hard reg splitting.
29150 2021-01-06 Martin Sebor <msebor@redhat.com>
29153 * builtins.c (new_delete_mismatch_p): New overload.
29154 (new_delete_mismatch_p (tree, tree)): Call it.
29156 2021-01-06 Alexandre Oliva <oliva@adacore.com>
29158 * Makefile.in (T_GLIMITS_H): New.
29159 (stmp-int-hdrs): Depend on it, use it.
29160 * config/t-vxworks (T_GLIMITS_H): Override it.
29161 (vxw-glimits.h): New.
29163 2021-01-06 Richard Biener <rguenther@suse.de>
29165 PR tree-optimization/98513
29166 * value-range.cc (intersect_ranges): Compare the upper bounds
29167 for the expected relation.
29169 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
29172 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
29174 * doc/standards.texi (HSAIL): Remove section.
29176 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
29178 * configure: Re-generate.
29180 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29182 * doc/invoke.texi (-std=c++20): Adjust for the publication of
29183 ISO 14882:2020 standard.
29184 * doc/standards.texi: Likewise.
29186 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29188 PR tree-optimization/94802
29189 * expr.h (maybe_optimize_sub_cmp_0): Declare.
29190 * expr.c: Include tree-pretty-print.h and flags.h.
29191 (maybe_optimize_sub_cmp_0): New function.
29192 (do_store_flag): Use it.
29193 * cfgexpand.c (expand_gimple_cond): Likewise.
29195 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29197 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
29198 * rtlanal.c (simple_regno_set): Tweak description to clarify the
29201 2021-01-05 Richard Biener <rguenther@suse.de>
29203 PR tree-optimization/98516
29204 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
29205 lanes when materializing on a VEC_PERM node.
29206 (vectorizable_slp_permutation): Dump the permute properly.
29208 2021-01-05 Richard Biener <rguenther@suse.de>
29210 * tree-vect-slp.c (vect_slp_region): Move debug counter
29211 to cover individual subgraphs.
29213 2021-01-05 Richard Biener <rguenther@suse.de>
29215 PR tree-optimization/98428
29216 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
29217 vector lane extracts for loop vectorization.
29219 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29221 PR tree-optimization/98514
29222 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
29224 (operand_rank): Change type from hash_map<tree, long> to
29225 hash_map<tree, int64_t>.
29226 (phi_rank): Change return type from long to int64_t.
29227 (loop_carried_phi): Change block_rank variable type from long to
29229 (propagate_rank): Change return type, rank parameter type and
29230 op_rank variable type from long to int64_t.
29231 (find_operand_rank): Change return type from long to int64_t
29232 and change slot variable type from long * to int64_t *.
29233 (insert_operand_rank): Change rank parameter type from long to
29235 (get_rank): Change return type and rank variable type from long to
29236 int64_t. Use PRId64 instead of ld to print the rank.
29237 (init_reassoc): Change rank variable type from long to int64_t
29238 and adjust correspondingly bb_rank and operand_rank initialization.
29240 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29242 PR tree-optimization/96928
29243 * tree-ssa-phiopt.c (xor_replacement): New function.
29244 (tree_ssa_phiopt_worker): Call it.
29246 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29248 PR tree-optimization/96930
29249 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
29250 from narrower value which has the same type as 1 << B, perform
29251 the right shift on the narrower value followed by extension.
29253 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29255 PR tree-optimization/96239
29256 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
29258 (get_status_for_store_merging): Don't return BB_INVALID for blocks
29259 with potential bswap optimizable CONSTRUCTORs.
29260 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
29263 2021-01-05 Richard Biener <rguenther@suse.de>
29265 PR tree-optimization/98381
29266 * tree.c (vector_element_bits): Properly compute bool vector
29268 * tree-vect-loop.c (vectorizable_live_operation): Properly
29269 compute the last lane bit offset.
29271 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29274 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
29275 Clear the top 64 bytes of the input XMM register.
29276 (sse_cvttps2pi): Ditto.
29278 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29281 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
29283 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
29286 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
29289 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
29291 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
29292 (umaddsidi4_split): Likewise.
29294 2021-01-05 liuhongt <hongtao.liu@intel.com>
29297 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
29298 define_insn_and_split for zero_extend of subreg HI of pmovskb
29300 (*sse2_pmovskb_zexthisi): Add new combine splitters for
29301 zero_extend of not of subreg HI of pmovskb result.
29303 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29306 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
29308 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
29309 convert_memory_address to convert symbolic immediates to ptr_mode
29310 before forcing them to memory.
29312 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29314 PR rtl-optimization/97144
29315 * recog.c (constrain_operands): Initialize matching_operand
29316 for each alternative, rather than only doing it once.
29318 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29320 PR rtl-optimization/98403
29321 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
29322 why we don't remove call clobbers.
29323 (function_info::apply_changes_to_insn): Don't attempt to add
29324 call clobbers here.
29326 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29328 PR tree-optimization/98371
29329 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
29330 (vect_analyze_loop): If an epilogue loop appears to be cheaper
29331 than the main loop, re-analyze it as a main loop before adopting
29334 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29337 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
29338 * aclocal.m4, configure: Regenerate.
29339 * Makefile.in (NETLIBS): Define.
29340 (BACKEND): Remove $(CODYLIB).
29342 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29344 PR rtl-optimization/98334
29345 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
29346 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
29348 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
29350 * tree-inline.c (expand_call_inline): Restore input_location.
29351 Return result from recursive call.
29353 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29355 PR tree-optimization/95401
29356 * config/aarch64/aarch64-sve-builtins.cc
29357 (gimple_folder::load_store_cookie): Use bits rather than bytes
29358 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
29359 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
29360 * tree-vect-stmts.c (vectorizable_store): Likewise.
29361 (vectorizable_load): Likewise.
29363 2021-01-04 Richard Biener <rguenther@suse.de>
29365 PR tree-optimization/98308
29366 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
29369 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29371 PR tree-optimization/95771
29372 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
29373 with precision smaller than int's precision and types with precision
29374 twice as large as long long. Formatting fixes.
29376 2021-01-04 Richard Biener <rguenther@suse.de>
29378 PR tree-optimization/98464
29379 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
29380 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
29381 (process_bb): Adjust.
29383 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
29386 * doc/invoke.texi (-fsanitize=address): Fix wording describing
29387 clash with -fsanitize=hwaddress.
29389 2021-01-04 Richard Biener <rguenther@suse.de>
29391 PR tree-optimization/98282
29392 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
29393 invariants as VN_NARY.
29395 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29398 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
29399 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
29400 to handle zero operands.
29402 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29404 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
29405 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
29406 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
29409 2021-01-04 Richard Biener <rguenther@suse.de>
29411 PR tree-optimization/98393
29412 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
29413 when hitting the limit.
29415 2021-01-04 Richard Biener <rguenther@suse.de>
29417 PR tree-optimization/98291
29418 * tree-vect-loop.c (vectorizable_reduction): Bypass
29419 associativity check for SLP reductions with VF 1.
29421 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29423 PR tree-optimization/96782
29424 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
29426 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
29428 * collect-utils.c (collect_execute): Check dumppfx.
29429 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
29430 to collect_execute.
29431 (do_link): Add new parameter atsuffix.
29432 (main): Handle -dumpdir option. Skip one argument for
29433 -o, -isystem and -B options.
29434 * gcc.c (make_at_file): New helper function.
29435 (close_at_file): Use it.
29437 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29439 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
29440 Amend handling for LD64_VERSION fallback defaults.
29442 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29444 * config.gcc: Compute default version information
29445 from the configured target. Likewise defaults for
29447 * config/darwin10.h: Removed.
29448 * config/darwin12.h: Removed.
29449 * config/darwin9.h: Removed.
29450 * config/rs6000/darwin8.h: Removed.
29452 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29454 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
29456 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29458 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
29459 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
29461 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29463 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
29465 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
29467 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29469 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
29470 for the Darwin10 unwinder stub from here ...
29471 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
29473 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29475 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
29476 (ASM_DEBUG_SPEC):Only define if the assembler supports
29478 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
29479 (DARWIN_PREFER_DWARF): Define.
29480 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
29481 (DARWIN_PREFER_DWARF): Likewise
29482 (DSYMUTIL_SPEC): Likewise.
29483 (COLLECT_RUN_DSYMUTIL): Likewise.
29484 (ASM_DEBUG_SPEC): Likewise.
29485 (ASM_DEBUG_OPTION_SPEC): Likewise.
29487 2021-01-02 Jan Hubicka <jh@suse.cz>
29489 * cfg.c (free_block): ggc_free bb.
29491 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29493 * gcc.c (process_command): Update copyright notice dates.
29494 * gcov-dump.c (print_version): Ditto.
29495 * gcov.c (print_version): Ditto.
29496 * gcov-tool.c (print_version): Ditto.
29497 * gengtype.c (create_file): Ditto.
29498 * doc/cpp.texi: Bump @copying's copyright year.
29499 * doc/cppinternals.texi: Ditto.
29500 * doc/gcc.texi: Ditto.
29501 * doc/gccint.texi: Ditto.
29502 * doc/gcov.texi: Ditto.
29503 * doc/install.texi: Ditto.
29504 * doc/invoke.texi: Ditto.
29506 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29508 * ChangeLog-2020: Rotate ChangeLog. New file.
29511 Copyright (C) 2021 Free Software Foundation, Inc.
29513 Copying and distribution of this file, with or without modification,
29514 are permitted in any medium without royalty provided the copyright
29515 notice and this notice are preserved.
projects / thirdparty / gcc.git / blob