]> git.ipfire.org Git - thirdparty/gcc.git/blob - gcc/ChangeLog
[ARM][GCC][1/x]: MVE ACLE intrinsics framework patch.
[thirdparty/gcc.git] / gcc / ChangeLog
1 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2 Mihail Ionescu <mihail.ionescu@arm.com>
3 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
4
5 * config.gcc (arm_mve.h): Include mve intrinsics header file.
6 * config/arm/aout.h (p0): Add new register name for MVE predicated
7 cases.
8 * config/arm-builtins.c (ARM_BUILTIN_SIMD_LANE_CHECK): Define macro
9 common to Neon and MVE.
10 (ARM_BUILTIN_NEON_LANE_CHECK): Renamed to ARM_BUILTIN_SIMD_LANE_CHECK.
11 (arm_init_simd_builtin_types): Disable poly types for MVE.
12 (arm_init_neon_builtins): Move a check to arm_init_builtins function.
13 (arm_init_builtins): Use ARM_BUILTIN_SIMD_LANE_CHECK instead of
14 ARM_BUILTIN_NEON_LANE_CHECK.
15 (mve_dereference_pointer): Add function.
16 (arm_expand_builtin_args): Call to mve_dereference_pointer when MVE is
17 enabled.
18 (arm_expand_neon_builtin): Moved to arm_expand_builtin function.
19 (arm_expand_builtin): Moved from arm_expand_neon_builtin function.
20 * config/arm/arm-c.c (__ARM_FEATURE_MVE): Define macro for MVE and MVE
21 with floating point enabled.
22 * config/arm/arm-protos.h (neon_immediate_valid_for_move): Renamed to
23 simd_immediate_valid_for_move.
24 (simd_immediate_valid_for_move): Renamed from
25 neon_immediate_valid_for_move function.
26 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Generate
27 error if vfpv2 feature bit is disabled and mve feature bit is also
28 disabled for HARD_FLOAT_ABI.
29 (use_return_insn): Check to not push VFP regs for MVE.
30 (aapcs_vfp_allocate): Add MVE check to have same Procedure Call Standard
31 as Neon.
32 (aapcs_vfp_allocate_return_reg): Likewise.
33 (thumb2_legitimate_address_p): Check to return 0 on valid Thumb-2
34 address operand for MVE.
35 (arm_rtx_costs_internal): MVE check to determine cost of rtx.
36 (neon_valid_immediate): Rename to simd_valid_immediate.
37 (simd_valid_immediate): Rename from neon_valid_immediate.
38 (simd_valid_immediate): MVE check on size of vector is 128 bits.
39 (neon_immediate_valid_for_move): Rename to
40 simd_immediate_valid_for_move.
41 (simd_immediate_valid_for_move): Rename from
42 neon_immediate_valid_for_move.
43 (neon_immediate_valid_for_logic): Modify call to neon_valid_immediate
44 function.
45 (neon_make_constant): Modify call to neon_valid_immediate function.
46 (neon_vector_mem_operand): Return VFP register for POST_INC or PRE_DEC
47 for MVE.
48 (output_move_neon): Add MVE check to generate vldm/vstm instrcutions.
49 (arm_compute_frame_layout): Calculate space for saved VFP registers for
50 MVE.
51 (arm_save_coproc_regs): Save coproc registers for MVE.
52 (arm_print_operand): Add case 'E' to print memory operands for MVE.
53 (arm_print_operand_address): Check to print register number for MVE.
54 (arm_hard_regno_mode_ok): Check for arm hard regno mode ok for MVE.
55 (arm_modes_tieable_p): Check to allow structure mode for MVE.
56 (arm_regno_class): Add VPR_REGNUM check.
57 (arm_expand_epilogue_apcs_frame): MVE check to calculate epilogue code
58 for APCS frame.
59 (arm_expand_epilogue): MVE check for enabling pop instructions in
60 epilogue.
61 (arm_print_asm_arch_directives): Modify function to disable print of
62 .arch_extension "mve" and "fp" for cases where MVE is enabled with
63 "SOFT FLOAT ABI".
64 (arm_vector_mode_supported_p): Check for modes available in MVE interger
65 and MVE floating point.
66 (arm_array_mode_supported_p): Add TARGET_HAVE_MVE check for array mode
67 pointer support.
68 (arm_conditional_register_usage): Enable usage of conditional regsiter
69 for MVE.
70 (fixed_regs[VPR_REGNUM]): Enable VPR_REG for MVE.
71 (arm_declare_function_name): Modify function to disable print of
72 .arch_extension "mve" and "fp" for cases where MVE is enabled with
73 "SOFT FLOAT ABI".
74 * config/arm/arm.h (TARGET_HAVE_MVE): Disable for soft float abi and
75 when target general registers are required.
76 (TARGET_HAVE_MVE_FLOAT): Likewise.
77 (FIXED_REGISTERS): Add bit for VFP_REG class which is enabled in arm.c
78 for MVE.
79 (CALL_USED_REGISTERS): Set bit for VFP_REG class in CALL_USED_REGISTERS
80 which indicate this is not available for across function calls.
81 (FIRST_PSEUDO_REGISTER): Modify.
82 (VALID_MVE_MODE): Define valid MVE mode.
83 (VALID_MVE_SI_MODE): Define valid MVE SI mode.
84 (VALID_MVE_SF_MODE): Define valid MVE SF mode.
85 (VALID_MVE_STRUCT_MODE): Define valid MVE struct mode.
86 (VPR_REGNUM): Add Vector Predication Register in arm_regs_in_sequence
87 for MVE.
88 (IS_VPR_REGNUM): Macro to check for VPR_REG register.
89 (REG_ALLOC_ORDER): Add VPR_REGNUM entry.
90 (enum reg_class): Add VPR_REG entry.
91 (REG_CLASS_NAMES): Add VPR_REG entry.
92 * config/arm/arm.md (VPR_REGNUM): Define.
93 (conds): Check is_mve_type attrbiute to differentiate "conditional" and
94 "unconditional" instructions.
95 (arm_movsf_soft_insn): Modify RTL to not allow for MVE.
96 (movdf_soft_insn): Modify RTL to not allow for MVE.
97 (vfp_pop_multiple_with_writeback): Enable for MVE.
98 (include "mve.md"): Include mve.md file.
99 * config/arm/arm_mve.h: Add MVE intrinsics head file.
100 * config/arm/constraints.md (Up): Constraint to enable "p0" register in MVE
101 for vector predicated operands.
102 * config/arm/iterators.md (VNIM1): Define.
103 (VNINOTM1): Define.
104 (VHFBF_split): Define
105 * config/arm/mve.md: New file.
106 (mve_mov<mode>): Define RTL for move, store and load in MVE.
107 (mve_mov<mode>): Define move RTL pattern with vec_duplicate operator for
108 second operand.
109 * config/arm/neon.md (neon_immediate_valid_for_move): Rename with
110 simd_immediate_valid_for_move.
111 (neon_mov<mode>): Split pattern and move expand pattern "movv8hf" which
112 is common to MVE and NEON to vec-common.md file.
113 (vec_init<mode><V_elem_l>): Add TARGET_HAVE_MVE check.
114 * config/arm/predicates.md (vpr_register_operand): Define.
115 * config/arm/t-arm: Add mve.md file.
116 * config/arm/types.md (mve_move): Add MVE instructions mve_move to
117 attribute "type".
118 (mve_store): Add MVE instructions mve_store to attribute "type".
119 (mve_load): Add MVE instructions mve_load to attribute "type".
120 (is_mve_type): Define attribute.
121 * config/arm/vec-common.md (mov<mode>): Modify RTL expand to support
122 standard move patterns in MVE along with NEON and IWMMXT with mode
123 iterator VNIM1.
124 (mov<mode>): Modify RTL expand to support standard move patterns in NEON
125 and IWMMXT with mode iterator V8HF.
126 (movv8hf): Define RTL expand to support standard "movv8hf" pattern in
127 NEON and MVE.
128 * config/arm/vfp.md (neon_immediate_valid_for_move): Rename to
129 simd_immediate_valid_for_move.
130
131
132 2020-03-16 H.J. Lu <hongjiu.lu@intel.com>
133
134 PR target/89229
135 * config/i386/i386.md (*movsi_internal): Call ix86_output_ssemov
136 for TYPE_SSEMOV. Remove ext_sse_reg_operand and TARGET_AVX512VL
137 check.
138 * config/i386/predicates.md (ext_sse_reg_operand): Removed.
139
140 2020-03-16 Jakub Jelinek <jakub@redhat.com>
141
142 PR debug/94167
143 * tree-inline.c (insert_init_stmt): Don't gimple_regimplify_operands
144 DEBUG_STMTs.
145
146 PR tree-optimization/94166
147 * tree-ssa-reassoc.c (sort_by_mach_mode): Use SSA_NAME_VERSION
148 as secondary comparison key.
149
150 2020-03-16 Bin Cheng <bin.cheng@linux.alibaba.com>
151
152 PR tree-optimization/94125
153 * tree-loop-distribution.c
154 (loop_distribution::break_alias_scc_partitions): Update post order
155 number for merged scc.
156
157 2020-03-15 H.J. Lu <hongjiu.lu@intel.com>
158
159 PR target/89229
160 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_SI and
161 MODE_SF.
162 * config/i386/i386.md (*movsf_internal): Call ix86_output_ssemov
163 for TYPE_SSEMOV. Remove TARGET_PREFER_AVX256, TARGET_AVX512VL
164 and ext_sse_reg_operand check.
165
166 2020-03-15 Lewis Hyatt <lhyatt@gmail.com>
167
168 * common.opt: Avoid redundancy in the help text.
169 * config/arc/arc.opt: Likewise.
170 * config/cr16/cr16.opt: Likewise.
171
172 2020-03-14 Jakub Jelinek <jakub@redhat.com>
173
174 PR middle-end/93566
175 * tree-nested.c (convert_nonlocal_omp_clauses,
176 convert_local_omp_clauses): Handle {,in_,task_}reduction clauses
177 with C/C++ array sections.
178
179 2020-03-14 H.J. Lu <hongjiu.lu@intel.com>
180
181 PR target/89229
182 * config/i386/i386.md (*movdi_internal): Call ix86_output_ssemov
183 for TYPE_SSEMOV. Remove ext_sse_reg_operand and TARGET_AVX512VL
184 check.
185
186 2020-03-14 Jakub Jelinek <jakub@redhat.com>
187
188 * gimple-fold.c (gimple_fold_builtin_strncpy): Change
189 "a an" to "an" in a comment.
190 * hsa-common.h (is_a_helper): Likewise.
191 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Likewise.
192 * config/arc/arc.c (arc600_corereg_hazard): Likewise.
193 * config/s390/s390.c (s390_indirect_branch_via_thunk): Likewise.
194
195 2020-03-13 Aaron Sawdey <acsawdey@linux.ibm.com>
196
197 PR target/92379
198 * config/rs6000/rs6000.c (num_insns_constant_multi): Don't shift a
199 64-bit value by 64 bits (UB).
200
201 2020-03-13 Vladimir Makarov <vmakarov@redhat.com>
202
203 PR rtl-optimization/92303
204 * lra-spills.c (remove_pseudos): Try to simplify memory subreg.
205
206 2020-03-13 Segher Boessenkool <segher@kernel.crashing.org>
207
208 PR rtl-optimization/94148
209 PR rtl-optimization/94042
210 * df-core.c (BB_LAST_CHANGE_AGE): Delete.
211 (df_worklist_propagate_forward): New parameter last_change_age, use
212 that instead of bb->aux.
213 (df_worklist_propagate_backward): Ditto.
214 (df_worklist_dataflow_doublequeue): Use a local array last_change_age.
215
216 2020-03-13 Richard Biener <rguenther@suse.de>
217
218 PR tree-optimization/94163
219 * tree-ssa-pre.c (create_expression_by_pieces): Check
220 whether alignment would be zero.
221
222 2020-03-13 Martin Liska <mliska@suse.cz>
223
224 PR lto/94157
225 * lto-wrapper.c (run_gcc): Use concat for appending
226 to collect_gcc_options.
227
228 2020-03-13 Jakub Jelinek <jakub@redhat.com>
229
230 PR target/94121
231 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use gen_int_mode
232 instead of GEN_INT.
233
234 2020-03-13 H.J. Lu <hongjiu.lu@intel.com>
235
236 PR target/89229
237 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DF.
238 * config/i386/i386.md (*movdf_internal): Call ix86_output_ssemov
239 for TYPE_SSEMOV. Remove TARGET_AVX512F, TARGET_PREFER_AVX256,
240 TARGET_AVX512VL and ext_sse_reg_operand check.
241
242 2020-03-13 Bu Le <bule1@huawei.com>
243
244 PR target/94154
245 * config/aarch64/aarch64.opt (-param=aarch64-float-recp-precision=)
246 (-param=aarch64-double-recp-precision=): New options.
247 * doc/invoke.texi: Document them.
248 * config/aarch64/aarch64.c (aarch64_emit_approx_div): Use them
249 instead of hard-coding the choice of 1 for float and 2 for double.
250
251 2020-03-13 Eric Botcazou <ebotcazou@adacore.com>
252
253 PR rtl-optimization/94119
254 * resource.h (clear_hashed_info_until_next_barrier): Declare.
255 * resource.c (clear_hashed_info_until_next_barrier): New function.
256 * reorg.c (add_to_delay_list): Fix formatting.
257 (relax_delay_slots): Call clear_hashed_info_until_next_barrier on
258 the next instruction after removing a BARRIER.
259
260 2020-03-13 Eric Botcazou <ebotcazou@adacore.com>
261
262 PR middle-end/92071
263 * expmed.c (store_integral_bit_field): For fields larger than a word,
264 call extract_bit_field on the value if the mode is BLKmode. Remove
265 specific path for big-endian targets and tidy things up a little bit.
266
267 2020-03-12 Richard Sandiford <richard.sandiford@arm.com>
268
269 PR rtl-optimization/90275
270 * cse.c (cse_insn): Delete no-op register moves too.
271
272 2020-03-12 Darius Galis <darius.galis@cyberthorstudios.com>
273
274 * config/rx/rx.md (CTRLREG_CPEN): Remove.
275 * config/rx/rx.c (rx_print_operand): Remove CTRLREG_CPEN support.
276
277 2020-03-12 Richard Biener <rguenther@suse.de>
278
279 PR tree-optimization/94103
280 * tree-ssa-sccvn.c (visit_reference_op_load): Avoid type
281 punning when the mode precision is not sufficient.
282
283 2020-03-12 H.J. Lu <hongjiu.lu@intel.com>
284
285 PR target/89229
286 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DI,
287 MODE_V1DF and MODE_V2SF.
288 * config/i386/mmx.md (MMXMODE:*mov<mode>_internal): Call
289 ix86_output_ssemov for TYPE_SSEMOV. Remove ext_sse_reg_operand
290 check.
291
292 2020-03-12 Jakub Jelinek <jakub@redhat.com>
293
294 * doc/tm.texi.in (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Change
295 ASM_OUTPUT_ALIGNED_DECL in description to ASM_OUTPUT_ALIGNED_LOCAL
296 and ASM_OUTPUT_DECL to ASM_OUTPUT_LOCAL.
297 * doc/tm.texi: Regenerated.
298
299 PR tree-optimization/94130
300 * tree-ssa-dse.c: Include gimplify.h.
301 (increment_start_addr): If stmt has lhs, drop the lhs from call and
302 set it after the call to the original value of the first argument.
303 Formatting fixes.
304 (decrement_count): Formatting fix.
305
306 2020-03-11 Delia Burduv <delia.burduv@arm.com>
307
308 * config/arm/arm-builtins.c
309 (arm_init_simd_builtin_scalar_types): New.
310 * config/arm/arm_neon.h (vld2_bf16): Used new builtin type.
311 (vld2q_bf16): Used new builtin type.
312 (vld3_bf16): Used new builtin type.
313 (vld3q_bf16): Used new builtin type.
314 (vld4_bf16): Used new builtin type.
315 (vld4q_bf16): Used new builtin type.
316 (vld2_dup_bf16): Used new builtin type.
317 (vld2q_dup_bf16): Used new builtin type.
318 (vld3_dup_bf16): Used new builtin type.
319 (vld3q_dup_bf16): Used new builtin type.
320 (vld4_dup_bf16): Used new builtin type.
321 (vld4q_dup_bf16): Used new builtin type.
322
323 2020-03-11 Jakub Jelinek <jakub@redhat.com>
324
325 PR target/94134
326 * config/pdp11/pdp11.c (pdp11_asm_output_var): Call switch_to_section
327 at the start to switch to data section. Don't print extra newline if
328 .globl directive has not been emitted.
329
330 2020-03-11 Richard Biener <rguenther@suse.de>
331
332 * match.pd ((T *)(ptr - ptr-cst) -> &MEM[ptr + -ptr-cst]):
333 New pattern.
334
335 2020-03-11 Eric Botcazou <ebotcazou@adacore.com>
336
337 PR middle-end/93961
338 * tree.c (variably_modified_type_p) <RECORD_TYPE>: Recurse into fields
339 whose type is a qualified union.
340
341 2020-03-11 Jakub Jelinek <jakub@redhat.com>
342
343 PR target/94121
344 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use absu_hwi
345 instead of abs_hwi, change moffset type to unsigned HOST_WIDE_INT.
346
347 PR bootstrap/93962
348 * value-prof.c (dump_histogram_value): Use abs_hwi instead of
349 std::abs.
350 (get_nth_most_common_value): Use abs_hwi instead of abs.
351
352 PR middle-end/94111
353 * dfp.c (decimal_to_binary): Only use decimal128ToString if from->cl
354 is rvc_normal, otherwise use real_to_decimal to print the number to
355 string.
356
357 PR tree-optimization/94114
358 * tree-loop-distribution.c (generate_memset_builtin): Call
359 rewrite_to_non_trapping_overflow even on mem.
360 (generate_memcpy_builtin): Call rewrite_to_non_trapping_overflow even
361 on dest and src.
362
363 2020-03-10 Jeff Law <law@redhat.com>
364
365 * config/bfin/bfin.md (movsi_insv): Add length attribute.
366
367 2020-03-10 Jiufu Guo <guojiufu@linux.ibm.com>
368
369 PR target/93709
370 * gcc/config/rs6000/rs6000.c (rs6000_emit_p9_fp_minmax): Check
371 NAN and SIGNED_ZEROR for smax/smin.
372
373 2020-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
374
375 PR target/90763
376 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
377 clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
378
379 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
380
381 * loop-iv.c (find_simple_exit): Make it static.
382 * cfgloop.h: Remove the corresponding prototype.
383
384 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
385
386 * ddg.c (create_ddg): Fix intendation.
387 (set_recurrence_length): Likewise.
388 (create_ddg_all_sccs): Likewise.
389
390 2020-03-10 Jakub Jelinek <jakub@redhat.com>
391
392 PR target/94088
393 * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
394 CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
395 is 32.
396
397 2020-03-09 Jason Merrill <jason@redhat.com>
398
399 * gdbinit.in (pgs): Fix typo in documentation.
400
401 2020-03-09 Vladimir Makarov <vmakarov@redhat.com>
402
403 Revert:
404
405 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
406
407 PR rtl-optimization/93564
408 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
409 do not honor reg alloc order.
410
411 2020-03-09 Andrew Pinski <apinski@marvell.com>
412
413 PR inline-asm/94095
414 * doc/extend.texi (x86 Operand Modifiers): Fix column
415 for 'A' modifier.
416
417 2020-03-09 Martin Liska <mliska@suse.cz>
418
419 PR target/93800
420 * config/rs6000/rs6000.c (rs6000_option_override_internal):
421 Remove set of str_align_loops and str_align_jumps as these
422 should be set in previous 2 conditions in the function.
423
424 2020-03-09 Jakub Jelinek <jakub@redhat.com>
425
426 PR rtl-optimization/94045
427 * params.opt (-param=max-find-base-term-values=): New option.
428 * alias.c (find_base_term): Add cut-off for number of visited VALUEs
429 in a single toplevel find_base_term call.
430
431 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
432
433 PR target/91598
434 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
435 * config/aarch64/aarch64-simd.md
436 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
437 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
438 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
439 * config/aarch64/arm_neon.h:
440 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
441 (vmlal_lane_u16): Likewise.
442 (vmlal_lane_s32): Likewise.
443 (vmlal_lane_u32): Likewise.
444 (vmlal_laneq_s16): Likewise.
445 (vmlal_laneq_u16): Likewise.
446 (vmlal_laneq_s32): Likewise.
447 (vmlal_laneq_u32): Likewise.
448 (vmull_lane_s16): Likewise.
449 (vmull_lane_u16): Likewise.
450 (vmull_lane_s32): Likewise.
451 (vmull_lane_u32): Likewise.
452 (vmull_laneq_s16): Likewise.
453 (vmull_laneq_u16): Likewise.
454 (vmull_laneq_s32): Likewise.
455 (vmull_laneq_u32): Likewise.
456 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
457 (Qlane): Likewise.
458
459 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
460
461 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
462 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
463 (aarch64_mls_elt<mode>): Likewise.
464 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
465 (aarch64_fma4_elt<mode>): Likewise.
466 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
467 (aarch64_fma4_elt_to_64v2df): Likewise.
468 (aarch64_fnma4_elt<mode>): Likewise.
469 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
470 (aarch64_fnma4_elt_to_64v2df): Likewise.
471
472 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
473
474 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
475 Specify movprfx attribute.
476 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
477
478 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
479
480 PR target/94065
481 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
482 cmodel=large.
483 (TARGET_NO_FP_IN_TOC): Same.
484 * config/rs6000/aix71.h: Same.
485 * config/rs6000/aix72.h: Same.
486
487 2020-03-06 Andrew Pinski <apinski@marvell.com>
488 Jeff Law <law@redhat.com>
489
490 PR rtl-optimization/93996
491 * haifa-sched.c (remove_notes): Be more careful when adding
492 REG_SAVE_NOTE.
493
494 2020-03-06 Delia Burduv <delia.burduv@arm.com>
495
496 * config/arm/arm_neon.h (vld2_bf16): New.
497 (vld2q_bf16): New.
498 (vld3_bf16): New.
499 (vld3q_bf16): New.
500 (vld4_bf16): New.
501 (vld4q_bf16): New.
502 (vld2_dup_bf16): New.
503 (vld2q_dup_bf16): New.
504 (vld3_dup_bf16): New.
505 (vld3q_dup_bf16): New.
506 (vld4_dup_bf16): New.
507 (vld4q_dup_bf16): New.
508 * config/arm/arm_neon_builtins.def
509 (vld2): Changed to VAR13 and added v4bf, v8bf
510 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
511 (vld3): Changed to VAR13 and added v4bf, v8bf
512 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
513 (vld4): Changed to VAR13 and added v4bf, v8bf
514 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
515 * config/arm/iterators.md (VDXBF2): New iterator.
516 *config/arm/neon.md (neon_vld2): Use new iterators.
517 (neon_vld2_dup<mode): Use new iterators.
518 (neon_vld3<mode>): Likewise.
519 (neon_vld3qa<mode>): Likewise.
520 (neon_vld3qb<mode>): Likewise.
521 (neon_vld3_dup<mode>): Likewise.
522 (neon_vld4<mode>): Likewise.
523 (neon_vld4qa<mode>): Likewise.
524 (neon_vld4qb<mode>): Likewise.
525 (neon_vld4_dup<mode>): Likewise.
526 (neon_vld2_dupv8bf): New.
527 (neon_vld3_dupv8bf): Likewise.
528 (neon_vld4_dupv8bf): Likewise.
529
530 2020-03-06 Delia Burduv <delia.burduv@arm.com>
531
532 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
533 (bfloat16x8x2_t): New typedef.
534 (bfloat16x4x3_t): New typedef.
535 (bfloat16x8x3_t): New typedef.
536 (bfloat16x4x4_t): New typedef.
537 (bfloat16x8x4_t): New typedef.
538 (vst2_bf16): New.
539 (vst2q_bf16): New.
540 (vst3_bf16): New.
541 (vst3q_bf16): New.
542 (vst4_bf16): New.
543 (vst4q_bf16): New.
544 * config/arm/arm-builtins.c (v2bf_UP): Define.
545 (VAR13): New.
546 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
547 * config/arm/arm-modes.def (V2BF): New mode.
548 * config/arm/arm-simd-builtin-types.def
549 (Bfloat16x2_t): New entry.
550 * config/arm/arm_neon_builtins.def
551 (vst2): Changed to VAR13 and added v4bf, v8bf
552 (vst3): Changed to VAR13 and added v4bf, v8bf
553 (vst4): Changed to VAR13 and added v4bf, v8bf
554 * config/arm/iterators.md (VDXBF): New iterator.
555 (VQ2BF): New iterator.
556 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
557 (neon_vst2<mode>): Used new iterators.
558 (neon_vst3<mode>): Used new iterators.
559 (neon_vst3<mode>): Used new iterators.
560 (neon_vst3qa<mode>): Used new iterators.
561 (neon_vst3qb<mode>): Used new iterators.
562 (neon_vst4<mode>): Used new iterators.
563 (neon_vst4<mode>): Used new iterators.
564 (neon_vst4qa<mode>): Used new iterators.
565 (neon_vst4qb<mode>): Used new iterators.
566
567 2020-03-06 Delia Burduv <delia.burduv@arm.com>
568
569 * config/aarch64/aarch64-simd-builtins.def
570 (bfcvtn): New built-in function.
571 (bfcvtn_q): New built-in function.
572 (bfcvtn2): New built-in function.
573 (bfcvt): New built-in function.
574 * config/aarch64/aarch64-simd.md
575 (aarch64_bfcvtn<q><mode>): New pattern.
576 (aarch64_bfcvtn2v8bf): New pattern.
577 (aarch64_bfcvtbf): New pattern.
578 * config/aarch64/arm_bf16.h (float32_t): New typedef.
579 (vcvth_bf16_f32): New intrinsic.
580 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
581 (vcvtq_low_bf16_f32): New intrinsic.
582 (vcvtq_high_bf16_f32): New intrinsic.
583 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
584 (UNSPEC_BFCVTN): New UNSPEC.
585 (UNSPEC_BFCVTN2): New UNSPEC.
586 (UNSPEC_BFCVT): New UNSPEC.
587 * config/arm/types.md (bf_cvt): New type.
588
589 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
590
591 * config/s390/s390.md ("tabort"): Get rid of two consecutive
592 blanks in format string.
593
594 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
595
596 PR target/89229
597 PR target/89346
598 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
599 * config/i386/i386.c (ix86_get_ssemov): New function.
600 (ix86_output_ssemov): Likewise.
601 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
602 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
603 check.
604 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
605 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
606 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
607 (*movti_internal): Likewise.
608 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
609
610 2020-03-05 Jeff Law <law@redhat.com>
611
612 PR tree-optimization/91890
613 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
614 Use gimple_or_expr_nonartificial_location.
615 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
616 Use gimple_or_expr_nonartificial_location.
617 * gimple.c (gimple_or_expr_nonartificial_location): New function.
618 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
619 * tree-ssa-strlen.c (maybe_warn_overflow): Use
620 gimple_or_expr_nonartificial_location.
621 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
622 (maybe_warn_pointless_strcmp): Likewise.
623
624 2020-03-05 Jakub Jelinek <jakub@redhat.com>
625
626 PR target/94046
627 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
628 SRC and MASK arguments to __m128 from __m128d.
629 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
630 from __m256d.
631 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
632 from __m128d.
633 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
634 argument to __m128i from __m128d.
635 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
636 __m256d.
637 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
638 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
639 __m256.
640
641 2020-03-05 Delia Burduv <delia.burduv@arm.com>
642
643 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
644 (vbfmlalbq_f32): New.
645 (vbfmlaltq_f32): New.
646 (vbfmlalbq_lane_f32): New.
647 (vbfmlaltq_lane_f32): New.
648 (vbfmlalbq_laneq_f32): New.
649 (vbfmlaltq_laneq_f32): New.
650 * config/arm/arm_neon_builtins.def (vmmla): New.
651 (vfmab): New.
652 (vfmat): New.
653 (vfmab_lane): New.
654 (vfmat_lane): New.
655 (vfmab_laneq): New.
656 (vfmat_laneq): New.
657 * config/arm/iterators.md (BF_MA): New int iterator.
658 (bt): New int attribute.
659 (VQXBF): Copy of VQX with V8BF.
660 * config/arm/neon.md (neon_vmmlav8bf): New insn.
661 (neon_vfma<bt>v8bf): New insn.
662 (neon_vfma<bt>_lanev8bf): New insn.
663 (neon_vfma<bt>_laneqv8bf): New expand.
664 (neon_vget_high<mode>): Changed iterator to VQXBF.
665 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
666 (UNSPEC_BFMAB): New UNSPEC.
667 (UNSPEC_BFMAT): New UNSPEC.
668
669 2020-03-05 Jakub Jelinek <jakub@redhat.com>
670
671 PR middle-end/93399
672 * tree-pretty-print.h (pretty_print_string): Declare.
673 * tree-pretty-print.c (pretty_print_string): Remove forward
674 declaration, no longer static. Change nbytes parameter type
675 from unsigned to size_t.
676 * print-rtl.c (print_value) <case CONST_STRING>: Use
677 pretty_print_string and for shrink way too long strings.
678
679 2020-03-05 Richard Biener <rguenther@suse.de>
680 Jakub Jelinek <jakub@redhat.com>
681
682 PR tree-optimization/93582
683 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
684 last operand as signed when looking for memset offset. Formatting
685 fix.
686
687 2020-03-04 Andrew Pinski <apinski@marvell.com>
688
689 PR bootstrap/93962
690 * value-prof.c (dump_histogram_value): Use std::abs.
691
692 2020-03-04 Martin Sebor <msebor@redhat.com>
693
694 PR tree-optimization/93986
695 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
696 operands to the same precision widest_int to avoid ICEs.
697
698 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
699
700 PR target/87560
701 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
702 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
703 for OPTION_MASK_ALTIVEC.
704
705 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
706
707 * config.gcc: Include the glibc-stdint.h header for zTPF.
708
709 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
710
711 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
712 direct FPR-GPR copies.
713 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
714 FPRs.
715
716 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
717
718 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
719 operands to the prologue_tpf expander.
720 (s390_emit_epilogue): Likewise.
721 (s390_option_override_internal): Do error checking and setup for
722 the new options.
723 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
724 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
725 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
726 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
727 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
728 operands for the check flag and the branch target.
729 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
730 ("mtpf-trace-hook-prologue-target")
731 ("mtpf-trace-hook-epilogue-check")
732 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
733 options.
734 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
735 options are for debugging purposes and will not be documented
736 here.
737
738 2020-03-04 Jakub Jelinek <jakub@redhat.com>
739
740 PR debug/93888
741 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
742
743 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
744 argument. Change pd argument so that it can be modified. Turn
745 constant non-CONSTRUCTOR store into non-constant if it is too large.
746 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
747 overflows.
748 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
749 callers.
750
751 2020-02-04 Richard Biener <rguenther@suse.de>
752
753 PR tree-optimization/93964
754 * graphite-isl-ast-to-gimple.c
755 (gcc_expression_from_isl_ast_expr_id): Add intermediate
756 conversion for pointer to integer converts.
757 * graphite-scop-detection.c (assign_parameter_index_in_region):
758 Relax assert.
759
760 2020-03-04 Martin Liska <mliska@suse.cz>
761
762 PR c/93886
763 PR c/93887
764 * doc/invoke.texi: Clarify --help=language and --help=common
765 interaction.
766
767 2020-03-04 Jakub Jelinek <jakub@redhat.com>
768
769 PR tree-optimization/94001
770 * tree-tailcall.c (process_assignment): Before comparing op1 to
771 *ass_var, verify *ass_var is non-NULL.
772
773 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
774
775 PR target/93995
776 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
777 the result of IOR.
778
779 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
780
781 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
782 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
783 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
784 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
785 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
786 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
787 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
788 (V_bf_low, V_bf_cvt_m): New mode attributes.
789 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
790 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
791 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
792 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
793 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
794
795 2020-03-03 Jakub Jelinek <jakub@redhat.com>
796
797 PR tree-optimization/93582
798 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
799 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
800 members, initialize them in the constructor and if mask is non-NULL,
801 artificially push_partial_def {} for the portions of the mask that
802 contain zeros.
803 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
804 val and return (void *)-1. Formatting fix.
805 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
806 Formatting fix.
807 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
808 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
809 data.mask_result.
810 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
811 mask.
812 (visit_stmt): Formatting fix.
813
814 2020-03-03 Richard Biener <rguenther@suse.de>
815
816 PR tree-optimization/93946
817 * alias.h (refs_same_for_tbaa_p): Declare.
818 * alias.c (refs_same_for_tbaa_p): New function.
819 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
820 zero.
821 * tree-ssa-scopedtables.h
822 (avail_exprs_stack::lookup_avail_expr): Add output argument
823 giving access to the hashtable entry.
824 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
825 Likewise.
826 * tree-ssa-dom.c: Include alias.h.
827 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
828 removing redundant store.
829 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
830 (ao_ref_init_from_vn_reference): Adjust prototype.
831 (vn_reference_lookup_pieces): Likewise.
832 (vn_reference_insert_pieces): Likewise.
833 * tree-ssa-sccvn.c: Track base alias set in addition to alias
834 set everywhere.
835 (eliminate_dom_walker::eliminate_stmt): Also check base alias
836 set when removing redundant stores.
837 (visit_reference_op_store): Likewise.
838 * dse.c (record_store): Adjust valdity check for redundant
839 store removal.
840
841 2020-03-03 Jakub Jelinek <jakub@redhat.com>
842
843 PR target/26877
844 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
845
846 PR rtl-optimization/94002
847 * explow.c (plus_constant): Punt if cst has VOIDmode and
848 get_pool_mode is different from mode.
849
850 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
851
852 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
853 address has an offset which fits the scalling constraint for a
854 load/store operation.
855 (legitimate_scaled_address_p): Update use
856 leigitimate_small_data_address_p.
857 (arc_print_operand): Likewise.
858 (arc_legitimate_address_p): Likewise.
859 (legitimate_small_data_address_p): Likewise.
860
861 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
862
863 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
864 (fnmasf4_fpu): Likewise.
865
866 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
867
868 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
869 32bit ops.
870 (subdi3): Likewise.
871 (adddi3_i): Remove pattern.
872 (subdi3_i): Likewise.
873
874 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
875
876 * config/arc/arc.md (eh_return): Add length info.
877
878 2020-03-02 David Malcolm <dmalcolm@redhat.com>
879
880 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
881
882 2020-03-02 David Malcolm <dmalcolm@redhat.com>
883
884 * doc/invoke.texi (Static Analyzer Options): Add
885 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
886 by -fanalyzer.
887
888 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
889
890 PR target/93997
891 * config/i386/i386.md (movstrict<mode>): Allow only
892 registers with VALID_INT_MODE_P modes.
893
894 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
895
896 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
897 (reduc_insn): Use 'U' and 'B' operand codes.
898 (reduc_<reduc_op>_scal_<mode>): Allow all types.
899 (reduc_<reduc_op>_scal_v64di): Delete.
900 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
901 (*plus_carry_dpp_shr_v64si): Change to ...
902 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
903 (mov_from_lane63_v64di): Change to ...
904 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
905 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
906 Support UNSPEC_MOV_DPP_SHR output formats.
907 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
908 Add "use_extends" reductions.
909 (print_operand_address): Add 'I' and 'U' codes.
910 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
911
912 2020-03-02 Martin Liska <mliska@suse.cz>
913
914 * lto-wrapper.c: Fix typo in comment about
915 C++ standard version.
916
917 2020-03-01 Martin Sebor <msebor@redhat.com>
918
919 PR c++/92721
920 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
921
922 2020-03-01 Martin Sebor <msebor@redhat.com>
923
924 PR middle-end/93829
925 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
926 of a pointer in the outermost ADDR_EXPRs.
927
928 2020-02-28 Jeff Law <law@redhat.com>
929
930 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
931 * config/v850/v850.c (v850_asm_trampoline_template): Update
932 accordingly.
933
934 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
935
936 PR target/93937
937 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
938 Delete insn.
939
940 2020-02-28 Martin Liska <mliska@suse.cz>
941
942 PR other/93965
943 * configure.ac: Improve detection of ld_date by requiring
944 either two dashes or none.
945 * configure: Regenerate.
946
947 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
948
949 PR rtl-optimization/93564
950 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
951 do not honor reg alloc order.
952
953 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
954
955 PR target/87612
956 * config/aarch64/aarch64.c (aarch64_override_options): Fix
957 misleading warning string.
958
959 2020-02-27 Martin Sebor <msebor@redhat.com>
960
961 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
962
963 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
964
965 PR target/93932
966 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
967 Split the insn into two parts. This insn only does variable
968 extract from a register.
969 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
970 variable extract from memory.
971 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
972 only does variable extract from a register.
973 (vsx_extract_v4sf_var_load): New insn, do variable extract from
974 memory.
975 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
976 into two parts. This insn only does variable extract from a
977 register.
978 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
979 do variable extract from memory.
980
981 2020-02-27 Martin Jambor <mjambor@suse.cz>
982 Feng Xue <fxue@os.amperecomputing.com>
983
984 PR ipa/93707
985 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
986 new function calls_same_node_or_its_all_contexts_clone_p.
987 (cgraph_edge_brings_value_p): Use it.
988 (cgraph_edge_brings_value_p): Likewise.
989 (self_recursive_pass_through_p): Return false if caller is a clone.
990 (self_recursive_agg_pass_through_p): Likewise.
991
992 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
993
994 PR middle-end/92152
995 * alias.c (ends_tbaa_access_path_p): Break out from ...
996 (component_uses_parent_alias_set_from): ... here.
997 * alias.h (ends_tbaa_access_path_p): Declare.
998 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
999 handle trailing arrays past end of tbaa access path.
1000 (aliasing_component_refs_p): ... here; likewise.
1001 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
1002 path; disambiguate also past end of it.
1003 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
1004 path.
1005
1006 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
1007
1008 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
1009 beginning of the file.
1010 (vcreate_bf16, vcombine_bf16): New.
1011 (vdup_n_bf16, vdupq_n_bf16): New.
1012 (vdup_lane_bf16, vdup_laneq_bf16): New.
1013 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1014 (vduph_lane_bf16, vduph_laneq_bf16): New.
1015 (vset_lane_bf16, vsetq_lane_bf16): New.
1016 (vget_lane_bf16, vgetq_lane_bf16): New.
1017 (vget_high_bf16, vget_low_bf16): New.
1018 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1019 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1020 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1021 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1022 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1023 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1024 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1025 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1026 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1027 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1028 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
1029 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1030 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1031 (vreinterpretq_bf16_p128): New.
1032 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1033 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1034 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1035 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1036 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1037 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1038 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1039 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1040 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1041 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1042 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1043 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1044 (vreinterpretq_p128_bf16): New.
1045 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
1046 (V_elem): Likewise.
1047 (V_elem_l): Likewise.
1048 (VD_LANE): Likewise.
1049 (VQX) Add V8BF.
1050 (V_DOUBLE): Likewise.
1051 (VDQX): Add V4BF and V8BF.
1052 (V_two_elem, V_three_elem, V_four_elem): Likewise.
1053 (V_reg): Likewise.
1054 (V_HALF): Likewise.
1055 (V_double_vector_mode): Likewise.
1056 (V_cmp_result): Likewise.
1057 (V_uf_sclr): Likewise.
1058 (V_sz_elem): Likewise.
1059 (Is_d_reg): Likewise.
1060 (V_mode_nunits): Likewise.
1061 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
1062
1063 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
1064
1065 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
1066 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
1067 (<expander><mode>3<exec>): Likewise.
1068 (<expander><mode>3): New.
1069 (v<expander><mode>3): New.
1070 (<expander><mode>3): New.
1071 (<expander><mode>3<exec>): Rename to ...
1072 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
1073 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
1074
1075 2020-02-27 Alexandre Oliva <oliva@adacore.com>
1076
1077 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
1078 them alone on vx7.
1079
1080 2020-02-27 Richard Biener <rguenther@suse.de>
1081
1082 PR tree-optimization/93508
1083 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
1084 non-_CHK variants. Valueize their length arguments.
1085
1086 2020-02-27 Richard Biener <rguenther@suse.de>
1087
1088 PR tree-optimization/93953
1089 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
1090 to the hash-map entry.
1091
1092 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
1093
1094 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
1095
1096 2020-02-27 Mark Williams <mwilliams@fb.com>
1097
1098 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
1099 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
1100 -ffile-prefix-map and -fmacro-prefix-map.
1101 * lto-streamer-out.c: Include file-prefix-map.h.
1102 (lto_output_location): Remap the file part of locations.
1103
1104 2020-02-27 Jakub Jelinek <jakub@redhat.com>
1105
1106 PR c/93949
1107 * gimplify.c (gimplify_init_constructor): Don't promote readonly
1108 DECL_REGISTER variables to TREE_STATIC.
1109
1110 PR tree-optimization/93582
1111 PR tree-optimization/93945
1112 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
1113 non-zero INTEGER_CST second argument and ref->offset or ref->size
1114 not a multiple of BITS_PER_UNIT.
1115
1116 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
1117
1118 * doc/install.texi (Binaries): Update description of BullFreeware.
1119
1120 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
1121
1122 PR c++/90467
1123
1124 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
1125 C++ Language Options, Warning Options, and Static Analyzer
1126 Options lists. Document negative form of options enabled by
1127 default. Move some things around to more accurately sort
1128 warnings by category.
1129 (C++ Dialect Options, Warning Options, Static Analyzer
1130 Options): Document negative form of options when enabled by
1131 default. Move some things around to more accurately sort
1132 warnings by category. Add some missing index entries.
1133 Light copy-editing.
1134
1135 2020-02-26 Carl Love <cel@us.ibm.com>
1136
1137 PR target/91276
1138 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
1139 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
1140 for the vector unsigned short arguments. It is also listed as the
1141 name of the built-in for arguments vector unsigned short,
1142 vector unsigned int and vector unsigned long long built-ins. The
1143 name of the builtins for these arguments should be:
1144 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
1145 __builtin_crypto_vpmsumd respectively.
1146
1147 2020-02-26 Richard Biener <rguenther@suse.de>
1148
1149 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
1150 and load permutation.
1151
1152 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
1153
1154 PR middle-end/93843
1155 * optabs-tree.c (supportable_convert_operation): Reject types with
1156 scalar modes.
1157
1158 2020-02-26 David Malcolm <dmalcolm@redhat.com>
1159
1160 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
1161
1162 2020-02-26 Jakub Jelinek <jakub@redhat.com>
1163
1164 PR tree-optimization/93820
1165 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
1166 argument to ALL_INTEGER_CST_P boolean.
1167 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
1168 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
1169 adjacent INTEGER_CST store into merged_store->only_constants like
1170 overlapping one.
1171
1172 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1173
1174 PR other/93912
1175 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
1176 -> probability.
1177 * cfghooks.c (verify_flow_info): Likewise.
1178 * predict.c (combine_predictions_for_bb): Likewise.
1179 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
1180 sucessor -> successor.
1181 (find_traces_1_round): Fix comment typo, destinarion -> destination.
1182 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
1183 successors.
1184 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
1185 message typo, sucessors -> successors.
1186
1187 2020-02-25 Martin Sebor <msebor@redhat.com>
1188
1189 * doc/extend.texi (attribute access): Correct an example.
1190
1191 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1192
1193 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
1194 Add simd_bf.
1195 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
1196 (VAR15, VAR16): New.
1197 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
1198 (VD): Enable for V4BF.
1199 (VDC): Likewise.
1200 (VQ): Enable for V8BF.
1201 (VQ2): Likewise.
1202 (VQ_NO2E): Likewise.
1203 (VDBL, Vdbl): Add V4BF.
1204 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
1205 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
1206 (bfloat16x8x2_t): Likewise.
1207 (bfloat16x4x3_t): Likewise.
1208 (bfloat16x8x3_t): Likewise.
1209 (bfloat16x4x4_t): Likewise.
1210 (bfloat16x8x4_t): Likewise.
1211 (vcombine_bf16): New.
1212 (vld1_bf16, vld1_bf16_x2): New.
1213 (vld1_bf16_x3, vld1_bf16_x4): New.
1214 (vld1q_bf16, vld1q_bf16_x2): New.
1215 (vld1q_bf16_x3, vld1q_bf16_x4): New.
1216 (vld1_lane_bf16): New.
1217 (vld1q_lane_bf16): New.
1218 (vld1_dup_bf16): New.
1219 (vld1q_dup_bf16): New.
1220 (vld2_bf16): New.
1221 (vld2q_bf16): New.
1222 (vld2_dup_bf16): New.
1223 (vld2q_dup_bf16): New.
1224 (vld3_bf16): New.
1225 (vld3q_bf16): New.
1226 (vld3_dup_bf16): New.
1227 (vld3q_dup_bf16): New.
1228 (vld4_bf16): New.
1229 (vld4q_bf16): New.
1230 (vld4_dup_bf16): New.
1231 (vld4q_dup_bf16): New.
1232 (vst1_bf16, vst1_bf16_x2): New.
1233 (vst1_bf16_x3, vst1_bf16_x4): New.
1234 (vst1q_bf16, vst1q_bf16_x2): New.
1235 (vst1q_bf16_x3, vst1q_bf16_x4): New.
1236 (vst1_lane_bf16): New.
1237 (vst1q_lane_bf16): New.
1238 (vst2_bf16): New.
1239 (vst2q_bf16): New.
1240 (vst3_bf16): New.
1241 (vst3q_bf16): New.
1242 (vst4_bf16): New.
1243 (vst4q_bf16): New.
1244
1245 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1246
1247 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
1248 (VALL_F16): Likewise.
1249 (VALLDI_F16): Likewise.
1250 (Vtype): Likewise.
1251 (Vetype): Likewise.
1252 (vswap_width_name): Likewise.
1253 (VSWAP_WIDTH): Likewise.
1254 (Vel): Likewise.
1255 (VEL): Likewise.
1256 (q): Likewise.
1257 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
1258 (vget_lane_bf16, vgetq_lane_bf16): New.
1259 (vcreate_bf16): New.
1260 (vdup_n_bf16, vdupq_n_bf16): New.
1261 (vdup_lane_bf16, vdup_laneq_bf16): New.
1262 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1263 (vduph_lane_bf16, vduph_laneq_bf16): New.
1264 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1265 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1266 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1267 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1268 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1269 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1270 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1271 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1272 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1273 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1274 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
1275 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
1276 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1277 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1278 (vreinterpretq_bf16_p128): New.
1279 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1280 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1281 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1282 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1283 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1284 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1285 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1286 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1287 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1288 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1289 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1290 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1291 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
1292 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
1293 (vreinterpretq_p128_bf16): New.
1294
1295 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
1296
1297 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
1298 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
1299 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
1300 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
1301 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
1302 * config/arm/iterators.md (VSF2BF): New attribute.
1303 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
1304 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
1305 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
1306
1307 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
1308
1309 * config/arm/arm.md (required_for_purecode): New attribute.
1310 (enabled): Handle required_for_purecode.
1311 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
1312 work with -mpure-code.
1313
1314 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1315
1316 PR rtl-optimization/93908
1317 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
1318 with mask.
1319
1320 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
1321
1322 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
1323
1324 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
1325
1326 * doc/install.texi (--enable-checking): Adjust wording.
1327
1328 2020-02-25 Richard Biener <rguenther@suse.de>
1329
1330 PR tree-optimization/93868
1331 * tree-vect-slp.c (slp_copy_subtree): New function.
1332 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
1333 re-arranging stmts in it.
1334
1335 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1336
1337 PR middle-end/93874
1338 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
1339 dummy function and remove it at the end.
1340
1341 PR translation/93864
1342 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
1343 paramter -> parameter.
1344 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
1345 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
1346
1347 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
1348
1349 * doc/install.texi (--enable-checking): Properly document current
1350 behavior.
1351 (--enable-stage1-checking): Minor clarification about bootstrap.
1352
1353 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1354
1355 PR analyzer/93032
1356 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
1357 -fanalyzer-checker=taint is also required.
1358 (-fanalyzer-checker=): Note that providing this option enables the
1359 given checker, and doing so may be required for checkers that are
1360 disabled by default.
1361
1362 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1363
1364 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
1365 significant control flow events; add a "3" which shows all
1366 control flow events; the old "3" becomes "4".
1367
1368 2020-02-24 Jakub Jelinek <jakub@redhat.com>
1369
1370 PR tree-optimization/93582
1371 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1372 pd.offset and pd.size to be counted in bits rather than bytes, add
1373 support for maxsizei that is not a multiple of BITS_PER_UNIT and
1374 handle bitfield stores and loads.
1375 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1376 uncomparable quantities - bytes vs. bits. Allow push_partial_def
1377 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1378 pd.offset/pd.size to be counted in bits rather than bytes.
1379 Formatting fix. Rename shadowed len variable to buflen.
1380
1381 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1382 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
1383
1384 PR driver/47785
1385 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1386 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1387 * opts-common.c (parse_options_from_collect_gcc_options): New function.
1388 (prepend_xassembler_to_collect_as_options): Likewise.
1389 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1390 (prepend_xassembler_to_collect_as_options): Likewise.
1391 * lto-opts.c (lto_write_options): Stream assembler options
1392 in COLLECT_AS_OPTIONS.
1393 * lto-wrapper.c (xassembler_options_error): New static variable.
1394 (get_options_from_collect_gcc_options): Move parsing options code to
1395 parse_options_from_collect_gcc_options and call it.
1396 (merge_and_complain): Validate -Xassembler options.
1397 (append_compiler_options): Handle OPT_Xassembler.
1398 (run_gcc): Append command line -Xassembler options to
1399 collect_gcc_options.
1400 * doc/invoke.texi: Add documentation about using Xassembler
1401 options with LTO.
1402
1403 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
1404
1405 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1406 for LTGT.
1407 (riscv_rtx_costs): Update cost model for LTGT.
1408
1409 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
1410
1411 PR rtl-optimization/93564
1412 * ira-color.c (struct update_cost_queue_elem): New member start.
1413 (queue_update_cost, get_next_update_cost): Add new arg start.
1414 (allocnos_conflict_p): New function.
1415 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1416 Add checking conflicts with allocnos_conflict_p.
1417 (update_costs_from_prefs, restore_costs_from_copies): Adjust
1418 update_costs_from_allocno calls.
1419 (update_conflict_hard_regno_costs): Add checking conflicts with
1420 allocnos_conflict_p. Adjust calls of queue_update_cost and
1421 get_next_update_cost.
1422 (assign_hard_reg): Adjust calls of queue_update_cost. Add
1423 debugging print.
1424 (bucket_allocno_compare_func): Restore previous version.
1425
1426 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
1427
1428 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1429 double-word size when handling aggregate return values.
1430 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1431 that homogeneous SFmode and DFmode aggregates are passed and returned
1432 in general registers.
1433
1434 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1435
1436 PR translation/93759
1437 * opts.c (print_filtered_help): Translate help before appending
1438 messages to it rather than after that.
1439
1440 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1441
1442 PR rtl-optimization/PR92989
1443 * lra-lives.c (process_bb_lives): Restore the original order
1444 of the bb liveness update. Call make_hard_regno_dead for each
1445 register clobbered at the start of an EH receiver.
1446
1447 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1448
1449 PR ipa/93763
1450 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1451 self-recursively generated.
1452
1453 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1454
1455 PR target/93860
1456 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1457 error string.
1458
1459 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1460
1461 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1462 Document new target supports option.
1463
1464 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1465
1466 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1467 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1468 * config/arm/iterators.md (MATMUL): New iterator.
1469 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1470 (mmla_sfx): New attribute.
1471 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1472 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1473 (UNSPEC_MATMUL_US): New.
1474
1475 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1476
1477 * config/arm/arm.md: Prevent scalar shifts from being used when big
1478 endian is enabled.
1479
1480 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1481 Richard Biener <rguenther@suse.de>
1482
1483 PR tree-optimization/93586
1484 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1485 after mismatched array refs; do not sure type size information to
1486 recover from unmatched referneces with !flag_strict_aliasing_p.
1487
1488 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1489
1490 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1491 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1492 (scatter_store<mode>): Rename to ...
1493 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1494 (scatter<mode>_exec): Delete. Move contents ...
1495 (mask_scatter_store<mode>): ... here, and rename that to ...
1496 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1497 Remove mode conversion.
1498 (mask_gather_load<mode>): Rename to ...
1499 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1500 Remove mode conversion.
1501 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1502
1503 2020-02-21 Martin Jambor <mjambor@suse.cz>
1504
1505 PR tree-optimization/93845
1506 * tree-sra.c (verify_sra_access_forest): Only test access size of
1507 scalar types.
1508
1509 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1510
1511 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1512 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1513 (addv64di3_exec): Likewise.
1514 (subv64di3): Likewise.
1515 (subv64di3_exec): Likewise.
1516 (addv64di3_zext): Likewise.
1517 (addv64di3_zext_exec): Likewise.
1518 (addv64di3_zext_dup): Likewise.
1519 (addv64di3_zext_dup_exec): Likewise.
1520 (addv64di3_zext_dup2): Likewise.
1521 (addv64di3_zext_dup2_exec): Likewise.
1522 (addv64di3_sext_dup2): Likewise.
1523 (addv64di3_sext_dup2_exec): Likewise.
1524 (<expander>v64di3): Likewise.
1525 (<expander>v64di3_exec): Likewise.
1526 (*<reduc_op>_dpp_shr_v64di): Likewise.
1527 (*plus_carry_dpp_shr_v64di): Likewise.
1528 * config/gcn/gcn.md (adddi3): Likewise.
1529 (addptrdi3): Likewise.
1530 (<expander>di3): Likewise.
1531
1532 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1533
1534 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1535
1536 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1537
1538 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1539 support. Use aarch64_emit_mult instead of emitting multiplication
1540 instructions directly.
1541 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1542 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1543
1544 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1545
1546 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1547 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1548 instead of emitting multiplication instructions directly.
1549 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1550 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1551 (@aarch64_frecps<mode>): New expanders.
1552
1553 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1554
1555 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1556 on and produce uint64_ts rather than ints.
1557 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1558 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1559
1560 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1561
1562 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1563 an unused xmsk register when handling approximate rsqrt.
1564
1565 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1566
1567 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1568 flag_finite_math_only condition.
1569
1570 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1571
1572 PR target/93828
1573 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1574 to destination operand for shufps alternative.
1575 (*vec_extractv2si_1): Ditto.
1576
1577 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1578
1579 PR target/93658
1580 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1581 vector modes.
1582
1583 2020-02-20 Martin Liska <mliska@suse.cz>
1584
1585 PR translation/93831
1586 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1587
1588 2020-02-20 Martin Liska <mliska@suse.cz>
1589
1590 PR translation/93830
1591 * common/config/avr/avr-common.c: Remote trailing "|".
1592
1593 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1594
1595 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1596 comment.
1597
1598 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1599
1600 PR tree-optimization/93767
1601 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1602 access-size bias from the offset calculations for negative strides.
1603
1604 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1605
1606 * collect2.c (c_file, o_file): Make const again.
1607 (ldout,lderrout, dump_ld_file): Remove.
1608 (tool_cleanup): Avoid calling not signal-safe functions.
1609 (maybe_run_lto_and_relink): Avoid possible signal handler
1610 access to unintialzed memory (lto_o_files).
1611 (main): Avoid leaking temp files in $TMPDIR.
1612 Initialize c_file/o_file with concat, which avoids exposing
1613 uninitialized memory to signal handler, which calls unlink(!).
1614 Avoid calling maybe_unlink when the main function returns,
1615 since the atexit handler is already doing this.
1616 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1617
1618 2020-02-19 Martin Jambor <mjambor@suse.cz>
1619
1620 PR tree-optimization/93776
1621 * tree-sra.c (create_access): Do not create zero size accesses.
1622 (get_access_for_expr): Do not search for zero sized accesses.
1623
1624 2020-02-19 Martin Jambor <mjambor@suse.cz>
1625
1626 PR tree-optimization/93667
1627 * tree-sra.c (scalarizable_type_p): Return false if record fields
1628 do not follow wach other.
1629
1630 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1631
1632 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1633 rather than fmv.x.s/fmv.s.x.
1634
1635 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1636
1637 * config/aarch64/aarch64-simd-builtins.def
1638 (intrinsic_vec_smult_lo_): New.
1639 (intrinsic_vec_umult_lo_): Likewise.
1640 (vec_widen_smult_hi_): Likewise.
1641 (vec_widen_umult_hi_): Likewise.
1642 * config/aarch64/aarch64-simd.md
1643 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1644 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1645 (vmull_high_s16): Likewise.
1646 (vmull_high_s32): Likewise.
1647 (vmull_high_u8): Likewise.
1648 (vmull_high_u16): Likewise.
1649 (vmull_high_u32): Likewise.
1650 (vmull_s8): Likewise.
1651 (vmull_s16): Likewise.
1652 (vmull_s32): Likewise.
1653 (vmull_u8): Likewise.
1654 (vmull_u16): Likewise.
1655 (vmull_u32): Likewise.
1656
1657 2020-02-18 Martin Liska <mliska@suse.cz>
1658
1659 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1660 bootstrap by missing removal of invalid sanity check.
1661
1662 2020-02-18 Martin Liska <mliska@suse.cz>
1663
1664 PR ipa/92518
1665 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1666 Always compare LHS of gimple_assign.
1667
1668 2020-02-18 Martin Liska <mliska@suse.cz>
1669
1670 PR ipa/93583
1671 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1672 and return type of functions.
1673 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1674 Drop MALLOC attribute for void functions.
1675 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1676 malloc_state for a new VOID clone.
1677
1678 2020-02-18 Martin Liska <mliska@suse.cz>
1679
1680 PR ipa/92924
1681 * common.opt: Add -fprofile-reproducibility.
1682 * doc/invoke.texi: Document it.
1683 * value-prof.c (dump_histogram_value):
1684 Document and support behavior for counters[0]
1685 being a negative value.
1686 (get_nth_most_common_value): Handle negative
1687 counters[0] in respect to flag_profile_reproducible.
1688
1689 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1690
1691 PR ipa/93797
1692 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1693 speculative_uid in messages. Remove trailing whitespace from error
1694 message. Use num_speculative_call_targets instead of
1695 num_speculative_targets in a message.
1696 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1697 edge messages and stmt instead of cal_stmt in reference message.
1698
1699 PR tree-optimization/93780
1700 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1701 before calling build_vector_type.
1702 (execute_update_addresses_taken): Likewise.
1703
1704 PR driver/93796
1705 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1706 typo, functoin -> function.
1707 * tree.c (free_lang_data_in_decl): Fix comment typo,
1708 functoin -> function.
1709 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1710
1711 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1712
1713 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1714 won't be printed.
1715 (print_option_information): Don't call get_option_url if URLs
1716 won't be printed.
1717
1718 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1719
1720 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1721 handling of register_common-less targets.
1722
1723 2020-02-17 Martin Liska <mliska@suse.cz>
1724
1725 PR ipa/93760
1726 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1727
1728 2020-02-17 Martin Liska <mliska@suse.cz>
1729
1730 PR translation/93755
1731 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1732 Fix double quotes.
1733
1734 2020-02-17 Martin Liska <mliska@suse.cz>
1735
1736 PR other/93756
1737 * config/rx/elf.opt: Fix typo.
1738
1739 2020-02-17 Richard Biener <rguenther@suse.de>
1740
1741 PR c/86134
1742 * opts-global.c (print_ignored_options): Use inform and
1743 amend message.
1744
1745 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1746
1747 PR target/93047
1748 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1749
1750 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1751
1752 PR target/93743
1753 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1754 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1755
1756 2020-02-15 Jason Merrill <jason@redhat.com>
1757
1758 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1759
1760 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1761
1762 PR tree-optimization/93744
1763 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1764 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1765 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1766 sure @2 in the first and @1 in the other patterns has no side-effects.
1767
1768 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1769 Bernd Edlinger <bernd.edlinger@hotmail.de>
1770
1771 PR 87488
1772 PR other/93168
1773 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1774 * configure.ac (--with-diagnostics-urls): New configuration
1775 option, based on --with-diagnostics-color.
1776 (DIAGNOSTICS_URLS_DEFAULT): New define.
1777 * config.h: Regenerate.
1778 * configure: Regenerate.
1779 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1780 DIAGNOSTICS_URLS_DEFAULT from configure-time
1781 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1782 and TERM_URLS environment variable.
1783 * diagnostic-url.h (diagnostic_url_format): New enum type.
1784 (diagnostic_urls_enabled_p): rename to...
1785 (determine_url_format): ... this, and change return type.
1786 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1787 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1788 the linux console, and mingw.
1789 (diagnostic_urls_enabled_p): rename to...
1790 (determine_url_format): ... this, and adjust.
1791 * pretty-print.h (pretty_printer::show_urls): rename to...
1792 (pretty_printer::url_format): ... this, and change to enum.
1793 * pretty-print.c (pretty_printer::pretty_printer,
1794 pp_begin_url, pp_end_url, test_urls): Adjust.
1795 * doc/install.texi (--with-diagnostics-urls): Document the new
1796 configuration option.
1797 (--with-diagnostics-color): Document the existing interaction
1798 with GCC_COLORS better.
1799 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1800 vindex reference. Update description of defaults based on the above.
1801 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1802 interacts with GCC_COLORS.
1803
1804 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1805
1806 PR target/93704
1807 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1808 conjunction with TARGET_GNU_TLS in early return.
1809
1810 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1811
1812 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1813 the mode is not wider than UNITS_PER_WORD.
1814
1815 2020-02-14 Martin Jambor <mjambor@suse.cz>
1816
1817 PR tree-optimization/93516
1818 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1819 access of the same type as the parent.
1820 (propagate_subaccesses_from_lhs): Likewise.
1821
1822 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1823
1824 PR target/93724
1825 * config/i386/avx512vbmi2intrin.h
1826 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1827 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1828 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1829 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1830 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1831 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1832 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1833 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1834 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1835 of lacking a closing parenthesis.
1836 * config/i386/avx512vbmi2vlintrin.h
1837 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1838 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1839 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1840 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1841 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1842 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1843 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1844 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1845 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1846 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1847 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1848 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1849 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1850 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1851 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1852 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1853 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1854 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1855
1856 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1857
1858 PR target/93656
1859 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1860 the target function entry.
1861
1862 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1863
1864 * common/config/arc/arc-common.c (arc_option_optimization_table):
1865 Disable if-conversion step when optimized for size.
1866
1867 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1868
1869 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1870 R12-R15 are always in ARCOMPACT16_REGS register class.
1871 * config/arc/arc.opt (mq-class): Deprecate.
1872 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1873 option.
1874 * doc/invoke.texi (mq-class): Update text.
1875 * common/config/arc/arc-common.c (arc_option_optimization_table):
1876 Update list.
1877
1878 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1879
1880 * config/arc/arc.c (arc_insn_cost): New function.
1881 (TARGET_INSN_COST): Define.
1882 * config/arc/arc.md (cost): New attribute.
1883 (add_n): Use arc_nonmemory_operand.
1884 (ashlsi3_insn): Likewise, also update constraints.
1885 (ashrsi3_insn): Likewise.
1886 (rotrsi3): Likewise.
1887 (add_shift): Likewise.
1888 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1889
1890 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1891
1892 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1893 registers.
1894 (umulsidi_600): Likewise.
1895
1896 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1897
1898 PR target/93696
1899 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1900 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1901 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1902 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1903 pass __A to the builtin followed by __W instead of __A followed by
1904 __B.
1905 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1906 _mm512_mask_popcnt_epi64): Likewise.
1907 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1908 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1909 _mm256_mask_popcnt_epi64): Likewise.
1910
1911 PR tree-optimization/93582
1912 * fold-const.h (shift_bytes_in_array_left,
1913 shift_bytes_in_array_right): Declare.
1914 * fold-const.c (shift_bytes_in_array_left,
1915 shift_bytes_in_array_right): New function, moved from
1916 gimple-ssa-store-merging.c, no longer static.
1917 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1918 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1919 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1920 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1921 shift_bytes_in_array.
1922 (verify_shift_bytes_in_array): Rename to ...
1923 (verify_shift_bytes_in_array_left): ... this. Use
1924 shift_bytes_in_array_left instead of shift_bytes_in_array.
1925 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1926 instead of verify_shift_bytes_in_array.
1927 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1928 / native_interpret_expr where the store covers all needed bits,
1929 punt on PDP-endian, otherwise allow all involved offsets and sizes
1930 not to be byte-aligned.
1931
1932 PR target/93673
1933 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1934 use const_0_to_255_operand predicate instead of immediate_operand.
1935 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1936 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1937 vgf2p8affineinvqb_<mode><mask_name>,
1938 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1939 const_0_to_255_operand predicated operands.
1940
1941 2020-02-12 Jeff Law <law@redhat.com>
1942
1943 * config/h8300/h8300.md (comparison shortening peepholes): Use
1944 a mode iterator to merge the HImode and SImode peepholes.
1945
1946 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1947
1948 PR middle-end/93663
1949 * real.c (is_even): Make static. Function comment fix.
1950 (is_halfway_below): Make static, don't assert R is not inf/nan,
1951 instead return false for those. Small formatting fixes.
1952
1953 2020-02-12 Martin Sebor <msebor@redhat.com>
1954
1955 PR middle-end/93646
1956 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1957 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1958 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1959 (strlen_check_and_optimize_call): Adjust callee name.
1960
1961 2020-02-12 Jeff Law <law@redhat.com>
1962
1963 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1964 (and (xor)) variant. Combine other two into single peephole.
1965
1966 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1967
1968 PR rtl-optimization/93565
1969 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1970
1971 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1972
1973 * config/aarch64/aarch64-simd.md
1974 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1975 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1976 generating separate ADDV and zero_extend patterns.
1977 * config/aarch64/iterators.md (VDQV_E): New iterator.
1978
1979 2020-02-12 Jeff Law <law@redhat.com>
1980
1981 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1982 expanders, splits, etc.
1983 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1984 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1985 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1986 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1987 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1988 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1989 function prototype.
1990 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1991
1992 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1993
1994 PR target/93670
1995 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1996 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1997 TARGET_AVX512DQ from condition.
1998 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1999 instead of <mask_mode512bit_condition> in condition. If
2000 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
2001 vextract*32x8.
2002 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
2003 from condition.
2004
2005 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
2006
2007 PR target/91052
2008 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
2009
2010 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
2011
2012 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
2013 where strlen is more legible.
2014 (rs6000_builtin_vectorized_libmass): Ditto.
2015 (rs6000_print_options_internal): Ditto.
2016
2017 2020-02-11 Martin Sebor <msebor@redhat.com>
2018
2019 PR tree-optimization/93683
2020 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
2021
2022 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
2023
2024 * config/rs6000/predicates.md (cint34_operand): Rename the
2025 -mprefixed-addr option to be -mprefixed.
2026 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
2027 the -mprefixed-addr option to be -mprefixed.
2028 (OTHER_FUTURE_MASKS): Likewise.
2029 (POWERPC_MASKS): Likewise.
2030 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
2031 the -mprefixed-addr option to be -mprefixed. Change error
2032 messages to refer to -mprefixed.
2033 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
2034 -mprefixed.
2035 (rs6000_legitimate_offset_address_p): Likewise.
2036 (rs6000_mode_dependent_address): Likewise.
2037 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
2038 "-mprefixed" for target attributes and pragmas.
2039 (address_to_insn_form): Rename the -mprefixed-addr option to be
2040 -mprefixed.
2041 (rs6000_adjust_insn_length): Likewise.
2042 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
2043 -mprefixed-addr option to be -mprefixed.
2044 (ASM_OUTPUT_OPCODE): Likewise.
2045 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
2046 -mprefixed-addr option to be -mprefixed.
2047 * config/rs6000/rs6000.opt (-mprefixed): Rename the
2048 -mprefixed-addr option to be prefixed. Change the option from
2049 being undocumented to being documented.
2050 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
2051 -mprefixed option. Update the -mpcrel documentation to mention
2052 -mprefixed.
2053
2054 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
2055
2056 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
2057 including FIRST_PSEUDO_REGISTER - 1.
2058 * ira-color.c (print_hard_reg_set): Ditto.
2059
2060 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2061
2062 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
2063 (USTERNOP_QUALIFIERS): New define.
2064 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
2065 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
2066 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
2067 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
2068 * config/arm/arm_neon.h (vusdot_s32): New.
2069 (vusdot_lane_s32): New.
2070 (vusdotq_lane_s32): New.
2071 (vsudot_lane_s32): New.
2072 (vsudotq_lane_s32): New.
2073 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
2074 * config/arm/iterators.md (DOTPROD_I8MM): New.
2075 (sup, opsuffix): Add <us/su>.
2076 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
2077 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
2078
2079 2020-02-11 Richard Biener <rguenther@suse.de>
2080
2081 PR tree-optimization/93661
2082 PR tree-optimization/93662
2083 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
2084 tree_to_poly_int64.
2085 * tree-sra.c (get_access_for_expr): Likewise.
2086
2087 2020-02-10 Jakub Jelinek <jakub@redhat.com>
2088
2089 PR target/93637
2090 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
2091 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
2092 Change condition from TARGET_AVX2 to TARGET_AVX.
2093
2094 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
2095
2096 PR other/93641
2097 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
2098 argument of strncmp.
2099
2100 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
2101
2102 Try to generate zero-based comparisons.
2103 * config/cris/cris.c (cris_reduce_compare): New function.
2104 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
2105 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
2106 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
2107
2108 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
2109
2110 PR target/91913
2111 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
2112 in Thumb state and also as a destination in Arm state. Add T16
2113 variants.
2114
2115 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
2116
2117 * md.texi (Define Subst): Match closing paren in example.
2118
2119 2020-02-10 Jakub Jelinek <jakub@redhat.com>
2120
2121 PR target/58218
2122 PR other/93641
2123 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
2124 arguments of strncmp.
2125
2126 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
2127
2128 PR ipa/93203
2129 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
2130 but different source value.
2131 (adjust_callers_for_value_intersection): New function.
2132 (gather_edges_for_value): Adjust order of callers to let a
2133 non-self-recursive caller be the first element.
2134 (self_recursive_pass_through_p): Add a new parameter "simple", and
2135 check generalized self-recursive pass-through jump function.
2136 (self_recursive_agg_pass_through_p): Likewise.
2137 (find_more_scalar_values_for_callers_subset): Compute value from
2138 pass-through jump function for self-recursive.
2139 (intersect_with_plats): Cleanup previous implementation code for value
2140 itersection with self-recursive call edge.
2141 (intersect_with_agg_replacements): Likewise.
2142 (intersect_aggregates_with_edge): Deduce value from pass-through jump
2143 function for self-recursive call edge. Cleanup previous implementation
2144 code for value intersection with self-recursive call edge.
2145 (decide_whether_version_node): Remove dead callers and adjust order
2146 to let a non-self-recursive caller be the first element.
2147
2148 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
2149
2150 * recog.c: Move pass_split_before_sched2 code in front of
2151 pass_split_before_regstack.
2152 (pass_data_split_before_sched2): Rename pass to split3 from split4.
2153 (pass_data_split_before_regstack): Rename pass to split4 from split3.
2154 (rest_of_handle_split_before_sched2): Remove.
2155 (pass_split_before_sched2::execute): Unconditionally call
2156 split_all_insns.
2157 (enable_split_before_sched2): New function.
2158 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
2159 (pass_split_before_regstack::gate): Ditto.
2160 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
2161 Update name check for renamed split4 pass.
2162 * config/sh/sh.c (register_sh_passes): Update pass insertion
2163 point for renamed split4 pass.
2164
2165 2020-02-09 Jakub Jelinek <jakub@redhat.com>
2166
2167 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
2168 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
2169 copying them around between host and target.
2170
2171 2020-02-08 Andrew Pinski <apinski@marvell.com>
2172
2173 PR target/91927
2174 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
2175 STRICT_ALIGNMENT also.
2176
2177 2020-02-08 Jim Wilson <jimw@sifive.com>
2178
2179 PR target/93532
2180 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
2181
2182 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
2183 Jakub Jelinek <jakub@redhat.com>
2184
2185 PR target/65782
2186 * config/i386/i386.h (CALL_USED_REGISTERS): Make
2187 xmm16-xmm31 call-used even in 64-bit ms-abi.
2188
2189 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
2190
2191 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
2192 (simd_ummla, simd_usmmla): Likewise.
2193 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
2194 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
2195 (vusmmlaq_s32): New.
2196
2197 2020-02-07 Richard Biener <rguenther@suse.de>
2198
2199 PR middle-end/93519
2200 * tree-inline.c (fold_marked_statements): Do a PRE walk,
2201 skipping unreachable regions.
2202 (optimize_inline_calls): Skip folding stmts when we didn't
2203 inline.
2204
2205 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
2206
2207 PR target/85667
2208 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
2209 Don't return aggregates with only SFmode and DFmode in SSE
2210 register.
2211 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
2212
2213 2020-02-07 Jakub Jelinek <jakub@redhat.com>
2214
2215 PR target/93122
2216 * config/rs6000/rs6000-logue.c
2217 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
2218 if it fails, move rs into end_addr and retry. Add
2219 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
2220 the insn pattern doesn't describe well what exactly happens to
2221 dwarf2cfi.c.
2222
2223 PR target/93594
2224 * config/i386/predicates.md (avx_identity_operand): Remove.
2225 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
2226 (avx_<castmode><avxsizesuffix>_<castmode>,
2227 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
2228 a VEC_CONCAT of the operand and UNSPEC_CAST.
2229 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
2230 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
2231 UNSPEC_CAST.
2232
2233 PR target/93611
2234 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
2235 recog_data.insn if distance_non_agu_define changed it.
2236
2237 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
2238
2239 PR target/93569
2240 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
2241 we only had X-FORM (reg+reg) addressing for vectors. Also before
2242 ISA 3.0, we only had X-FORM addressing for scalars in the
2243 traditional Altivec registers.
2244
2245 2020-02-06 <zhongyunde@huawei.com>
2246 Vladimir Makarov <vmakarov@redhat.com>
2247
2248 PR rtl-optimization/93561
2249 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
2250 hard register range.
2251
2252 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2253
2254 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
2255 attribute.
2256
2257 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
2258
2259 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
2260 where the low and the high 32 bits are equal to each other specially,
2261 with an rldimi instruction.
2262
2263 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2264
2265 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
2266
2267 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2268
2269 * config/arm/arm-tables.opt: Regenerate.
2270
2271 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2272
2273 PR target/87763
2274 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
2275 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
2276 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
2277
2278 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2279
2280 PR rtl-optimization/87763
2281 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
2282
2283 2020-02-06 Delia Burduv <delia.burduv@arm.com>
2284
2285 * config/aarch64/aarch64-simd-builtins.def
2286 (bfmlaq): New built-in function.
2287 (bfmlalb): New built-in function.
2288 (bfmlalt): New built-in function.
2289 (bfmlalb_lane): New built-in function.
2290 (bfmlalt_lane): New built-in function.
2291 * config/aarch64/aarch64-simd.md
2292 (aarch64_bfmmlaqv4sf): New pattern.
2293 (aarch64_bfmlal<bt>v4sf): New pattern.
2294 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
2295 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
2296 (vbfmlalbq_f32): New intrinsic.
2297 (vbfmlaltq_f32): New intrinsic.
2298 (vbfmlalbq_lane_f32): New intrinsic.
2299 (vbfmlaltq_lane_f32): New intrinsic.
2300 (vbfmlalbq_laneq_f32): New intrinsic.
2301 (vbfmlaltq_laneq_f32): New intrinsic.
2302 * config/aarch64/iterators.md (BF_MLA): New int iterator.
2303 (bt): New int attribute.
2304
2305 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
2306
2307 * config/i386/i386.md (*pushtf): Emit "#" instead of
2308 calling gcc_unreachable in insn output.
2309 (*pushxf): Ditto.
2310 (*pushdf): Ditto.
2311 (*pushsf_rex64): Ditto for alternatives other than 1.
2312 (*pushsf): Ditto for alternatives other than 1.
2313
2314 2020-02-06 Martin Liska <mliska@suse.cz>
2315
2316 PR gcov-profile/91971
2317 PR gcov-profile/93466
2318 * coverage.c (coverage_init): Revert mangling of
2319 path into filename. It can lead to huge filename length.
2320 Creation of subfolders seem more natural.
2321
2322 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2323
2324 PR target/93300
2325 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
2326 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
2327 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
2328
2329 2020-02-06 Jakub Jelinek <jakub@redhat.com>
2330
2331 PR target/93594
2332 * config/i386/predicates.md (avx_identity_operand): New predicate.
2333 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
2334 define_insn_and_split.
2335
2336 PR libgomp/93515
2337 * omp-low.c (use_pointer_for_field): For nested constructs, also
2338 look for map clauses on target construct.
2339 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
2340 taskreg_nesting_level.
2341
2342 PR libgomp/93515
2343 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
2344 shared clause, call omp_notice_variable on outer context if any.
2345
2346 2020-02-05 Jason Merrill <jason@redhat.com>
2347
2348 PR c++/92003
2349 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
2350 non-zero address even if weak and not yet defined.
2351
2352 2020-02-05 Martin Sebor <msebor@redhat.com>
2353
2354 PR tree-optimization/92765
2355 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
2356 * tree-ssa-strlen.c (compute_string_length): Remove.
2357 (determine_min_objsize): Remove.
2358 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
2359 Avoid using type size as the upper bound on string length.
2360 (handle_builtin_string_cmp): Add an argument. Adjust.
2361 (strlen_check_and_optimize_call): Pass additional argument to
2362 handle_builtin_string_cmp.
2363
2364 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
2365
2366 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
2367 (*pushdi2_rex64 peephole2): Unconditionally split after
2368 epilogue_completed.
2369 (*ashl<mode>3_doubleword): Ditto.
2370 (*<shift_insn><mode>3_doubleword): Ditto.
2371
2372 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
2373
2374 PR target/93568
2375 * config/rs6000/rs6000.c (get_vector_offset): Fix
2376
2377 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
2378
2379 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2380
2381 2020-02-05 David Malcolm <dmalcolm@redhat.com>
2382
2383 * doc/analyzer.texi
2384 (Special Functions for Debugging the Analyzer): Update description
2385 of __analyzer_dump_exploded_nodes.
2386
2387 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2388
2389 PR target/92190
2390 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2391 include sets and not clobbers in the vzeroupper pattern.
2392 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2393 the parallel has 17 (64-bit) or 9 (32-bit) elts.
2394 (*avx_vzeroupper_1): New define_insn_and_split.
2395
2396 PR target/92190
2397 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2398 don't run when !optimize.
2399 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2400 when !optimize.
2401
2402 2020-02-05 Richard Biener <rguenther@suse.de>
2403
2404 PR middle-end/90648
2405 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2406 checks before matching calls.
2407
2408 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2409
2410 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2411 function comment typo.
2412
2413 PR middle-end/93555
2414 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2415 simd_clone_create failed when i == 0, adjust clone->nargs by
2416 clone->inbranch.
2417
2418 2020-02-05 Martin Liska <mliska@suse.cz>
2419
2420 PR c++/92717
2421 * doc/invoke.texi: Document that one should
2422 not combine ASLR and -fpch.
2423
2424 2020-02-04 Richard Biener <rguenther@suse.de>
2425
2426 PR tree-optimization/93538
2427 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2428
2429 2020-02-04 Richard Biener <rguenther@suse.de>
2430
2431 PR tree-optimization/91123
2432 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2433 (vn_walk_cb_data::last_vuse): New member.
2434 (vn_walk_cb_data::saved_operands): Likewsie.
2435 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2436 (vn_walk_cb_data::push_partial_def): Use finish.
2437 (vn_reference_lookup_2): Update last_vuse and use finish if
2438 we've saved operands.
2439 (vn_reference_lookup_3): Use finish and update calls to
2440 push_partial_defs everywhere. When translating through
2441 memcpy or aggregate copies save off operands and alias-set.
2442 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2443 operation for redundant store removal.
2444
2445 2020-02-04 Richard Biener <rguenther@suse.de>
2446
2447 PR tree-optimization/92819
2448 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2449 generating more stmts than before.
2450
2451 2020-02-04 Martin Liska <mliska@suse.cz>
2452
2453 * config/arm/arm.c (arm_gen_far_branch): Move the function
2454 outside of selftests.
2455
2456 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2457
2458 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2459 function to adjust PC-relative vector addresses.
2460 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2461 handle vectors with PC-relative addresses.
2462
2463 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2464
2465 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2466 reference.
2467 (hard_reg_and_mode_to_addr_mask): Delete.
2468 (rs6000_adjust_vec_address): If the original vector address
2469 was REG+REG or REG+OFFSET and the element is not zero, do the add
2470 of the elements in the original address before adding the offset
2471 for the vector element. Use address_to_insn_form to validate the
2472 address using the register being loaded, rather than guessing
2473 whether the address is a DS-FORM or DQ-FORM address.
2474
2475 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2476
2477 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2478 to calculate the offset in memory from the start of a vector of a
2479 particular element. Add code to keep the element number in
2480 bounds if the element number is variable.
2481 (rs6000_adjust_vec_address): Move calculation of offset of the
2482 vector element to get_vector_offset.
2483 (rs6000_split_vec_extract_var): Do not do the initial AND of
2484 element here, move the code to get_vector_offset.
2485
2486 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2487
2488 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2489 gcc_asserts.
2490
2491 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2492
2493 * config/rs6000/constraints.md: Improve documentation.
2494
2495 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2496
2497 PR target/93548
2498 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2499 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2500
2501 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2502
2503 * config.gcc: Remove "carrizo" support.
2504 * config/gcn/gcn-opts.h (processor_type): Likewise.
2505 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2506 * config/gcn/gcn.opt (gpu_type): Likewise.
2507 * config/gcn/t-omp-device: Likewise.
2508
2509 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2510
2511 PR target/91816
2512 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2513 * config/arm/arm.c (arm_gen_far_branch): New function
2514 arm_gen_far_branch.
2515 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2516
2517 2020-02-03 Julian Brown <julian@codesourcery.com>
2518 Tobias Burnus <tobias@codesourcery.com>
2519
2520 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2521
2522 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2523
2524 PR target/93533
2525 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2526 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2527 result.
2528
2529 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2530
2531 PR rtl-optimization/91333
2532 * ira-color.c (struct allocno_color_data): Add member
2533 hard_reg_prefs.
2534 (init_allocno_threads): Set the member up.
2535 (bucket_allocno_compare_func): Add compare hard reg
2536 prefs.
2537
2538 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2539
2540 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2541
2542 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2543 * config.in: Regenerated.
2544 * configure: Regenerated.
2545 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2546 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2547 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2548
2549 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2550
2551 * configure: Regenerate.
2552
2553 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2554
2555 PR rtl-optimization/91333
2556 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2557 reg preferences comparison up.
2558
2559 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2560
2561 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2562 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2563 aarch64-sve-builtins-base.h.
2564 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2565 aarch64-sve-builtins-base.cc.
2566 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2567 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2568 (svcvtnt): Declare.
2569 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2570 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2571 (svcvtnt): New functions.
2572 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2573 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2574 (svcvtnt): New functions.
2575 (svcvt): Add a form that converts f32 to bf16.
2576 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2577 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2578 Declare.
2579 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2580 Treat B as bfloat16_t.
2581 (ternary_bfloat_lane_base): New class.
2582 (ternary_bfloat_def): Likewise.
2583 (ternary_bfloat): New shape.
2584 (ternary_bfloat_lane_def): New class.
2585 (ternary_bfloat_lane): New shape.
2586 (ternary_bfloat_lanex2_def): New class.
2587 (ternary_bfloat_lanex2): New shape.
2588 (ternary_bfloat_opt_n_def): New class.
2589 (ternary_bfloat_opt_n): New shape.
2590 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2591 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2592 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2593 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2594 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2595 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2596 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2597 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2598 the pattern off the narrow mode instead of the wider one.
2599 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2600 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2601 (sve_fp_op): Handle them.
2602 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2603 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2604
2605 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2606
2607 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2608 * config/aarch64/aarch64-modes.def (BF): Move definition before
2609 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2610 (SVE_MODES): Handle BF modes.
2611 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2612 BF modes.
2613 (aarch64_full_sve_mode): Likewise.
2614 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2615 and VNx32BF.
2616 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2617 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2618 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2619 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2620 new SVE BF modes.
2621 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2622 type_class_index.
2623 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2624 (TYPES_all_data): Add bf16.
2625 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2626 (register_tuple_type): Increase buffer size.
2627 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2628 (bf16): New type suffix.
2629 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2630 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2631 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2632 Change type from all_data to all_arith.
2633 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2634 (svminp): Likewise.
2635
2636 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2637 Matthew Malcomson <matthew.malcomson@arm.com>
2638 Richard Sandiford <richard.sandiford@arm.com>
2639
2640 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2641 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2642 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2643 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2644 __ARM_FEATURE_MATMUL_FP64.
2645 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2646 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2647 be disabled at the same time.
2648 (f32mm): New extension.
2649 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2650 (AARCH64_FL_F64MM): Bump to the next bit up.
2651 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2652 (TARGET_SVE_F64MM): New macros.
2653 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2654 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2655 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2656 (UNSPEC_ZIP2Q): New unspeccs.
2657 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2658 (optab, sur, perm_insn): Handle the new unspecs.
2659 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2660 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2661 TARGET_SVE_F64MM instead of separate tests.
2662 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2663 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2664 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2665 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2666 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2667 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2668 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2669 (TYPES_s_signed): New macro.
2670 (TYPES_s_integer): Use it.
2671 (TYPES_d_float): New macro.
2672 (TYPES_d_data): Use it.
2673 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2674 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2675 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2676 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2677 (svmmla): New shape.
2678 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2679 template parameters.
2680 (ternary_resize2_lane_base): Likewise.
2681 (ternary_resize2_base): New class.
2682 (ternary_qq_lane_base): Likewise.
2683 (ternary_intq_uintq_lane_def): Likewise.
2684 (ternary_intq_uintq_lane): New shape.
2685 (ternary_intq_uintq_opt_n_def): New class
2686 (ternary_intq_uintq_opt_n): New shape.
2687 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2688 (ternary_uintq_intq_def): New class.
2689 (ternary_uintq_intq): New shape.
2690 (ternary_uintq_intq_lane_def): New class.
2691 (ternary_uintq_intq_lane): New shape.
2692 (ternary_uintq_intq_opt_n_def): New class.
2693 (ternary_uintq_intq_opt_n): New shape.
2694 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2695 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2696 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2697 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2698 Generalize to...
2699 (svdotprod_lane_impl): ...this new class.
2700 (svmmla_impl, svusdot_impl): New classes.
2701 (svdot_lane): Update to use svdotprod_lane_impl.
2702 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2703 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2704 functions.
2705 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2706 function, with no types defined.
2707 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2708 AARCH64_FL_I8MM functions.
2709 (svmmla): New AARCH64_FL_F32MM function.
2710 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2711 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2712 AARCH64_FL_F64MM function.
2713 (REQUIRED_EXTENSIONS):
2714
2715 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2716
2717 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2718 alternative only.
2719
2720 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2721
2722 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2723 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2724 (*movti_internal): Do not check for
2725 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2726 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2727 just after check for TARGET_AVX.
2728 (*movdf_internal): Ditto.
2729 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2730 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2731 * config/i386/sse.md (mov<mode>_internal): Only check
2732 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2733 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2734 (<sse>_andnot<mode>3<mask_name>): Move check for
2735 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2736 (<code><mode>3<mask_name>): Ditto.
2737 (*andnot<mode>3): Ditto.
2738 (*andnottf3): Ditto.
2739 (*<code><mode>3): Ditto.
2740 (*<code>tf3): Ditto.
2741 (*andnot<VI:mode>3): Remove
2742 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2743 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2744 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2745 (sse4_1_blendv<ssemodesuffix>): Ditto.
2746 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2747 Explain that tune applies to 128bit instructions only.
2748
2749 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2750
2751 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2752 to definition of hsa_kernel_description. Parse assembly to find SGPR
2753 and VGPR count of kernel and store in hsa_kernel_description.
2754
2755 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2756
2757 PR rtl-optimization/91838
2758 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2759 to truncate if allowed or reject combination.
2760
2761 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2762
2763 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2764 (find_inv_vars_cb): Likewise.
2765
2766 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2767
2768 * calls.c (special_function_p): Split out the check for DECL_NAME
2769 being non-NULL and fndecl being extern at file scope into a
2770 new maybe_special_function_p and call it. Drop check for fndecl
2771 being non-NULL that was after a usage of DECL_NAME (fndecl).
2772 * tree.h (maybe_special_function_p): New inline function.
2773
2774 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2775
2776 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2777 (mask_gather_load<mode>): ... here, and zero-initialize the
2778 destination.
2779 (maskload<mode>di): Zero-initialize the destination.
2780 * config/gcn/gcn.c:
2781
2782 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2783
2784 PR analyzer/93356
2785 * doc/analyzer.texi (Limitations): Note that constraints on
2786 floating-point values are currently ignored.
2787
2788 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2789
2790 PR lto/93384
2791 * symtab.c (symtab_node::noninterposable_alias): If localalias
2792 already exists, but is not usable, append numbers after it until
2793 a unique name is found. Formatting fix.
2794
2795 PR middle-end/93505
2796 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2797 rotate counts.
2798
2799 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2800
2801 * config/gcn/gcn.c (print_operand): Handle LTGT.
2802 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2803
2804 2020-01-30 Richard Biener <rguenther@suse.de>
2805
2806 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2807 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2808
2809 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2810
2811 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2812 without a DECL in .data.rel.ro.local.
2813
2814 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2815
2816 PR target/93494
2817 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2818 returned.
2819
2820 PR target/91824
2821 * config/i386/sse.md
2822 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2823 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2824 any_extend code iterator instead of always zero_extend.
2825 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2826 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2827 Use any_extend code iterator instead of always zero_extend.
2828 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2829 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2830 Use any_extend code iterator instead of always zero_extend.
2831 (*sse2_pmovmskb_ext): New define_insn.
2832 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2833
2834 PR target/91824
2835 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2836 (*popcountsi2_zext_falsedep): New define_insn.
2837
2838 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2839
2840 * config.in: Regenerated.
2841 * configure: Regenerated.
2842
2843 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2844
2845 PR bootstrap/93409
2846 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2847 LLVM's assembler changed the default in version 9.
2848
2849 2020-01-24 Jeff Law <law@redhat.com>
2850
2851 PR tree-optimization/89689
2852 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2853
2854 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2855
2856 Revert:
2857
2858 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2859
2860 PR rtl-optimization/87763
2861 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2862 simplification to handle subregs as well as bare regs.
2863 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2864
2865 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2866
2867 PR target/93221
2868 * ira.c (ira): Revert use of simplified LRA algorithm.
2869
2870 2020-01-29 Martin Jambor <mjambor@suse.cz>
2871
2872 PR tree-optimization/92706
2873 * tree-sra.c (struct access): Fields first_link, last_link,
2874 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2875 next_rhs_queued and grp_rhs_queued respectively, new fields
2876 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2877 (struct assign_link): Field next renamed to next_rhs, new field
2878 next_lhs. Updated comment.
2879 (work_queue_head): Renamed to rhs_work_queue_head.
2880 (lhs_work_queue_head): New variable.
2881 (add_link_to_lhs): New function.
2882 (relink_to_new_repr): Also relink LHS lists.
2883 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2884 (add_access_to_lhs_work_queue): New function.
2885 (pop_access_from_work_queue): Renamed to
2886 pop_access_from_rhs_work_queue.
2887 (pop_access_from_lhs_work_queue): New function.
2888 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2889 work_queue.
2890 (child_would_conflict_in_lacc): Renamed to
2891 child_would_conflict_in_acc. Adjusted parameter names.
2892 (create_artificial_child_access): New parameter set_grp_read, use it.
2893 (subtree_mark_written_and_enqueue): Renamed to
2894 subtree_mark_written_and_rhs_enqueue.
2895 (propagate_subaccesses_across_link): Renamed to
2896 propagate_subaccesses_from_rhs.
2897 (propagate_subaccesses_from_lhs): New function.
2898 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2899 RHSs.
2900
2901 2020-01-29 Martin Jambor <mjambor@suse.cz>
2902
2903 PR tree-optimization/92706
2904 * tree-sra.c (struct access): Adjust comment of
2905 grp_total_scalarization.
2906 (find_access_in_subtree): Look for single children spanning an entire
2907 access.
2908 (scalarizable_type_p): Allow register accesses, adjust callers.
2909 (completely_scalarize): Remove function.
2910 (scalarize_elem): Likewise.
2911 (create_total_scalarization_access): Likewise.
2912 (sort_and_splice_var_accesses): Do not track total scalarization
2913 flags.
2914 (analyze_access_subtree): New parameter totally, adjust to new meaning
2915 of grp_total_scalarization.
2916 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2917 (can_totally_scalarize_forest_p): New function.
2918 (create_total_scalarization_access): Likewise.
2919 (create_total_access_and_reshape): Likewise.
2920 (total_should_skip_creating_access): Likewise.
2921 (totally_scalarize_subtree): Likewise.
2922 (analyze_all_variable_accesses): Perform total scalarization after
2923 subaccess propagation using the new functions above.
2924 (initialize_constant_pool_replacements): Output initializers by
2925 traversing the access tree.
2926
2927 2020-01-29 Martin Jambor <mjambor@suse.cz>
2928
2929 * tree-sra.c (verify_sra_access_forest): New function.
2930 (verify_all_sra_access_forests): Likewise.
2931 (create_artificial_child_access): Set parent.
2932 (analyze_all_variable_accesses): Call the verifier.
2933
2934 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2935
2936 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2937 if called on indirect edge.
2938 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2939 speculative call if needed.
2940
2941 2020-01-29 Richard Biener <rguenther@suse.de>
2942
2943 PR tree-optimization/93428
2944 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2945 permutation when the load node is created.
2946 (vect_analyze_slp_instance): Re-use it here.
2947
2948 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2949
2950 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2951
2952 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2953
2954 PR rtl-optimization/93272
2955 * ira-lives.c (process_out_of_region_eh_regs): New function.
2956 (process_bb_node_lives): Call it.
2957
2958 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2959
2960 * coverage.c (read_counts_file): Make error message lowercase.
2961
2962 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2963
2964 * profile-count.c (profile_quality_display_names): Fix ordering.
2965
2966 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2967
2968 PR lto/93318
2969 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2970 hash only when edge is first within the sequence.
2971 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2972 (symbol_table::create_edge): Do not set target_prob.
2973 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2974 the call site hash.
2975 (cgraph_edge::make_speculative): Drop target_prob parameter.
2976 (cgraph_edge::speculative_call_info): Remove.
2977 (cgraph_edge::first_speculative_call_target): New member function.
2978 (update_call_stmt_hash_for_removing_direct_edge): New function.
2979 (cgraph_edge::resolve_speculation): Rewrite to new API.
2980 (cgraph_edge::speculative_call_for_target): New member function.
2981 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2982 multiple speculation targets.
2983 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2984 of profile.
2985 (verify_speculative_call): Verify that targets form an interval.
2986 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2987 (cgraph_edge::first_speculative_call_target): New member function.
2988 (cgraph_edge::next_speculative_call_target): New member function.
2989 (cgraph_edge::speculative_call_target_ref): New member function.
2990 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2991 (cgraph_edge): Remove target_prob.
2992 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2993 Fix handling of speculative calls.
2994 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2995 * ipa-fnsummary.c (analyze_function_body): Likewise.
2996 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2997 * ipa-profile.c (dump_histogram): Fix formating.
2998 (ipa_profile_generate_summary): Watch for overflows.
2999 (ipa_profile): Do not require probablity to be 1/2; update to new API.
3000 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
3001 (update_indirect_edges_after_inlining): Update to new API.
3002 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
3003 profiles.
3004 * profile-count.h: (profile_probability::adjusted): New.
3005 * tree-inline.c (copy_bb): Update to new speculative call API; fix
3006 updating of profile.
3007 * value-prof.c (gimple_ic_transform): Rename to ...
3008 (dump_ic_profile): ... this one; update dumping.
3009 (stream_in_histogram_value): Fix formating.
3010 (gimple_value_profile_transformations): Update.
3011
3012 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
3013
3014 PR target/91461
3015 * config/i386/i386.md (*movoi_internal_avx): Remove
3016 TARGET_SSE_TYPELESS_STORES check.
3017 (*movti_internal): Prefer TARGET_AVX over
3018 TARGET_SSE_TYPELESS_STORES.
3019 (*movtf_internal): Likewise.
3020 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
3021 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
3022 from TARGET_SSE_TYPELESS_STORES.
3023
3024 2020-01-28 David Malcolm <dmalcolm@redhat.com>
3025
3026 * diagnostic-core.h (warning_at): Rename overload to...
3027 (warning_meta): ...this.
3028 (emit_diagnostic_valist): Delete decl of overload taking
3029 diagnostic_metadata.
3030 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
3031 (warning_at): Rename overload taking diagnostic_metadata to...
3032 (warning_meta): ...this.
3033
3034 2020-01-28 Richard Biener <rguenther@suse.de>
3035
3036 PR tree-optimization/93439
3037 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
3038 * tree-cfg.c (move_sese_region_to_fn): ... here.
3039 (verify_types_in_gimple_reference): Verify used cliques are
3040 tracked.
3041
3042 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
3043
3044 PR target/91399
3045 * config/i386/i386-options.c (set_ix86_tune_features): Add an
3046 argument of a pointer to struct gcc_options and pass it to
3047 parse_mtune_ctrl_str.
3048 (ix86_function_specific_restore): Pass opts to
3049 set_ix86_tune_features.
3050 (ix86_option_override_internal): Likewise.
3051 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
3052 gcc_options and use it for x_ix86_tune_ctrl_string.
3053
3054 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3055
3056 PR rtl-optimization/87763
3057 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
3058 simplification to handle subregs as well as bare regs.
3059 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
3060
3061 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3062
3063 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
3064 for reduction chains that (now) include a call.
3065
3066 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3067
3068 PR tree-optimization/92822
3069 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
3070 out the don't-care elements of a vector whose significant elements
3071 are duplicates, make the don't-care elements duplicates too.
3072
3073 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3074
3075 PR tree-optimization/93434
3076 * tree-predcom.c (split_data_refs_to_components): Record which
3077 components have had aliasing loads removed. Prevent store-store
3078 commoning for all such components.
3079
3080 2020-01-28 Jakub Jelinek <jakub@redhat.com>
3081
3082 PR target/93418
3083 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
3084 -1 or is_vshift is true, use new_vector with number of elts npatterns
3085 rather than new_unary_operation.
3086
3087 PR tree-optimization/93454
3088 * gimple-fold.c (fold_array_ctor_reference): Perform
3089 elt_size.to_uhwi () just once, instead of calling it in every
3090 iteration. Punt if that value is above size of the temporary
3091 buffer. Decrease third native_encode_expr argument when
3092 bufoff + elt_sz is above size of buf.
3093
3094 2020-01-27 Joseph Myers <joseph@codesourcery.com>
3095
3096 * config/mips/mips.c (mips_declare_object_name)
3097 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
3098
3099 2020-01-27 Martin Liska <mliska@suse.cz>
3100
3101 PR gcov-profile/93403
3102 * tree-profile.c (gimple_init_gcov_profiler): Generate
3103 both __gcov_indirect_call_profiler_v4 and
3104 __gcov_indirect_call_profiler_v4_atomic.
3105
3106 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3107
3108 PR target/92822
3109 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
3110 expander.
3111 (@aarch64_split_simd_mov<mode>): Use it.
3112 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
3113 Leave the vec_extract patterns to handle 2-element vectors.
3114 (aarch64_simd_mov_from_<mode>high): Likewise.
3115 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
3116 (vec_extractv2dfv1df): Likewise.
3117
3118 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3119
3120 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
3121 jump conditions for *compare_condjump<GPI:mode>.
3122
3123 2020-01-27 David Malcolm <dmalcolm@redhat.com>
3124
3125 PR analyzer/93276
3126 * digraph.cc (test_edge::test_edge): Specify template for base
3127 class initializer.
3128
3129 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3130
3131 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
3132
3133 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3134
3135 * config/arc/arc-protos.h (gen_mlo): Remove.
3136 (gen_mhi): Likewise.
3137 * config/arc/arc.c (AUX_MULHI): Define.
3138 (arc_must_save_reister): Special handling for r58/59.
3139 (arc_compute_frame_size): Consider mlo/mhi registers.
3140 (arc_save_callee_saves): Emit fp/sp move only when emit_move
3141 paramter is true.
3142 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
3143 mlo/mhi name selection.
3144 (arc_restore_callee_saves): Don't early restore blink when ISR.
3145 (arc_expand_prologue): Add mlo/mhi saving.
3146 (arc_expand_epilogue): Add mlo/mhi restoring.
3147 (gen_mlo): Remove.
3148 (gen_mhi): Remove.
3149 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
3150 numbering when MUL64 option is used.
3151 (DWARF2_FRAME_REG_OUT): Define.
3152 * config/arc/arc.md (arc600_stall): New pattern.
3153 (VUNSPEC_ARC_ARC600_STALL): Define.
3154 (mulsi64): Use correct mlo/mhi registers.
3155 (mulsi_600): Clean it up.
3156 * config/arc/predicates.md (mlo_operand): Remove any dependency on
3157 TARGET_BIG_ENDIAN.
3158 (mhi_operand): Likewise.
3159
3160 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3161 Petro Karashchenko <petro.karashchenko@ring.com>
3162
3163 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
3164 attributes if needed.
3165 (prepare_move_operands): Generate special unspec instruction for
3166 direct access.
3167 (arc_isuncached_mem_p): Propagate uncached attribute to each
3168 structure member.
3169 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
3170 (VUNSPEC_ARC_STDI): Likewise.
3171 (ALLI): New mode iterator.
3172 (mALLI): New mode attribute.
3173 (lddi): New instruction pattern.
3174 (stdi): Likewise.
3175 (stdidi_split): Split instruction for architectures which are not
3176 supporting ll64 option.
3177 (lddidi_split): Likewise.
3178
3179 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3180
3181 PR rtl-optimization/92989
3182 * lra-lives.c (process_bb_lives): Update the live-in set before
3183 processing additional clobbers.
3184
3185 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3186
3187 PR rtl-optimization/93170
3188 * cselib.c (cselib_invalidate_regno_val): New function, split out
3189 from...
3190 (cselib_invalidate_regno): ...here.
3191 (cselib_invalidated_by_call_p): New function.
3192 (cselib_process_insn): Iterate over all the hard-register entries in
3193 REG_VALUES and invalidate any that cross call-clobbered registers.
3194
3195 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3196
3197 * dojump.c (split_comparison): Use HONOR_NANS rather than
3198 HONOR_SNANS when splitting LTGT.
3199
3200 2020-01-27 Martin Liska <mliska@suse.cz>
3201
3202 PR driver/91220
3203 * opts.c (print_filtered_help): Exclude language-specific
3204 options from --help=common unless enabled in all FEs.
3205
3206 2020-01-27 Martin Liska <mliska@suse.cz>
3207
3208 * opts.c (print_help): Exclude params from
3209 all except --help=param.
3210
3211 2020-01-27 Martin Liska <mliska@suse.cz>
3212
3213 PR target/93274
3214 * config/i386/i386-features.c (make_resolver_func):
3215 Align the code with ppc64 target implementation.
3216 Do not generate a unique name for resolver function.
3217
3218 2020-01-27 Richard Biener <rguenther@suse.de>
3219
3220 PR tree-optimization/93397
3221 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
3222 converted reduction chain SLP graph adjustment.
3223
3224 2020-01-26 Marek Polacek <polacek@redhat.com>
3225
3226 PR sanitizer/93436
3227 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
3228 null DECL_NAME.
3229
3230 2020-01-26 Jason Merrill <jason@redhat.com>
3231
3232 PR c++/92601
3233 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
3234 of complete types.
3235
3236 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
3237
3238 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
3239 (rx_setmem): Likewise.
3240
3241 2020-01-26 Jakub Jelinek <jakub@redhat.com>
3242
3243 PR target/93412
3244 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
3245 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
3246 drop <di> from constraint of last operand.
3247
3248 PR target/93430
3249 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
3250 TARGET_AVX2 and V4DFmode not in the split condition, but in the
3251 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
3252
3253 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
3254
3255 PR ipa/93166
3256 * ipa-cp.c (get_info_about_necessary_edges): Remove value
3257 check assertion.
3258
3259 2020-01-24 Jeff Law <law@redhat.com>
3260
3261 PR tree-optimization/92788
3262 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
3263 not EDGE_ABNORMAL.
3264
3265 2020-01-24 Jakub Jelinek <jakub@redhat.com>
3266
3267 PR target/93395
3268 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
3269 *avx_vperm_broadcast_<mode>,
3270 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
3271 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
3272 Move before avx2_perm<mode>/avx512f_perm<mode>.
3273
3274 PR target/93376
3275 * simplify-rtx.c (simplify_const_unary_operation,
3276 simplify_const_binary_operation): Punt for mode precision above
3277 MAX_BITSIZE_MODE_ANY_INT.
3278
3279 2020-01-24 Andrew Pinski <apinski@marvell.com>
3280
3281 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
3282 alu.shift_reg to 0.
3283
3284 2020-01-24 Jeff Law <law@redhat.com>
3285
3286 PR target/13721
3287 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
3288 for REGs. Call output_operand_lossage to get more reasonable
3289 diagnostics.
3290
3291 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
3292
3293 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
3294 gcn_fp_compare_operator.
3295 (vec_cmpu<mode>di): Use gcn_compare_operator.
3296 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
3297 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
3298 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
3299 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
3300 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
3301 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
3302 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
3303 gcn_fp_compare_operator.
3304 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
3305 gcn_fp_compare_operator.
3306 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
3307 gcn_fp_compare_operator.
3308 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
3309 gcn_fp_compare_operator.
3310
3311 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
3312
3313 * doc/install.texi (Cross-Compiler-Specific Options): Document
3314 `--with-toolexeclibdir' option.
3315
3316 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
3317
3318 * target.def (flags_regnum): Also mention effect on delay slot filling.
3319 * doc/tm.texi: Regenerate.
3320
3321 2020-01-23 Jeff Law <law@redhat.com>
3322
3323 PR translation/90162
3324 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
3325
3326 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
3327
3328 PR target/92269
3329 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
3330 profiling label
3331
3332 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3333
3334 PR rtl-optimization/93402
3335 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
3336 USE insns.
3337
3338 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3339
3340 * config.in: Regenerated.
3341 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
3342 for TARGET_LIBC_GNUSTACK.
3343 * configure: Regenerated.
3344 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
3345 found to be 2.31 or greater.
3346
3347 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3348
3349 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
3350 TARGET_SOFT_FLOAT.
3351 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
3352 (mips_asm_file_end): New function. Delegate to
3353 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
3354 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3355
3356 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3357
3358 PR target/93376
3359 * config/i386/i386-modes.def (POImode): New mode.
3360 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
3361 * config/i386/i386.md (DPWI): New mode attribute.
3362 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
3363 (QWI): Rename to...
3364 (QPWI): ... this. Use POI instead of OI for TImode.
3365 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
3366 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
3367 instead of <QWI>.
3368
3369 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3370
3371 PR target/93341
3372 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3373 unspec.
3374 (speculation_tracker_rev): New pattern.
3375 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3376 Use speculation_tracker_rev to track the inverse condition.
3377
3378 2020-01-23 Richard Biener <rguenther@suse.de>
3379
3380 PR tree-optimization/93381
3381 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3382 alias-set of the def as argument and record the first one.
3383 (vn_walk_cb_data::first_set): New member.
3384 (vn_reference_lookup_3): Pass the alias-set of the current def
3385 to push_partial_def. Fix alias-set used in the aggregate copy
3386 case.
3387 (vn_reference_lookup): Consistently set *last_vuse_ptr.
3388 * real.c (clear_significand_below): Fix out-of-bound access.
3389
3390 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3391
3392 PR target/93346
3393 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3394 New define_insn patterns.
3395
3396 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3397
3398 * doc/sourcebuild.texi (check-function-bodies): Add an
3399 optional target/xfail selector.
3400
3401 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3402
3403 PR rtl-optimization/93124
3404 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3405 bare USE and CLOBBER insns.
3406
3407 2020-01-22 Andrew Pinski <apinski@marvell.com>
3408
3409 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3410
3411 2020-01-22 David Malcolm <dmalcolm@redhat.com>
3412
3413 PR analyzer/93307
3414 * gdbinit.in (break-on-saved-diagnostic): Update for move of
3415 diagnostic_manager into "ana" namespace.
3416 * selftest-run-tests.c (selftest::run_tests): Update for move of
3417 selftest::run_analyzer_selftests to
3418 ana::selftest::run_analyzer_selftests.
3419
3420 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
3421
3422 * cfgexpand.c (union_stack_vars): Update the size.
3423
3424 2020-01-22 Richard Biener <rguenther@suse.de>
3425
3426 PR tree-optimization/93381
3427 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3428 throughout, handle all conversions the same.
3429
3430 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3431
3432 PR target/93335
3433 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3434 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3435 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3436 Call force_reg on high_in2 unconditionally.
3437
3438 2020-01-22 Martin Liska <mliska@suse.cz>
3439
3440 PR tree-optimization/92924
3441 * profile.c (compute_value_histograms): Divide
3442 all counter values.
3443
3444 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3445
3446 PR target/91298
3447 * output.h (assemble_name_resolve): Declare.
3448 * varasm.c (assemble_name_resolve): New function.
3449 (assemble_name): Use it.
3450 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3451
3452 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3453
3454 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3455 update_web_docs_git instead of update_web_docs_svn.
3456
3457 2020-01-21 Andrew Pinski <apinski@marvell.com>
3458
3459 PR target/9311
3460 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3461 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3462 (*tlsgd_small_<mode>): Likewise.
3463 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3464 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3465 register. Convert that register back to dest using convert_mode.
3466
3467 2020-01-21 Jim Wilson <jimw@sifive.com>
3468
3469 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3470 instead of XINT.
3471
3472 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3473 Uros Bizjak <ubizjak@gmail.com>
3474
3475 PR target/93319
3476 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3477 with ptr_mode.
3478 (legitimize_tls_address): Do GNU2 TLS address computation in
3479 ptr_mode and zero-extend result to Pmode.
3480 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3481 :P with :PTR and Pmode with ptr_mode.
3482 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3483 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3484 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3485
3486 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3487
3488 PR target/93333
3489 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3490 the last two operands are CONST_INT_P before using them as such.
3491
3492 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3493
3494 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3495 to get the integer element types.
3496
3497 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3498
3499 * config/aarch64/aarch64-sve-builtins.h
3500 (function_expander::convert_to_pmode): Declare.
3501 * config/aarch64/aarch64-sve-builtins.cc
3502 (function_expander::convert_to_pmode): New function.
3503 (function_expander::get_contiguous_base): Use it.
3504 (function_expander::prepare_gather_address_operands): Likewise.
3505 * config/aarch64/aarch64-sve-builtins-sve2.cc
3506 (svwhilerw_svwhilewr_impl::expand): Likewise.
3507
3508 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3509
3510 PR target/92424
3511 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3512 cfun->machine->label_is_assembled.
3513 (aarch64_print_patchable_function_entry): New.
3514 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3515 * config/aarch64/aarch64.h (struct machine_function): New field,
3516 label_is_assembled.
3517
3518 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3519
3520 PR ipa/93315
3521 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3522 NULL on exit.
3523
3524 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3525
3526 PR lto/93318
3527 * cgraph.c (cgraph_edge::resolve_speculation,
3528 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3529 call_stmt_site_hash.
3530
3531 2020-01-21 Martin Liska <mliska@suse.cz>
3532
3533 * config/rs6000/rs6000.c (common_mode_defined): Remove
3534 unused variable.
3535
3536 2020-01-21 Richard Biener <rguenther@suse.de>
3537
3538 PR tree-optimization/92328
3539 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3540 type when value-numbering same-sized store by inserting a
3541 VIEW_CONVERT_EXPR.
3542 (eliminate_dom_walker::eliminate_stmt): When eliminating
3543 a redundant store handle bit-reinterpretation of the same value.
3544
3545 2020-01-21 Andrew Pinski <apinski@marvel.com>
3546
3547 PR tree-opt/93321
3548 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3549 from ...
3550 (prepare_block_for_update): This. Use a worklist instead of
3551 recursing.
3552
3553 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3554
3555 * gcc/config/arm/arm.c (clear_operation_p):
3556 Initialise last_regno, skip first iteration
3557 based on the first_set value and use ints instead
3558 of the unnecessary HOST_WIDE_INTs.
3559
3560 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3561
3562 PR target/93073
3563 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3564 compare_mode other than SFmode or DFmode.
3565
3566 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3567
3568 PR target/93304
3569 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3570 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3571 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3572
3573 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3574
3575 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3576
3577 2020-01-20 Andrew Pinski <apinski@marvell.com>
3578
3579 PR middle-end/93242
3580 * targhooks.c (default_print_patchable_function_entry): Use
3581 output_asm_insn to emit the nop instruction.
3582
3583 2020-01-20 Fangrui Song <maskray@google.com>
3584
3585 PR middle-end/93194
3586 * targhooks.c (default_print_patchable_function_entry): Align to
3587 POINTER_SIZE.
3588
3589 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3590
3591 PR target/93319
3592 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3593 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3594 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3595 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3596 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3597 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3598 Remove the {q} suffix from lea.
3599 (*tls_dynamic_gnu2_call_64): Renamed to ...
3600 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3601 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3602 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3603 Pass Pmode to gen_tls_dynamic_gnu2_64.
3604
3605 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3606
3607 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3608
3609 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3610
3611 * config/aarch64/aarch64-sve-builtins-base.cc
3612 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3613
3614 2020-01-20 Richard Biener <rguenther@suse.de>
3615
3616 PR debug/92763
3617 * dwarf2out.c (prune_unused_types): Unconditionally mark
3618 called function DIEs.
3619
3620 2020-01-20 Martin Liska <mliska@suse.cz>
3621
3622 PR tree-optimization/93199
3623 * tree-eh.c (struct leh_state): Add
3624 new field outer_non_cleanup.
3625 (cleanup_is_dead_in): Pass leh_state instead
3626 of eh_region. Add a checking that state->outer_non_cleanup
3627 points to outer non-clean up region.
3628 (lower_try_finally): Record outer_non_cleanup
3629 for this_state.
3630 (lower_catch): Likewise.
3631 (lower_eh_filter): Likewise.
3632 (lower_eh_must_not_throw): Likewise.
3633 (lower_cleanup): Likewise.
3634
3635 2020-01-20 Richard Biener <rguenther@suse.de>
3636
3637 PR tree-optimization/93094
3638 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3639 (vect_transform_loop): Likewise.
3640 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3641 loop_vectorized_call to vect_transform_loop.
3642 * tree-vect-loop.c (vect_transform_loop): Pass down
3643 loop_vectorized_call to vect_loop_versioning.
3644 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3645 the earlier discovered loop_vectorized_call.
3646
3647 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3648
3649 * doc/contribute.texi: Update for SVN -> Git transition.
3650 * doc/install.texi: Likewise.
3651
3652 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3653
3654 PR lto/93318
3655 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3656 speculative targets.
3657 (verify_speculative_call): New function
3658 (cgraph_node::verify_node): Use it.
3659 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3660 speculations.
3661
3662 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3663
3664 PR lto/93318
3665 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3666 (cgraph_edge::make_direct): Remove all indirect targets.
3667 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3668 (cgraph_node::verify_node): Verify that only one call_stmt or
3669 lto_stmt_uid is set.
3670 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3671 lto_stmt_uid.
3672 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3673 (lto_output_ref): Simplify streaming of stmt.
3674 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3675
3676 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3677
3678 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3679 Mark parameter unused.
3680
3681 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3682
3683 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3684
3685 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3686
3687 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3688
3689 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3690
3691 * Makefile.in: Add coroutine-passes.o.
3692 * builtin-types.def (BT_CONST_SIZE): New.
3693 (BT_FN_BOOL_PTR): New.
3694 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3695 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3696 * coroutine-builtins.def: New file.
3697 * coroutine-passes.cc: New file.
3698 * function.h (struct GTY function): Add a bit to indicate that the
3699 function is a coroutine component.
3700 * internal-fn.c (expand_CO_FRAME): New.
3701 (expand_CO_YIELD): New.
3702 (expand_CO_SUSPN): New.
3703 (expand_CO_ACTOR): New.
3704 * internal-fn.def (CO_ACTOR): New.
3705 (CO_YIELD): New.
3706 (CO_SUSPN): New.
3707 (CO_FRAME): New.
3708 * passes.def: Add pass_coroutine_lower_builtins,
3709 pass_coroutine_early_expand_ifns.
3710 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3711 (make_pass_coroutine_early_expand_ifns): New.
3712 * doc/invoke.texi: Document the fcoroutines command line
3713 switch.
3714
3715 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3716
3717 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3718
3719 PR target/93312
3720 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3721 after checking the argument is a REG. Don't use REGNO (reg)
3722 again to set last_regno, reuse regno variable instead.
3723
3724 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3725
3726 * doc/analyzer.texi (Limitations): Add note about NaN.
3727
3728 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3729 Sudakshina Das <sudi.das@arm.com>
3730
3731 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3732 and valid immediate.
3733 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3734 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3735 * config/arm/constraints.md (Pg): New.
3736 * config/arm/predicates.md (long_shift_imm): New.
3737 (arm_reg_or_long_shift_imm): Likewise.
3738 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3739 (thumb2_lsll): Likewise.
3740 (thumb2_lsrl): New.
3741
3742 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3743 Sudakshina Das <sudi.das@arm.com>
3744
3745 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3746 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3747 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3748 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3749 * config/arm/thumb2.md (thumb2_asrl): New.
3750 (thumb2_lsll): Likewise.
3751
3752 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3753
3754 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3755 unused variable.
3756
3757 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3758
3759 * gdbinit.in (help-gcc-hooks): New command.
3760 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3761 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3762 documentation.
3763
3764 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3765
3766 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3767 correct target macro.
3768
3769 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3770
3771 * config/aarch64/aarch64-protos.h
3772 (aarch64_sve_ld1ro_operand_p): New.
3773 * config/aarch64/aarch64-sve-builtins-base.cc
3774 (class load_replicate): New.
3775 (class svld1ro_impl): New.
3776 (class svld1rq_impl): Change to inherit from load_replicate.
3777 (svld1ro): New sve intrinsic function base.
3778 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3779 New DEF_SVE_FUNCTION.
3780 * config/aarch64/aarch64-sve-builtins-base.h
3781 (svld1ro): New decl.
3782 * config/aarch64/aarch64-sve-builtins.cc
3783 (function_expander::add_mem_operand): Modify assert to allow
3784 OImode.
3785 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3786 pattern.
3787 * config/aarch64/aarch64.c
3788 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3789 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3790 (aarch64_sve_ld1ro_operand_p): New.
3791 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3792 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3793 * config/aarch64/predicates.md
3794 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3795
3796 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3797
3798 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3799 Introduce this ACLE specified predefined macro.
3800 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3801 (fp): Disabling this disables f64mm.
3802 (simd): Disabling this disables f64mm.
3803 (fp16): Disabling this disables f64mm.
3804 (sve): Disabling this disables f64mm.
3805 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3806 (AARCH64_ISA_F64MM): New.
3807 (TARGET_F64MM): New.
3808 * doc/invoke.texi (f64mm): Document new option.
3809
3810 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3811
3812 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3813 (neoversen1_tunings): Likewise.
3814
3815 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3816
3817 PR target/92692
3818 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3819 Add assert to ensure prolog has been emitted.
3820 (aarch64_split_atomic_op): Likewise.
3821 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3822 Use epilogue_completed rather than reload_completed.
3823 (aarch64_atomic_exchange<mode>): Likewise.
3824 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3825 (atomic_nand<mode>): Likewise.
3826 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3827 (atomic_fetch_nand<mode>): Likewise.
3828 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3829 (atomic_nand_fetch<mode>): Likewise.
3830
3831 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3832
3833 PR target/93133
3834 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3835 for FP modes.
3836 (REVERSE_CONDITION): Delete.
3837 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3838 (CCFP_CCFPE): Likewise.
3839 (e): New mode attribute.
3840 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3841 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3842 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3843 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3844 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3845 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3846 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3847 name of generator from gen_ccmpdi to gen_ccmpccdi.
3848 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3849 the previous comparison but aren't able to, use the new ccmp_rev
3850 patterns instead.
3851
3852 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3853
3854 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3855 than testing directly for INTEGER_CST.
3856 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3857
3858 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3859
3860 PR tree-optimization/93292
3861 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3862 get_vectype_for_scalar_type returns NULL.
3863
3864 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3865
3866 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3867 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3868
3869 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3870
3871 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3872 dump.
3873 * params.opt: (max-predicted-iterations): Set bounds.
3874 * predict.c (real_almost_one, real_br_prob_base,
3875 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3876 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3877 probabilities; do not truncate to reg_br_prob_bases.
3878 (estimate_loops_at_level): Pass max_cyclic_prob.
3879 (estimate_loops): Compute max_cyclic_prob.
3880 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3881 of back edge prob.
3882 * profile-count.c (profile_probability::to_sreal): New.
3883 * profile-count.h (class sreal): Move up in file.
3884 (profile_probability::to_sreal): Declare.
3885
3886 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3887
3888 * config/arm/arm.c
3889 (arm_invalid_conversion): New function for target hook.
3890 (arm_invalid_unary_op): New function for target hook.
3891 (arm_invalid_binary_op): New function for target hook.
3892
3893 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3894
3895 * config.gcc: Add arm_bf16.h.
3896 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3897 (arm_simd_builtin_std_type): Add BFmode.
3898 (arm_init_simd_builtin_types): Define element types for vector types.
3899 (arm_init_bf16_types): New function.
3900 (arm_init_builtins): Add arm_init_bf16_types function call.
3901 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3902 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3903 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3904 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3905 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3906 (arm_mangle_type): Add __bf16.
3907 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3908 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3909 arm_bf16_ptr_type_node.
3910 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3911 define_split between ARM registers.
3912 * config/arm/arm_bf16.h: New file.
3913 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3914 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3915 (VQXMOV): Add V8BF.
3916 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3917 * config/arm/vfp.md: Add BFmode to movhf patterns.
3918
3919 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3920 Andre Vieira <andre.simoesdiasvieira@arm.com>
3921
3922 * config/arm/arm-cpus.in (mve, mve_float): New features.
3923 (dsp, mve, mve.fp): New options.
3924 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3925 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3926 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3927
3928 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3929 Thomas Preud'homme <thomas.preudhomme@arm.com>
3930
3931 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3932 Armv8-M Mainline.
3933 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3934 error for using -mcmse when targeting Armv8.1-M Mainline.
3935
3936 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3937 Thomas Preud'homme <thomas.preudhomme@arm.com>
3938
3939 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3940 address in r4 when targeting Armv8.1-M Mainline.
3941 (nonsecure_call_value_internal): Likewise.
3942 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3943 a register match_operand again. Emit BLXNS when targeting
3944 Armv8.1-M Mainline.
3945 (nonsecure_call_value_reg_thumb2): Likewise.
3946
3947 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3948 Thomas Preud'homme <thomas.preudhomme@arm.com>
3949
3950 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3951 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3952 variable as true when floating-point ABI is not hard. Replace
3953 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3954 Generate VLSTM and VLLDM instruction respectively before and
3955 after a function call to cmse_nonsecure_call function.
3956 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3957 (VUNSPEC_VLLDM): Likewise.
3958 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3959 (lazy_load_multiple_insn): Likewise.
3960
3961 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3962 Thomas Preud'homme <thomas.preudhomme@arm.com>
3963
3964 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3965 (arm_emit_vfp_multi_reg_pop): Likewise.
3966 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3967 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3968 restore callee-saved VFP registers.
3969
3970 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3971 Thomas Preud'homme <thomas.preudhomme@arm.com>
3972
3973 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3974 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3975 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3976 callee-saved GPRs as well as clear ip register before doing a nonsecure
3977 call then restore callee-saved GPRs after it when targeting
3978 Armv8.1-M Mainline.
3979 (arm_reorg): Adapt to function rename.
3980
3981 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3982 Thomas Preud'homme <thomas.preudhomme@arm.com>
3983
3984 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3985 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3986 clear_vfp_multiple pattern based on a new vfp parameter.
3987 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3988 targeting Armv8.1-M Mainline.
3989 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3990 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3991 whether VFP registers are available before looking call_used_regs for a
3992 VFP register.
3993 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3994 of prototype of clear_operation_p.
3995 (clear_vfp_multiple_operation): New predicate.
3996 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3997 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3998
3999 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4000 Thomas Preud'homme <thomas.preudhomme@arm.com>
4001
4002 * config/arm/arm-protos.h (clear_operation_p): Declare.
4003 * config/arm/arm.c (clear_operation_p): New function.
4004 (cmse_clear_registers): Generate clear_multiple instruction pattern if
4005 targeting Armv8.1-M Mainline or successor.
4006 (output_return_instruction): Only output APSR register clearing if
4007 Armv8.1-M Mainline instructions not available.
4008 (thumb_exit): Likewise.
4009 * config/arm/predicates.md (clear_multiple_operation): New predicate.
4010 * config/arm/thumb2.md (clear_apsr): New define_insn.
4011 (clear_multiple): Likewise.
4012 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
4013
4014 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4015 Thomas Preud'homme <thomas.preudhomme@arm.com>
4016
4017 * config/arm/arm.c (fp_sysreg_names): Declare and define.
4018 (use_return_insn): Also return false for Armv8.1-M Mainline.
4019 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
4020 Mainline instructions are available.
4021 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
4022 when targeting Armv8.1-M Mainline Security Extensions.
4023 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
4024 Mainline entry function.
4025 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
4026 targeting Armv8.1-M Mainline or successor.
4027 (arm_expand_epilogue): Fix indentation of caller-saved register
4028 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
4029 entry function.
4030 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
4031 (FP_SYSREGS): Likewise.
4032 (enum vfp_sysregs_encoding): Define enum.
4033 (fp_sysreg_names): Declare.
4034 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
4035 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
4036 (pop_fpsysreg_insn): Likewise.
4037
4038 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4039 Thomas Preud'homme <thomas.preudhomme@arm.com>
4040
4041 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
4042 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
4043 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
4044 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
4045 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
4046 (ARMv8_1m_main): New feature group.
4047 (armv8.1-m.main): New architecture.
4048 * config/arm/arm-tables.opt: Regenerate.
4049 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
4050 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
4051 (arm_options_perform_arch_sanity_checks): Error out when targeting
4052 Armv8.1-M Mainline Security Extensions.
4053 * config/arm/arm.h (arm_arch8_1m_main): Declare.
4054
4055 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4056
4057 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
4058 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
4059 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
4060 aarch64_bfdot_laneq): New.
4061 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
4062 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
4063 vbfdotq_laneq_f32): New.
4064 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
4065 VBFMLA_W, VBF): New.
4066 (isquadop): Add V4BF, V8BF.
4067
4068 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4069
4070 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
4071 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
4072 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
4073 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
4074 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
4075 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
4076 usdot_laneq, sudot_lane,sudot_laneq): New.
4077 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
4078 (aarch64_<sur>dot_lane): New.
4079 * config/aarch64/arm_neon.h (vusdot_s32): New.
4080 (vusdotq_s32): New.
4081 (vusdot_lane_s32): New.
4082 (vsudot_lane_s32): New.
4083 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
4084 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
4085
4086 2020-01-16 Martin Liska <mliska@suse.cz>
4087
4088 * value-prof.c (dump_histogram_value): Fix
4089 obvious spacing issue.
4090
4091 2020-01-16 Andrew Pinski <apinski@marvell.com>
4092
4093 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
4094 !storage_order_barrier_p.
4095
4096 2020-01-16 Andrew Pinski <apinski@marvell.com>
4097
4098 * sched-int.h (_dep): Add unused bit-field field for the padding.
4099 * sched-deps.c (init_dep_1): Init unused field.
4100
4101 2020-01-16 Andrew Pinski <apinski@marvell.com>
4102
4103 * optabs.h (create_expand_operand): Initialize target field also.
4104
4105 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
4106
4107 PR tree-optimization/92429
4108 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
4109 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
4110 control folding.
4111 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
4112 tree.
4113
4114 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
4115
4116 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
4117 aarch64_sve_int_mode to each mode.
4118
4119 2020-01-15 David Malcolm <dmalcolm@redhat.com>
4120
4121 * doc/analyzer.texi (Overview): Add note about
4122 -fdump-ipa-analyzer.
4123
4124 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
4125
4126 PR tree-optimization/93231
4127 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
4128 input_type is unsigned. Use tree_to_shwi for shift constant.
4129 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
4130 (simplify_count_trailing_zeroes): Add test to handle known non-zero
4131 inputs more efficiently.
4132
4133 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
4134
4135 * config/i386/i386.md (*movsf_internal): Do not require
4136 SSE2 ISA for alternatives 14 and 15.
4137
4138 2020-01-15 Richard Biener <rguenther@suse.de>
4139
4140 PR middle-end/93273
4141 * tree-eh.c (sink_clobbers): If we already visited the destination
4142 block do not defer insertion.
4143 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
4144 the purpose of defered insertion.
4145
4146 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4147
4148 * BASE-VER: Bump to 10.0.1.
4149
4150 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
4151
4152 PR tree-optimization/93247
4153 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
4154 type of the stmt that we're going to vectorize.
4155
4156 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
4157
4158 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
4159 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
4160 type from the lhs.
4161
4162 2020-01-15 Martin Liska <mliska@suse.cz>
4163
4164 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
4165 2 calls of streamer_read_hwi in a function call.
4166
4167 2020-01-15 Richard Biener <rguenther@suse.de>
4168
4169 * alias.c (record_alias_subset): Avoid redundant work when
4170 subset is already recorded.
4171
4172 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4173
4174 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
4175 the analyzer options provide CWE identifiers.
4176
4177 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4178
4179 * tree-diagnostic-path.cc (path_summary::event_range::print):
4180 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
4181 using get_pure_location.
4182
4183 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4184
4185 PR tree-optimization/93262
4186 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
4187 perform head trimming only if the last argument is constant,
4188 either all ones, or larger or equal to head trim, in the latter
4189 case decrease the last argument by head_trim.
4190
4191 PR tree-optimization/93249
4192 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
4193 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
4194 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
4195 perform head trim unless we can prove there are no '\0' chars
4196 from the source among the first head_trim chars.
4197
4198 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4199
4200 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
4201
4202 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4203
4204 PR target/93009
4205 * config/i386/sse.md
4206 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
4207 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
4208 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
4209 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
4210 just a single alternative instead of two, make operands 1 and 2
4211 commutative.
4212
4213 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
4214
4215 PR lto/91576
4216 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
4217 TYPE_MODE.
4218
4219 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4220
4221 * Makefile.in (lang_opt_files): Add analyzer.opt.
4222 (ANALYZER_OBJS): New.
4223 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
4224 tristate.o and ANALYZER_OBJS.
4225 (TEXI_GCCINT_FILES): Add analyzer.texi.
4226 * common.opt (-fanalyzer): New driver option.
4227 * config.in: Regenerate.
4228 * configure: Regenerate.
4229 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
4230 (gccdepdir): Also create depdir for "analyzer" subdir.
4231 * digraph.cc: New file.
4232 * digraph.h: New file.
4233 * doc/analyzer.texi: New file.
4234 * doc/gccint.texi ("Static Analyzer") New menu item.
4235 (analyzer.texi): Include it.
4236 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
4237 ("Warning Options"): Add static analysis warnings to the list.
4238 (-Wno-analyzer-double-fclose): New option.
4239 (-Wno-analyzer-double-free): New option.
4240 (-Wno-analyzer-exposure-through-output-file): New option.
4241 (-Wno-analyzer-file-leak): New option.
4242 (-Wno-analyzer-free-of-non-heap): New option.
4243 (-Wno-analyzer-malloc-leak): New option.
4244 (-Wno-analyzer-possible-null-argument): New option.
4245 (-Wno-analyzer-possible-null-dereference): New option.
4246 (-Wno-analyzer-null-argument): New option.
4247 (-Wno-analyzer-null-dereference): New option.
4248 (-Wno-analyzer-stale-setjmp-buffer): New option.
4249 (-Wno-analyzer-tainted-array-index): New option.
4250 (-Wno-analyzer-use-after-free): New option.
4251 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
4252 (-Wno-analyzer-use-of-uninitialized-value): New option.
4253 (-Wanalyzer-too-complex): New option.
4254 (-fanalyzer-call-summaries): New warning.
4255 (-fanalyzer-checker=): New warning.
4256 (-fanalyzer-fine-grained): New warning.
4257 (-fno-analyzer-state-merge): New warning.
4258 (-fno-analyzer-state-purge): New warning.
4259 (-fanalyzer-transitivity): New warning.
4260 (-fanalyzer-verbose-edges): New warning.
4261 (-fanalyzer-verbose-state-changes): New warning.
4262 (-fanalyzer-verbosity=): New warning.
4263 (-fdump-analyzer): New warning.
4264 (-fdump-analyzer-callgraph): New warning.
4265 (-fdump-analyzer-exploded-graph): New warning.
4266 (-fdump-analyzer-exploded-nodes): New warning.
4267 (-fdump-analyzer-exploded-nodes-2): New warning.
4268 (-fdump-analyzer-exploded-nodes-3): New warning.
4269 (-fdump-analyzer-supergraph): New warning.
4270 * doc/sourcebuild.texi (dg-require-dot): New.
4271 (dg-check-dot): New.
4272 * gdbinit.in (break-on-saved-diagnostic): New command.
4273 * graphviz.cc: New file.
4274 * graphviz.h: New file.
4275 * ordered-hash-map-tests.cc: New file.
4276 * ordered-hash-map.h: New file.
4277 * passes.def (pass_analyzer): Add before
4278 pass_ipa_whole_program_visibility.
4279 * selftest-run-tests.c (selftest::run_tests): Call
4280 selftest::ordered_hash_map_tests_cc_tests.
4281 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
4282 decl.
4283 * shortest-paths.h: New file.
4284 * timevar.def (TV_ANALYZER): New timevar.
4285 (TV_ANALYZER_SUPERGRAPH): Likewise.
4286 (TV_ANALYZER_STATE_PURGE): Likewise.
4287 (TV_ANALYZER_PLAN): Likewise.
4288 (TV_ANALYZER_SCC): Likewise.
4289 (TV_ANALYZER_WORKLIST): Likewise.
4290 (TV_ANALYZER_DUMP): Likewise.
4291 (TV_ANALYZER_DIAGNOSTICS): Likewise.
4292 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
4293 * tree-pass.h (make_pass_analyzer): New decl.
4294 * tristate.cc: New file.
4295 * tristate.h: New file.
4296
4297 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
4298
4299 PR target/93254
4300 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
4301 alternatives 9 and 10.
4302
4303 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4304
4305 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
4306 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
4307 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
4308 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
4309 (selftest::hash_map_tests_c_tests): Call it.
4310 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
4311 New static constant, using the value of = H::empty_zero_p.
4312 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
4313 from default_hash_traits <Value>.
4314 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
4315 from Traits.
4316 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
4317 * hash-table.h (hash_table::alloc_entries): Guard the loop of
4318 calls to mark_empty with !Descriptor::empty_zero_p.
4319 (hash_table::empty_slow): Conditionalize the memset call with a
4320 check that Descriptor::empty_zero_p; otherwise, loop through the
4321 entries calling mark_empty on them.
4322 * hash-traits.h (int_hash::empty_zero_p): New static constant.
4323 (pointer_hash::empty_zero_p): Likewise.
4324 (pair_hash::empty_zero_p): Likewise.
4325 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
4326 Likewise.
4327 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
4328 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
4329 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
4330 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
4331 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
4332 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
4333 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
4334 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
4335 * tree-vectorizer.h
4336 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
4337 Likewise.
4338
4339 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
4340
4341 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
4342 fix typo on return value.
4343
4344 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
4345
4346 PR ipa/69678
4347 * cgraph.c (symbol_table::create_edge): Init speculative_id and
4348 target_prob.
4349 (cgraph_edge::make_speculative): Add param for setting speculative_id
4350 and target_prob.
4351 (cgraph_edge::speculative_call_info): Update comments and find reference
4352 by speculative_id for multiple indirect targets.
4353 (cgraph_edge::resolve_speculation): Decrease the speculations
4354 for indirect edge, drop it's speculative if not direct target
4355 left. Update comments.
4356 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4357 (cgraph_node::dump): Print num_speculative_call_targets.
4358 (cgraph_node::verify_node): Don't report error if speculative
4359 edge not include statement.
4360 (cgraph_edge::num_speculative_call_targets_p): New function.
4361 * cgraph.h (int common_target_id): Remove.
4362 (int common_target_probability): Remove.
4363 (num_speculative_call_targets): New variable.
4364 (make_speculative): Add param for setting speculative_id.
4365 (cgraph_edge::num_speculative_call_targets_p): New declare.
4366 (target_prob): New variable.
4367 (speculative_id): New variable.
4368 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
4369 call summaries for multiple speculative call targets.
4370 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
4371 * ipa-profile.c (struct speculative_call_target): New struct.
4372 (class speculative_call_summary): New class.
4373 (class speculative_call_summaries): New class.
4374 (call_sums): New variable.
4375 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4376 (ipa_profile_write_edge_summary): New function.
4377 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4378 (ipa_profile_dump_all_summaries): New function.
4379 (ipa_profile_read_edge_summary): New function.
4380 (ipa_profile_read_summary_section): New function.
4381 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4382 (ipa_profile): Generate num_speculative_call_targets from
4383 profile summaries.
4384 * ipa-ref.h (speculative_id): New variable.
4385 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4386 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4387 common_target_probability. Stream out speculative_id and
4388 num_speculative_call_targets.
4389 (input_edge): Likewise.
4390 * predict.c (dump_prediction): Remove edges count assert to be
4391 precise.
4392 * symtab.c (symtab_node::create_reference): Init speculative_id.
4393 (symtab_node::clone_references): Clone speculative_id.
4394 (symtab_node::clone_referring): Clone speculative_id.
4395 (symtab_node::clone_reference): Clone speculative_id.
4396 (symtab_node::clear_stmts_in_references): Clear speculative_id.
4397 * tree-inline.c (copy_bb): Duplicate all the speculative edges
4398 if indirect call contains multiple speculative targets.
4399 * value-prof.h (check_ic_target): Remove.
4400 * value-prof.c (gimple_value_profile_transformations):
4401 Use void function gimple_ic_transform.
4402 * value-prof.c (gimple_ic_transform): Handle topn case.
4403 Fix comment typos. Change it to a void function.
4404
4405 2020-01-13 Andrew Pinski <apinski@marvell.com>
4406
4407 * config/aarch64/aarch64-cores.def (octeontx2): New define.
4408 (octeontx2t98): New define.
4409 (octeontx2t96): New define.
4410 (octeontx2t93): New define.
4411 (octeontx2f95): New define.
4412 (octeontx2f95n): New define.
4413 (octeontx2f95mm): New define.
4414 * config/aarch64/aarch64-tune.md: Regenerate.
4415 * doc/invoke.texi (-mcpu=): Document the new cpu types.
4416
4417 2020-01-13 Jason Merrill <jason@redhat.com>
4418
4419 PR c++/33799 - destroy return value if local cleanup throws.
4420 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4421
4422 2020-01-13 Martin Liska <mliska@suse.cz>
4423
4424 * ipa-cp.c (get_max_overall_size): Use newly
4425 renamed param param_ipa_cp_unit_growth.
4426 * params.opt: Remove legacy param name.
4427
4428 2020-01-13 Martin Sebor <msebor@redhat.com>
4429
4430 PR tree-optimization/93213
4431 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4432 stores to be eliminated.
4433
4434 2020-01-13 Martin Liska <mliska@suse.cz>
4435
4436 * opts.c (print_help): Do not print CL_PARAM
4437 and CL_WARNING for CL_OPTIMIZATION.
4438
4439 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4440
4441 PR driver/92757
4442 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4443 depending on optimization settings.
4444
4445 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4446
4447 PR tree-optimization/90838
4448 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4449 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4450 argument rather than to initialize temporary for targets that
4451 don't use the mode argument at all. Initialize ctzval to avoid
4452 warning at -O0.
4453
4454 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4455
4456 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4457 * tree-core.h: Document it.
4458 * gimplify.c (gimplify_omp_workshare): Set it.
4459 * omp-low.c (lower_omp_target): Use it.
4460 * tree-pretty-print.c (dump_omp_clause): Print it.
4461
4462 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4463 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4464
4465 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4466
4467 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4468 * common.opt (fdiagnostics-path-format=): New option.
4469 (diagnostic_path_format): New enum.
4470 (fdiagnostics-show-path-depths): New option.
4471 * coretypes.h (diagnostic_event_id_t): New forward decl.
4472 * diagnostic-color.c (color_dict): Add "path".
4473 * diagnostic-event-id.h: New file.
4474 * diagnostic-format-json.cc (json_from_expanded_location): Make
4475 non-static.
4476 (json_end_diagnostic): Call context->make_json_for_path if it
4477 exists and the diagnostic has a path.
4478 (diagnostic_output_format_init): Clear context->print_path.
4479 * diagnostic-path.h: New file.
4480 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4481 when printing a run of events in a diagnostic_path so that they
4482 all get the same color.
4483 (layout::m_diagnostic_path_p): New field.
4484 (layout::layout): Initialize it.
4485 (layout::print_any_labels): Don't colorize the label text for an
4486 event in a diagnostic_path.
4487 (gcc_rich_location::add_location_if_nearby): Add
4488 "restrict_to_current_line_spans" and "label" params. Pass the
4489 former to layout.maybe_add_location_range; pass the latter
4490 when calling add_range.
4491 * diagnostic.c: Include "diagnostic-path.h".
4492 (diagnostic_initialize): Initialize context->path_format and
4493 context->show_path_depths.
4494 (diagnostic_show_any_path): New function.
4495 (diagnostic_path::interprocedural_p): New function.
4496 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4497 (simple_diagnostic_path::num_events): New function.
4498 (simple_diagnostic_path::get_event): New function.
4499 (simple_diagnostic_path::add_event): New function.
4500 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4501 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4502 (debug): New overload taking a diagnostic_path *.
4503 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4504 * diagnostic.h (enum diagnostic_path_format): New enum.
4505 (json::value): New forward decl.
4506 (diagnostic_context::path_format): New field.
4507 (diagnostic_context::show_path_depths): New field.
4508 (diagnostic_context::print_path): New callback field.
4509 (diagnostic_context::make_json_for_path): New callback field.
4510 (diagnostic_show_any_path): New decl.
4511 (json_from_expanded_location): New decl.
4512 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4513 (-fdiagnostics-show-path-depths): New option.
4514 (-fdiagnostics-color): Add "path" to description of default
4515 GCC_COLORS; describe it.
4516 (-fdiagnostics-format=json): Document how diagnostic paths are
4517 represented in the JSON output format.
4518 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4519 Add optional params "restrict_to_current_line_spans" and "label".
4520 * opts.c (common_handle_option): Handle
4521 OPT_fdiagnostics_path_format_ and
4522 OPT_fdiagnostics_show_path_depths.
4523 * pretty-print.c: Include "diagnostic-event-id.h".
4524 (pp_format): Implement "%@" format code for printing
4525 diagnostic_event_id_t *.
4526 (selftest::test_pp_format): Add tests for "%@".
4527 * selftest-run-tests.c (selftest::run_tests): Call
4528 selftest::tree_diagnostic_path_cc_tests.
4529 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4530 * toplev.c (general_init): Initialize global_dc->path_format and
4531 global_dc->show_path_depths.
4532 * tree-diagnostic-path.cc: New file.
4533 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4534 non-static. Drop "diagnostic" param in favor of storing the
4535 original value of "where" and re-using it.
4536 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4537 maybe_unwind_expanded_macro_loc.
4538 (tree_diagnostics_defaults): Initialize context->print_path and
4539 context->make_json_for_path.
4540 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4541 decl.
4542 (default_tree_make_json_for_path): New decl.
4543 (maybe_unwind_expanded_macro_loc): New decl.
4544
4545 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4546
4547 PR tree-optimization/93210
4548 * fold-const.h (native_encode_initializer,
4549 can_native_interpret_type_p): Declare.
4550 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4551 simplify.
4552 (native_encode_initializer): New function, moved from dwarf2out.c.
4553 Adjust to native_encode_expr compatible arguments, including dry-run
4554 and partial extraction modes. Don't handle STRING_CST.
4555 (can_native_interpret_type_p): No longer static.
4556 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4557 offset / BITS_PER_UNIT fits into int and don't call it if
4558 can_native_interpret_type_p fails. If suboff is NULL and for
4559 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4560 native_encode_initializer.
4561 (fold_const_aggregate_ref_1): Formatting fix.
4562 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4563 (tree_add_const_value_attribute): Adjust caller.
4564
4565 PR tree-optimization/90838
4566 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4567 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4568 CTZ_DEFINED_VALUE_AT_ZERO.
4569
4570 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4571
4572 PR inline-asm/93027
4573 * lra-constraints.c (match_reload): Permit input operands have the
4574 same mode as output while other input operands have a different
4575 mode.
4576
4577 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4578
4579 PR tree-optimization/90838
4580 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4581 (check_ctz_string): Likewise.
4582 (optimize_count_trailing_zeroes): Likewise.
4583 (simplify_count_trailing_zeroes): Likewise.
4584 (pass_forwprop::execute): Try ctz simplification.
4585 * match.pd: Add matching for ctz idioms.
4586
4587 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4588
4589 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4590 for target hook.
4591 (aarch64_invalid_unary_op): New function for target hook.
4592 (aarch64_invalid_binary_op): New function for target hook.
4593
4594 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4595
4596 * config.gcc: Add arm_bf16.h.
4597 * config/aarch64/aarch64-builtins.c
4598 (aarch64_simd_builtin_std_type): Add BFmode.
4599 (aarch64_init_simd_builtin_types): Define element types for vector
4600 types.
4601 (aarch64_init_bf16_types): New function.
4602 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4603 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4604 modes.
4605 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4606 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4607 patterns.
4608 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4609 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4610 * config/aarch64/aarch64.c
4611 (aarch64_classify_vector_mode): Add support for BF types.
4612 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4613 (aarch64_vq_mode): Add support for BF types.
4614 (aarch64_simd_container_mode): Add support for BF types.
4615 (aarch64_mangle_type): Add support for BF scalar type.
4616 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4617 * config/aarch64/arm_bf16.h: New file.
4618 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4619 * config/aarch64/iterators.md: Add BF types to mode attributes.
4620 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4621
4622 2020-01-10 Jason Merrill <jason@redhat.com>
4623
4624 PR c++/93173 - incorrect tree sharing.
4625 * gimplify.c (copy_if_shared): No longer static.
4626 * gimplify.h: Declare it.
4627
4628 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4629
4630 * doc/invoke.texi (-msve-vector-bits=): Document that
4631 -msve-vector-bits=128 now generates VL-specific code for
4632 little-endian targets.
4633 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4634 build_vector_type_for_mode to construct the data vector types.
4635 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4636 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4637 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4638 for 128-bit vectors.
4639
4640 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4641
4642 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4643 invocation.
4644
4645 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4646
4647 * config/aarch64/aarch64-builtins.c
4648 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4649 rather than checking the number of elements and the element mode.
4650
4651 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4652
4653 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4654 get_related_vectype_for_scalar_type rather than build_vector_type
4655 to create the index type for a conditional reduction.
4656
4657 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4658
4659 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4660 for any type of gather or scatter, including strided accesses.
4661
4662 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4663
4664 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4665 comment.
4666
4667 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4668
4669 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4670 get_dr_vinfo_offset
4671 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4672 parameter and its use to reset DR_OFFSET's.
4673 (vect_transform_loop): Remove orig_drs_init argument.
4674 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4675 member of dr_vec_info rather than the offset of the associated
4676 data_reference's innermost_loop_behavior.
4677 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4678 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4679 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4680 get_dr_vinfo_offset.
4681 (vectorizable_store): Likewise.
4682 (vectorizable_load): Likewise.
4683
4684 2020-01-10 Richard Biener <rguenther@suse.de>
4685
4686 * gimple-ssa-store-merging
4687 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4688
4689 2020-01-10 Martin Liska <mliska@suse.cz>
4690
4691 PR ipa/93217
4692 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4693 encapsulation that was there before r280040.
4694
4695 2020-01-10 Richard Biener <rguenther@suse.de>
4696
4697 PR middle-end/93199
4698 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4699 sequences to avoid walking them again for secondary opportunities.
4700 (pass_lower_eh_dispatch::execute): Instead actually insert
4701 them here.
4702
4703 2020-01-10 Richard Biener <rguenther@suse.de>
4704
4705 PR middle-end/93199
4706 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4707 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4708 avoid quadraticness.
4709
4710 2020-01-10 Martin Jambor <mjambor@suse.cz>
4711
4712 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4713 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4714 to get param_ipa_sra_max_replacements.
4715 (param_splitting_across_edge): Pass the caller to
4716 pull_accesses_from_callee.
4717
4718 2020-01-10 Martin Jambor <mjambor@suse.cz>
4719
4720 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4721 * ipa-cp.c (max_new_size): Removed.
4722 (orig_overall_size): New variable.
4723 (get_max_overall_size): New function.
4724 (estimate_local_effects): Use it. Adjust dump.
4725 (decide_about_value): Likewise.
4726 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4727 orig_overall_size. Adjust dump.
4728 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4729
4730 2020-01-10 Martin Jambor <mjambor@suse.cz>
4731
4732 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4733 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4734 instead of param_ipa_max_agg_items.
4735 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4736 optimization info for the callee.
4737
4738 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4739
4740 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4741 markers if debug_inline_points is false.
4742
4743 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4744
4745 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4746 extra_objs.
4747 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4748 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4749 aarch64-sve-builtins-sve2.h.
4750 (aarch64-sve-builtins-sve2.o): New rule.
4751 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4752 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4753 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4754 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4755 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4756 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4757 TARGET_SVE2_SM4.
4758 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4759 instructions that are handled here.
4760 (@cond_asrd<mode>): Generalize to...
4761 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4762 (*cond_asrd<mode>_2): Generalize to...
4763 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4764 (*cond_asrd<mode>_z): Generalize to...
4765 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4766 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4767 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4768 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4769 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4770 pattern.
4771 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4772 (@aarch64_scatter_stnt<mode>): Likewise.
4773 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4774 (@aarch64_mul_lane_<mode>): Likewise.
4775 (@aarch64_sve_suqadd<mode>_const): Likewise.
4776 (*<sur>h<addsub><mode>): Generalize to...
4777 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4778 new pattern.
4779 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4780 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4781 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4782 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4783 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4784 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4785 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4786 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4787 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4788 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4789 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4790 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4791 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4792 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4793 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4794 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4795 (@aarch64_sve2_xar<mode>): Likewise.
4796 (@aarch64_sve2_bcax<mode>): Likewise.
4797 (*aarch64_sve2_eor3<mode>): Rename to...
4798 (@aarch64_sve2_eor3<mode>): ...this.
4799 (@aarch64_sve2_bsl<mode>): New expander.
4800 (@aarch64_sve2_nbsl<mode>): Likewise.
4801 (@aarch64_sve2_bsl1n<mode>): Likewise.
4802 (@aarch64_sve2_bsl2n<mode>): Likewise.
4803 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4804 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4805 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4806 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4807 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4808 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4809 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4810 (<su>mull<bt><Vwide>): Generalize to...
4811 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4812 pattern.
4813 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4814 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4815 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4816 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4817 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4818 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4819 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4820 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4821 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4822 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4823 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4824 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4825 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4826 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4827 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4828 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4829 (<SHRNB:r>shrnb<mode>): Generalize to...
4830 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4831 new pattern.
4832 (<SHRNT:r>shrnt<mode>): Generalize to...
4833 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4834 new pattern.
4835 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4836 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4837 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4838 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4839 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4840 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4841 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4842 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4843 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4844 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4845 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4846 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4847 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4848 (@aarch64_sve2_cvtnt<mode>): Likewise.
4849 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4850 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4851 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4852 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4853 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4854 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4855 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4856 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4857 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4858 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4859 (@aarch64_sve2_pmul<mode>): Likewise.
4860 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4861 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4862 (@aarch64_sve2_tbl2<mode>): Likewise.
4863 (@aarch64_sve2_tbx<mode>): Likewise.
4864 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4865 (@aarch64_sve2_histcnt<mode>): Likewise.
4866 (@aarch64_sve2_histseg<mode>): Likewise.
4867 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4868 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4869 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4870 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4871 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4872 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4873 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4874 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4875 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4876 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4877 (SVE2_PMULL_PAIR_I): New mode iterators.
4878 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4879 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4880 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4881 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4882 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4883 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4884 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4885 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4886 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4887 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4888 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4889 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4890 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4891 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4892 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4893 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4894 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4895 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4896 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4897 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4898 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4899 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4900 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4901 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4902 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4903 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4904 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4905 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4906 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4907 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4908 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4909 further down file.
4910 (VNARROW, Ventype): New mode attributes.
4911 (Vewtype): Handle VNx2DI. Fix typo in comment.
4912 (VDOUBLE): New mode attribute.
4913 (sve_lane_con): Handle VNx8HI.
4914 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4915 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4916 (sve_int_op, sve_int_op_rev): Handle the above codes.
4917 (sve_pred_int_rhs2_operand): Likewise.
4918 (MULLBT, SHRNB, SHRNT): Delete.
4919 (SVE_INT_SHIFT_IMM): New int iterator.
4920 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4921 and UNSPEC_WHILEHS for TARGET_SVE2.
4922 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4923 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4924 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4925 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4926 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4927 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4928 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4929 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4930 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4931 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4932 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4933 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4934 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4935 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4936 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4937 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4938 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4939 (optab): Handle the new unspecs.
4940 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4941 and UNSPEC_RSHRNT.
4942 (lr): Handle the new unspecs.
4943 (bt): Delete.
4944 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4945 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4946 (sve_int_qsub_op): New int attributes.
4947 (sve_fp_op, rot): Handle the new unspecs.
4948 * config/aarch64/aarch64-sve-builtins.h
4949 (function_resolver::require_matching_pointer_type): Declare.
4950 (function_resolver::resolve_unary): Add an optional boolean argument.
4951 (function_resolver::finish_opt_n_resolution): Add an optional
4952 type_suffix_index argument.
4953 (gimple_folder::redirect_call): Declare.
4954 (gimple_expander::prepare_gather_address_operands): Add an optional
4955 bool parameter.
4956 * config/aarch64/aarch64-sve-builtins.cc: Include
4957 aarch64-sve-builtins-sve2.h.
4958 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4959 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4960 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4961 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4962 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4963 (TYPES_s_unsigned): Likewise.
4964 (TYPES_s_integer): Use TYPES_s_unsigned.
4965 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4966 (TYPES_sd_integer): Use them.
4967 (TYPES_d_unsigned): New macro.
4968 (TYPES_d_integer): Use it.
4969 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4970 (TYPES_cvt_narrow): Likewise.
4971 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4972 (preds_mx): New variable.
4973 (function_builder::add_overloaded_function): Allow the new feature
4974 set to be more restrictive than the original one.
4975 (function_resolver::infer_pointer_type): Remove qualifiers from
4976 the pointer type before printing it.
4977 (function_resolver::require_matching_pointer_type): New function.
4978 (function_resolver::resolve_sv_displacement): Handle functions
4979 that don't support 32-bit vector indices or svint32_t vector offsets.
4980 (function_resolver::finish_opt_n_resolution): Take the inferred type
4981 as a separate argument.
4982 (function_resolver::resolve_unary): Optionally treat all forms in
4983 the same way as normal merging functions.
4984 (gimple_folder::redirect_call): New function.
4985 (function_expander::prepare_gather_address_operands): Add an argument
4986 that says whether scaled forms are available. If they aren't,
4987 handle scaling of vector indices and don't add the extension and
4988 scaling operands.
4989 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4990 fall back to using cond_* instead.
4991 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4992 Split out the member variables into...
4993 (rtx_code_function_base): ...this new base class.
4994 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4995 (unspec_based_function): Split out the member variables into...
4996 (unspec_based_function_base): ...this new base class.
4997 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4998 (unspec_based_function_exact_insn): New class.
4999 (unspec_based_add_function, unspec_based_add_lane_function)
5000 (unspec_based_lane_function, unspec_based_pred_function)
5001 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
5002 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
5003 (unspec_based_sub_function, unspec_based_sub_lane_function): New
5004 typedefs.
5005 (unspec_based_fused_function): New class.
5006 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
5007 (unspec_based_fused_lane_function): New class.
5008 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
5009 typedefs.
5010 (CODE_FOR_MODE1): New macro.
5011 (fixed_insn_function): New class.
5012 (while_comparison): Likewise.
5013 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
5014 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
5015 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
5016 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
5017 (load_gather_sv_restricted, shift_left_imm_long): Declare.
5018 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
5019 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
5020 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
5021 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
5022 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
5023 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
5024 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
5025 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
5026 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
5027 Also add an initial argument for unary_convert_narrowt, regardless
5028 of the predication type.
5029 (build_32_64): Allow loads and stores to specify MODE_none.
5030 (build_sv_index64, build_sv_uint_offset): New functions.
5031 (long_type_suffix): New function.
5032 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
5033 (binary_imm_long_base, load_gather_sv_base): Likewise.
5034 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
5035 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
5036 (unary_narrowb_base, unary_narrowt_base): Likewise.
5037 (binary_long_lane_def, binary_long_lane): New shape.
5038 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
5039 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
5040 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
5041 (binary_to_uint_def, binary_to_uint): Likewise.
5042 (binary_wide_def, binary_wide): Likewise.
5043 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
5044 (compare_def, compare): Likewise.
5045 (compare_ptr_def, compare_ptr): Likewise.
5046 (load_ext_gather_index_restricted_def,
5047 load_ext_gather_index_restricted): Likewise.
5048 (load_ext_gather_offset_restricted_def,
5049 load_ext_gather_offset_restricted): Likewise.
5050 (load_gather_sv_def): Inherit from load_gather_sv_base.
5051 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
5052 (shift_left_imm_def, shift_left_imm): Likewise.
5053 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
5054 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
5055 (store_scatter_index_restricted_def,
5056 store_scatter_index_restricted): Likewise.
5057 (store_scatter_offset_restricted_def,
5058 store_scatter_offset_restricted): Likewise.
5059 (tbl_tuple_def, tbl_tuple): Likewise.
5060 (ternary_long_lane_def, ternary_long_lane): Likewise.
5061 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
5062 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
5063 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
5064 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
5065 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
5066 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
5067 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
5068 (ternary_uint_def, ternary_uint): Likewise.
5069 (unary_convert): Fix typo in comment.
5070 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
5071 (unary_long_def, unary_long): Likewise.
5072 (unary_narrowb_def, unary_narrowb): Likewise.
5073 (unary_narrowt_def, unary_narrowt): Likewise.
5074 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
5075 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
5076 (unary_to_int_def, unary_to_int): Likewise.
5077 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
5078 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
5079 (svasrd_impl): Delete.
5080 (svcadd_impl::expand): Handle integer operations too.
5081 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
5082 new functions to derive the unspec numbers.
5083 (svmla_svmls_lane_impl): Replace with...
5084 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
5085 integer operations too.
5086 (svwhile_impl): Rename to...
5087 (svwhilelx_impl): ...this and inherit from while_comparison.
5088 (svasrd): Use unspec_based_function.
5089 (svmla_lane): Use svmla_lane_impl.
5090 (svmls_lane): Use svmls_lane_impl.
5091 (svrecpe, svrsqrte): Handle unsigned integer operations too.
5092 (svwhilele, svwhilelt): Use svwhilelx_impl.
5093 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
5094 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
5095 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
5096 * config/aarch64/aarch64-sve-builtins.def: Include
5097 aarch64-sve-builtins-sve2.def.
5098
5099 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5100
5101 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
5102 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
5103 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
5104 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
5105 immediates as well as vector ones.
5106 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
5107 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
5108 (aarch64_sve_qsub_immediate): Update calls accordingly.
5109
5110 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5111
5112 * config/aarch64/aarch64-sve2.md: Add banner comments.
5113 (<su>mulh<r>s<mode>3): Move further up file.
5114 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
5115 (*aarch64_sve2_sra<mode>): Move further down file.
5116 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
5117
5118 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5119
5120 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
5121 and UNSPEC_WHILEWR.
5122 (while_optab_cmp): Handle them.
5123 * config/aarch64/aarch64-sve.md
5124 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
5125 and add a "@" marker.
5126 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
5127 instead of gen_aarch64_sve2_while_ptest.
5128 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
5129
5130 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5131
5132 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
5133 (UNSPEC_WHILELE): ...this.
5134 (UNSPEC_WHILE_LO): Rename to...
5135 (UNSPEC_WHILELO): ...this.
5136 (UNSPEC_WHILE_LS): Rename to...
5137 (UNSPEC_WHILELS): ...this.
5138 (UNSPEC_WHILE_LT): Rename to...
5139 (UNSPEC_WHILELT): ...this.
5140 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
5141 (cmp_op, while_optab_cmp): Likewise.
5142 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
5143 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
5144 (svwhilelt): Likewise.
5145
5146 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5147
5148 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
5149 (unary_to_uint): Define.
5150 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
5151 (unary_count): Rename to...
5152 (unary_to_uint_def, unary_to_uint): ...this.
5153 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
5154
5155 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5156
5157 * config/aarch64/aarch64-sve-builtins-functions.h
5158 (code_for_mode_function): New class.
5159 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
5160 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
5161 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
5162 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
5163 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
5164
5165 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5166
5167 * config/aarch64/iterators.md (addsub): New code attribute.
5168 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
5169 Re-express as...
5170 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
5171 in the asm string and attributes. Fix indentation.
5172 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
5173 Re-express as...
5174 (@aarch64_sve_<optab><mode>): ...this.
5175 * config/aarch64/aarch64-sve-builtins.h
5176 (function_expander::expand_signed_unpred_op): Delete.
5177 * config/aarch64/aarch64-sve-builtins.cc
5178 (function_expander::expand_signed_unpred_op): Likewise.
5179 (function_expander::map_to_rtx_codes): If the optab isn't defined,
5180 try using code_for_aarch64_sve instead.
5181 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
5182 (svqsub_impl): Likewise.
5183 (svqadd, svqsub): Use rtx_code_function instead.
5184
5185 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5186
5187 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
5188 (HADDSUB, sur, addsub): Remove them.
5189
5190 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5191
5192 * tree-nrv.c (pass_return_slot::execute): Handle all internal
5193 functions the same way, rather than singling out those that
5194 aren't mapped directly to optabs.
5195
5196 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5197
5198 * target.def (compatible_vector_types_p): New target hook.
5199 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
5200 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
5201 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
5202 * doc/tm.texi: Regenerate.
5203 * gimple-expr.c: Include target.h.
5204 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
5205 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
5206 function.
5207 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
5208 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
5209 Use the original predicate if it already has a suitable type.
5210
5211 2020-01-09 Martin Jambor <mjambor@suse.cz>
5212
5213 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
5214 resolve_speculation and redirect_call_stmt_to_callee static. Change
5215 return type of set_call_stmt to cgraph_edge *.
5216 * auto-profile.c (afdo_indirect_call): Adjust call to
5217 redirect_call_stmt_to_callee.
5218 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
5219 make the this pointer explicit, adjust self-recursive calls and the
5220 call top make_direct. Return the resulting edge.
5221 (cgraph_edge::remove): Make this pointer explicit.
5222 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
5223 (cgraph_edge::make_direct): Likewise, adjust call to
5224 resolve_speculation.
5225 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
5226 call to set_call_stmt.
5227 (cgraph_update_edges_for_call_stmt_node): Update call to
5228 set_call_stmt and remove.
5229 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
5230 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
5231 (cgraph_node::create_edge_including_clones): Moved "first" definition
5232 of edge to the block where it was used. Adjusted calls to
5233 set_call_stmt.
5234 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
5235 cgraph_edge::remove.
5236 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
5237 make_direct and redirect_call_stmt_to_callee.
5238 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
5239 resolve_speculation and make_direct.
5240 * ipa-inline-transform.c (inline_transform): Adjust call to
5241 redirect_call_stmt_to_callee.
5242 (check_speculations_1):: Adjust call to resolve_speculation.
5243 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
5244 resolve-speculation.
5245 (inline_small_functions): Adjust call to resolve_speculation.
5246 (ipa_inline): Likewise.
5247 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
5248 make_direct.
5249 * ipa-visibility.c (function_and_variable_visibility): Make iteration
5250 safe with regards to edge removal, adjust calls to
5251 redirect_call_stmt_to_callee.
5252 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
5253 and redirect_call_stmt_to_callee.
5254 * multiple_target.c (create_dispatcher_calls): Adjust call to
5255 redirect_call_stmt_to_callee
5256 (redirect_to_specific_clone): Likewise.
5257 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
5258 Adjust calls to cgraph_edge::remove.
5259 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
5260 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
5261 (expand_call_inline): Adjust call to cgraph_edge::remove.
5262
5263 2020-01-09 Martin Liska <mliska@suse.cz>
5264
5265 * params.opt: Set Optimization for
5266 param_max_speculative_devirt_maydefs.
5267
5268 2020-01-09 Martin Sebor <msebor@redhat.com>
5269
5270 PR middle-end/93200
5271 PR fortran/92956
5272 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
5273
5274 2020-01-09 Martin Liska <mliska@suse.cz>
5275
5276 * auto-profile.c (auto_profile): Use opt_for_fn
5277 for a parameter.
5278 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
5279 (propagate_vals_across_arith_jfunc): Likewise.
5280 (hint_time_bonus): Likewise.
5281 (incorporate_penalties): Likewise.
5282 (good_cloning_opportunity_p): Likewise.
5283 (perform_estimation_of_a_value): Likewise.
5284 (estimate_local_effects): Likewise.
5285 (ipcp_propagate_stage): Likewise.
5286 * ipa-fnsummary.c (decompose_param_expr): Likewise.
5287 (set_switch_stmt_execution_predicate): Likewise.
5288 (analyze_function_body): Likewise.
5289 * ipa-inline-analysis.c (offline_size): Likewise.
5290 * ipa-inline.c (early_inliner): Likewise.
5291 * ipa-prop.c (ipa_analyze_node): Likewise.
5292 (ipcp_transform_function): Likewise.
5293 * ipa-sra.c (process_scan_results): Likewise.
5294 (ipa_sra_summarize_function): Likewise.
5295 * params.opt: Rename ipcp-unit-growth to
5296 ipa-cp-unit-growth. Add Optimization for various
5297 IPA-related parameters.
5298
5299 2020-01-09 Richard Biener <rguenther@suse.de>
5300
5301 PR middle-end/93054
5302 * gimplify.c (gimplify_expr): Deal with NOP definitions.
5303
5304 2020-01-09 Richard Biener <rguenther@suse.de>
5305
5306 PR tree-optimization/93040
5307 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
5308
5309 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
5310
5311 * common/config/avr/avr-common.c (avr_option_optimization_table)
5312 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
5313
5314 2020-01-09 Martin Liska <mliska@suse.cz>
5315
5316 * cgraphclones.c (symbol_table::materialize_all_clones):
5317 Use cgraph_node::dump_name.
5318
5319 2020-01-09 Jakub Jelinek <jakub@redhat.com>
5320
5321 PR inline-asm/93202
5322 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
5323 output_operand_lossage instead of gcc_unreachable.
5324 * doc/md.texi (riscv f constraint): Fix typo.
5325
5326 PR target/93141
5327 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
5328 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
5329 CONST_SCALAR_INT_P instead of CONST_INT_P.
5330 (*subv<mode>4_1): Rename to ...
5331 (subv<mode>4_1): ... this.
5332 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5333 define_insn_and_split patterns.
5334 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5335 patterns.
5336
5337 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5338
5339 * vec.c (class selftest::count_dtor): New class.
5340 (selftest::test_auto_delete_vec): New test.
5341 (selftest::vec_c_tests): Call it.
5342 * vec.h (class auto_delete_vec): New class template.
5343 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
5344
5345 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5346
5347 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
5348
5349 2020-01-08 Jim Wilson <jimw@sifive.com>
5350
5351 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
5352 use of TLS_MODEL_LOCAL_EXEC when not pic.
5353
5354 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5355
5356 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
5357 memory leak.
5358
5359 2020-01-08 Jakub Jelinek <jakub@redhat.com>
5360
5361 PR target/93187
5362 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
5363 *stack_protect_set_3 peephole2): Also check that the second
5364 insns source is general_operand.
5365
5366 PR target/93174
5367 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
5368 predicate for output operand instead of register_operand.
5369 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
5370 memory destination and non-memory operands[2].
5371
5372 2020-01-08 Martin Liska <mliska@suse.cz>
5373
5374 * cgraph.c (cgraph_node::dump): Use ::dump_name or
5375 ::dump_asm_name instead of (::name or ::asm_name).
5376 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5377 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5378 (analyze_functions): Likewise.
5379 (expand_all_functions): Likewise.
5380 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5381 (propagate_bits_across_jump_function): Likewise.
5382 (dump_profile_updates): Likewise.
5383 (ipcp_store_bits_results): Likewise.
5384 (ipcp_store_vr_results): Likewise.
5385 * ipa-devirt.c (dump_targets): Likewise.
5386 * ipa-fnsummary.c (analyze_function_body): Likewise.
5387 * ipa-hsa.c (check_warn_node_versionable): Likewise.
5388 (process_hsa_functions): Likewise.
5389 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5390 (set_alias_uids): Likewise.
5391 * ipa-inline-transform.c (save_inline_function_body): Likewise.
5392 * ipa-inline.c (recursive_inlining): Likewise.
5393 (inline_to_all_callers_1): Likewise.
5394 (ipa_inline): Likewise.
5395 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5396 (ipa_propagate_frequency): Likewise.
5397 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5398 (remove_described_reference): Likewise.
5399 * ipa-pure-const.c (worse_state): Likewise.
5400 (check_retval_uses): Likewise.
5401 (analyze_function): Likewise.
5402 (propagate_pure_const): Likewise.
5403 (propagate_nothrow): Likewise.
5404 (dump_malloc_lattice): Likewise.
5405 (propagate_malloc): Likewise.
5406 (pass_local_pure_const::execute): Likewise.
5407 * ipa-visibility.c (optimize_weakref): Likewise.
5408 (function_and_variable_visibility): Likewise.
5409 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5410 (ipa_discover_variable_flags): Likewise.
5411 * lto-streamer-out.c (output_function): Likewise.
5412 (output_constructor): Likewise.
5413 * tree-inline.c (copy_bb): Likewise.
5414 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5415 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5416
5417 2020-01-08 Richard Biener <rguenther@suse.de>
5418
5419 PR middle-end/93199
5420 * tree-eh.c (sink_clobbers): Update virtual operands for
5421 the first and last stmt only. Add a dry-run capability.
5422 (pass_lower_eh_dispatch::execute): Perform clobber sinking
5423 after CFG manipulations and in RPO order to catch all
5424 secondary opportunities reliably.
5425
5426 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5427
5428 PR target/93182
5429 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5430
5431 2019-01-08 Richard Biener <rguenther@suse.de>
5432
5433 PR middle-end/93199
5434 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5435 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5436 virtual operand, also updating SSA use.
5437 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5438 Update stmt after resetting virtual operand.
5439 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5440 * gimple-iterator.c (gsi_remove): When not removing the stmt
5441 permanently do not delink immediate uses or mark the stmt modified.
5442
5443 2020-01-08 Martin Liska <mliska@suse.cz>
5444
5445 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5446 (ipa_call_context::estimate_size_and_time): Likewise.
5447 (inline_analyze_function): Likewise.
5448
5449 2020-01-08 Martin Liska <mliska@suse.cz>
5450
5451 * cgraph.c (cgraph_node::dump): Use systematically
5452 dump_asm_name.
5453
5454 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5455
5456 Add -nodevicespecs option for avr.
5457
5458 PR target/93182
5459 * config/avr/avr.opt (-nodevicespecs): New driver option.
5460 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5461 "-specs=device-specs/..." if that option is not set.
5462 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5463
5464 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5465
5466 Implement 64-bit double functions for avr.
5467
5468 PR target/92055
5469 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5470 --with-double-comparison.
5471 * doc/install.texi: Document them.
5472 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5473 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5474 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5475 * doc/invoke.texi (AVR Built-in Macros): Document them.
5476 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5477 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5478 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5479
5480 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5481
5482 PR target/93188
5483 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5484 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5485 when only building rm-profile multilibs.
5486
5487 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5488
5489 PR ipa/93084
5490 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5491 lattice for a value to check.
5492 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5493 finite propagation in self-recursive scc.
5494
5495 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5496
5497 * ipa-inline.c (caller_growth_limits): Restore the AND.
5498
5499 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5500
5501 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5502 (VEC_ALLREG_ALT): New iterator.
5503 (VEC_ALLREG_INT_MODE): New iterator.
5504 (VCMP_MODE): New iterator.
5505 (VCMP_MODE_INT): New iterator.
5506 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5507 (vec_cmp<u>v64qidi): New define_expand.
5508 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5509 (vec_cmpu<mode>di_exec): New define_expand.
5510 (vec_cmp<u>v64qidi_exec): New define_expand.
5511 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5512 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5513 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5514 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5515 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5516 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5517 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5518 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5519 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5520 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5521 this.
5522 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5523 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5524
5525 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5526
5527 * config/gcn/constraints.md (DA): Update description and match.
5528 (DB): Likewise.
5529 (Db): New constraint.
5530 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5531 parameter.
5532 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5533 Implement 'Db' mixed immediate type.
5534 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5535 (addcv64si3_dup<exec_vcc>): Delete.
5536 (subcv64si3<exec_vcc>): Rework constraints.
5537 (addv64di3): Rework constraints.
5538 (addv64di3_exec): Rework constraints.
5539 (subv64di3): Rework constraints.
5540 (addv64di3_dup): Delete.
5541 (addv64di3_dup_exec): Delete.
5542 (addv64di3_zext): Rework constraints.
5543 (addv64di3_zext_exec): Rework constraints.
5544 (addv64di3_zext_dup): Rework constraints.
5545 (addv64di3_zext_dup_exec): Rework constraints.
5546 (addv64di3_zext_dup2): Rework constraints.
5547 (addv64di3_zext_dup2_exec): Rework constraints.
5548 (addv64di3_sext_dup2): Rework constraints.
5549 (addv64di3_sext_dup2_exec): Rework constraints.
5550
5551 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5552
5553 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5554 existing target checks.
5555
5556 2020-01-07 Richard Biener <rguenther@suse.de>
5557
5558 * doc/install.texi: Bump minimal supported MPC version.
5559
5560 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5561
5562 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5563 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5564 * langhooks.c: Include stor-layout.h.
5565 (lhd_simulate_enum_decl): New function.
5566 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5567 handle_arm_sve_h for the LTO frontend.
5568 (register_vector_type): Cope with null returns from pushdecl.
5569
5570 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5571
5572 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5573 (aarch64_sve::nvectors_if_data_type): Replace with...
5574 (aarch64_sve::builtin_type_p): ...this.
5575 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5576 (find_vector_type): Delete.
5577 (add_sve_type_attribute): New function.
5578 (lookup_sve_type_attribute): Likewise.
5579 (register_builtin_types): Add an "SVE type" attribute to each type.
5580 (register_tuple_type): Likewise.
5581 (svbool_type_p, nvectors_if_data_type): Delete.
5582 (mangle_builtin_type): Use lookup_sve_type_attribute.
5583 (builtin_type_p): Likewise. Add an overload that returns the
5584 number of constituent vector and predicate registers.
5585 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5586 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5587 instead of aarch64_sve_argument_p.
5588 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5589 (aarch64_pass_by_reference): Likewise.
5590 (aarch64_function_value_1): Likewise.
5591 (aarch64_return_in_memory): Likewise.
5592 (aarch64_layout_arg): Likewise.
5593
5594 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5595
5596 PR tree-optimization/93156
5597 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5598 least significant bit is always clear.
5599
5600 PR tree-optimization/93118
5601 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5602 simplifier with two intermediate conversions.
5603
5604 2020-01-07 Martin Liska <mliska@suse.cz>
5605
5606 * params.opt: Add Optimization for various parameters.
5607
5608 2020-01-07 Martin Liska <mliska@suse.cz>
5609
5610 PR ipa/83411
5611 * doc/extend.texi: Explain cloning for target_clone
5612 attribute.
5613
5614 2020-01-07 Martin Liska <mliska@suse.cz>
5615
5616 PR tree-optimization/92860
5617 * common.opt: Make in Optimization option
5618 as it is affected by -O0, which is an Optimization
5619 option.
5620 * tree-inline.c (tree_inlinable_function_p):
5621 Use opt_for_fn for warn_inline.
5622 (expand_call_inline): Likewise.
5623
5624 2020-01-07 Martin Liska <mliska@suse.cz>
5625
5626 PR tree-optimization/92860
5627 * common.opt: Make flag_ree as optimization
5628 attribute.
5629
5630 2020-01-07 Martin Liska <mliska@suse.cz>
5631
5632 PR optimization/92860
5633 * params.opt: Mark param_min_crossjump_insns with Optimization
5634 keyword.
5635
5636 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5637
5638 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5639 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5640
5641 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5642
5643 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5644 helper function to return the valid addressing formats for a given
5645 hard register and mode.
5646 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5647
5648 * config/rs6000/constraints.md (Q constraint): Update
5649 documentation.
5650 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5651 documentation.
5652
5653 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5654 Use 'Q' for doing vector extract from memory.
5655 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5656 memory.
5657 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5658 doing vector extract from memory.
5659 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5660 extract from memory.
5661
5662 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5663 for the offset being 34-bits when -mcpu=future is used.
5664
5665 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5666
5667 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5668 instead of cmpib_comparison_operator in cmpib patterns.
5669 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5670 of cmpib_comparison_operator. Revise comment.
5671
5672 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5673
5674 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5675 in an IFN_DIV_POW2 node to be equal.
5676
5677 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5678
5679 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5680 (vect_check_scalar_mask): ...this.
5681 (vectorizable_store, vectorizable_load): Update call accordingly.
5682 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5683 argument in calls to conditional internal functions.
5684
5685 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5686
5687 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5688 '0' matching inputs.
5689 (subv64di3_exec): Likewise.
5690
5691 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5692
5693 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5694 * doc/md.texi (movstr): Likewise.
5695
5696 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5697
5698 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5699 clobber.
5700
5701 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5702
5703 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5704 Depend on...
5705 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5706 to a temporary file and use move-if-change to update the real
5707 file where necessary.
5708
5709 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5710
5711 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5712 rather than Upa for CPY /M.
5713
5714 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5715
5716 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5717 immediate.
5718
5719 2020-01-06 Martin Liska <mliska@suse.cz>
5720
5721 PR tree-optimization/92860
5722 * params.opt: Mark param_max_combine_insns with Optimization
5723 keyword.
5724
5725 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5726
5727 PR target/93141
5728 * config/i386/i386.md (SWIDWI): New mode iterator.
5729 (DWI, dwi): Add TImode variants.
5730 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5731 <general_hilo_operand> instead of <general_operand>. Use
5732 CONST_SCALAR_INT_P instead of CONST_INT_P.
5733 (*addv<mode>4_1): Rename to ...
5734 (addv<mode>4_1): ... this.
5735 (QWI): New mode attribute.
5736 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5737 define_insn_and_split patterns.
5738 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5739 patterns.
5740 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5741 <general_hilo_operand> instead of <general_operand>.
5742 (*addcarry<mode>_1): New define_insn.
5743 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5744
5745 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5746
5747 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5748 Use "call" instead of "set".
5749
5750 2020-01-03 Martin Jambor <mjambor@suse.cz>
5751
5752 PR ipa/92917
5753 * ipa-cp.c (print_all_lattices): Skip functions without info.
5754
5755 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5756
5757 PR target/93089
5758 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5759 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5760 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5761 for 'e' simd clones.
5762
5763 PR target/93089
5764 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5765 entry.
5766 (mprefer-vector-width=): Add Save.
5767 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5768 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5769 (ix86_debug_options, ix86_function_specific_print): Adjust
5770 ix86_target_string callers.
5771 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5772 (ix86_valid_target_attribute_tree): Likewise.
5773 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5774 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5775 ix86_target_string caller.
5776
5777 PR target/93110
5778 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5779 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5780 instead of gen_int_shift_amount + convert_modes.
5781
5782 PR rtl-optimization/93088
5783 * loop-iv.c (find_single_def_src): Punt after looking through
5784 128 reg copies for regs with single definitions. Move definitions
5785 to first uses.
5786
5787 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5788
5789 * config/arm/arm-c.c (arm_cpu_builtins): Define
5790 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5791 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5792 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5793 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5794 * config/arm/arm-tables.opt: Regenerated.
5795 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5796 arm_arch_i8mm and arm_arch_bf16 when enabled.
5797 * config/arm/arm.h (TARGET_I8MM): New macro.
5798 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5799 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5800 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5801 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5802 (v8_6_a_simd_variants): New.
5803 (v8_*_a_simd_variants): Add i8mm and bf16.
5804 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5805
5806 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5807
5808 PR ipa/93087
5809 * predict.c (compute_function_frequency): Don't call
5810 warn_function_cold on functions that already have cold attribute.
5811
5812 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5813
5814 PR target/67834
5815 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5816 COMDAT group function labels in .data.rel.ro.local section.
5817 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5818
5819 PR target/93111
5820 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5821 comparison_operator in B and S integer comparisons. Likewise, use
5822 ordered_comparison_operator instead of cmpib_comparison_operator in
5823 cmpib patterns.
5824 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5825
5826 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5827
5828 Update copyright years.
5829
5830 * gcc.c (process_command): Update copyright notice dates.
5831 * gcov-dump.c (print_version): Ditto.
5832 * gcov.c (print_version): Ditto.
5833 * gcov-tool.c (print_version): Ditto.
5834 * gengtype.c (create_file): Ditto.
5835 * doc/cpp.texi: Bump @copying's copyright year.
5836 * doc/cppinternals.texi: Ditto.
5837 * doc/gcc.texi: Ditto.
5838 * doc/gccint.texi: Ditto.
5839 * doc/gcov.texi: Ditto.
5840 * doc/install.texi: Ditto.
5841 * doc/invoke.texi: Ditto.
5842
5843 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5844
5845 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5846 summary.
5847
5848 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5849
5850 PR tree-optimization/93098
5851 * match.pd (popcount): For shift amounts, use integer_onep
5852 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5853 tests. Make sure that precision is power of two larger than or equal
5854 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5855 instead of ULL suffixed constants. Formatting fixes.
5856 \f
5857 Copyright (C) 2020 Free Software Foundation, Inc.
5858
5859 Copying and distribution of this file, with or without modification,
5860 are permitted in any medium without royalty provided the copyright
5861 notice and this notice are preserved.