1 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
3 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
5 2018-09-17 David Malcolm <dmalcolm@redhat.com>
7 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
8 format_string_diagnostic_t.
10 * substring-locations.c
11 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
12 (format_warning_n_va): Convert to...
13 (format_string_diagnostic_t::emit_warning_n_va): ...this.
14 (format_warning_va): Convert to...
15 (format_string_diagnostic_t::emit_warning_va): ...this.
16 (format_warning_at_substring): Convert to...
17 (format_string_diagnostic_t::emit_warning): ...this.
18 (format_warning_at_substring_n): Convert to...
19 (format_string_diagnostic_t::emit_warning_n): ...this.
20 * substring-locations.h (class format_string_diagnostic_t): New
22 (format_warning_va): Convert to
23 format_string_diagnostic_t::emit_warning_va.
24 (format_warning_n_va): Convert to
25 format_string_diagnostic_t::emit_warning_n_va.
26 (format_warning_at_substring): Convert to
27 format_string_diagnostic_t::emit_warning.
28 (format_warning_at_substring_n): Convert to
29 format_string_diagnostic_t::emit_warning_n.
31 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
32 Bernd Schmidt <bernds_cb1@t-online.de>
34 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
37 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
39 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
40 operand 0 predicate to nonimmediate operand.
41 (rint<mode>2_frndint): Remove insn pattern.
42 (rint<mode>2): Change operand 1 predicate to general_operand.
43 Extend operand 1 to XFmode and generate rintxf2 insn.
44 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
45 Do not use X87MODEF mode macro.
46 (frndintxf2_<rounding>_i387): Rename from
47 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
48 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
49 to XFmode and generate significandxf3 insn.
51 2018-09-17 Richard Biener <rguenther@suse.de>
53 PR tree-optimization/87328
54 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
55 visiting unexecutable backedges when not iterating.
56 (do_rpo_vn): Mark all edges not executable even when not
59 2018-09-17 Martin Jambor <mjambor@suse.cz>
62 * doc/invoke.texi (Warning Options): Likewise.
64 2018-09-17 Richard Biener <rguenther@suse.de>
66 PR tree-optimization/87301
67 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
68 clean EH info from leftover copy assignments.
70 2018-09-17 Martin Liska <mliska@suse.cz>
73 * gcov.c (output_intermediate_file): Fix out of bounds
76 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
78 * config/arc/arc.c: Object attributes for core4 not reflected
80 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
83 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
85 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
87 2018-09-17 Martin Liska <mliska@suse.cz>
89 * doc/gcov.texi: Document new option --use-hotness-colors.
90 * gcov.c (struct source_info): Declare new field.
91 (source_info::source_info): Set default for maximum_count.
92 (print_usage): Add new -q option.
93 (process_args): Process it.
94 (accumulate_line_info): Save src->maximum_count.
95 (output_line_beginning): Make color line number if
96 flag_use_hotness_colors is set.
97 (output_line_details): Pass default argument value.
98 (output_lines): Pass src->maximum_count.
100 2018-09-17 Martin Liska <mliska@suse.cz>
102 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
103 Use processor_names table.
104 * config/i386/i386.c (ix86_default_align): Use
105 processor_cost_table for alignment values.
106 (ix86_option_override_internal): Use processor_names.
107 (ix86_function_specific_print): Likewise.
108 * config/i386/i386.h (struct processor_costs):
109 Add alignment values.
110 (struct ptt): Remove and replace with const char *.
111 * config/i386/x86-tune-costs.h (struct processor_costs):
112 Declare default alignments for all costs.
114 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
116 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
117 symbolics or VR_VARYING ranges for ABS_EXPR.
118 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
119 when range will wrap.
121 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
124 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
125 before and after a JUMP_TABLE_DATA.
127 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
130 * dojump.c (do_compare_and_jump): Canonicalize function pointers
131 when one operand is a function pointer. Use POINTER_TYPE_P and
132 FUNC_OR_METHOD_TYPE_P.
133 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
134 * fold-const.c (build_range_check): Likewise.
135 * match.pd (simple_comparison): Likewise.
137 2018-09-14 David Malcolm <dmalcolm@redhat.com>
140 * spellcheck.c (get_edit_distance_cutoff): New function.
141 (selftest::test_edit_distance_unit_test_oneway): Rename to...
142 (selftest::test_get_edit_distance_one_way): ...this.
143 (selftest::test_get_edit_distance_unit): Rename to...
144 (selftest::test_get_edit_distance_both_ways): ...this.
145 (selftest::test_edit_distances): Move tests to this new function,
146 and test some more pairs of strings. Update for above renaming.
147 (selftest::get_old_cutoff): New function.
148 (selftest::test_get_edit_distance_cutoff): New function.
149 (selftest::assert_suggested_for): New function.
150 (ASSERT_SUGGESTED_FOR): New macro.
151 (selftest::assert_not_suggested_for): New function.
152 (ASSERT_NOT_SUGGESTED_FOR): New macro.
153 (selftest::test_suggestions): New function.
154 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
155 tests to selftest::test_edit_distances and call it. Add calls to
156 selftest::test_get_edit_distance_cutoff and
157 selftest::test_suggestions.
158 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
159 (best_match::consider): Replace hard-coded cutoff calculation with
161 (best_match::get_cutoff): New declaration.
162 (best_match::get_best_meaningful_candidate): Likewise.
164 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
166 * builtins.c (fold_builtin_strlen): Remove TODO comment.
168 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
171 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
173 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
174 terminated string literal.
176 2018-09-14 Martin Sebor <msebor@redhat.com>
178 * builtins.c (unterminated_array): Handle ARRAY_REF.
179 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
180 * builtins.h (unterminated_array): Declare extern.
181 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
183 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
186 2018-09-14 Martin Sebor <msebor@redhat.com>
187 Jeff Law <law@redhat.com>
189 * builtins.c (unterminated_array): New.
190 (expand_builtin_strcpy): Adjust.
191 (expand_builtin_strcpy_args): Detect unterminated arrays.
192 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
194 * gimple-fold.h (get_maxval_strlen): Add argument.
195 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
197 * gimple-fold.c (get_range_strlen): Add argument.
198 (get_maxval_strlen): Adjust.
199 * gimple-fold.h (get_range_strlen): Add argument.
201 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
203 * config/i386/movdirintrin.h: Fix copyright year.
205 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
207 * reg-stack.c: Include regs.h.
208 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
209 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
210 FIRST_STACK_REG, not DFmode.
211 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
212 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
213 (change stack): Default register mode to the reg_raw_mode of
214 FIRST_STACK_REG, not DFmode.
215 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
216 (*swapxf): Rename from swapxf.
218 2018-09-14 Carl Love <cel@us.ibm.com>
220 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
221 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
223 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
226 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
229 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
232 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
234 2018-09-14 Jason Merrill <jason@redhat.com>
236 Fix --enable-gather-detailed-mem-stats.
237 * hash-table.c (hash_table_usage): Change from variable to function.
238 * hash-table.h: Adjust.
239 * Makefile.in: Add missing dependencies on hash-table.h.
241 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
243 PR tree-optimization/87259
245 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
246 execute_cse_reciprocals_1 has tried transforming.
248 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
250 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
251 VR_VARYING for PLUS/MINUS_EXPR.
253 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
255 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
258 2018-09-14 Richard Biener <rguenther@suse.de>
261 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
262 bits for the merged partition.
264 2018-09-13 Martin Sebor <msebor@redhat.com>
265 Bernd Edlinger <bernd.edlinger@hotmail.de>
267 * builtins.h (c_srlen): Add argument.
268 * builtins.c (warn_string_no_nul): New function.
269 (c_strlen): Add argument and use it. Update recursive calls.
270 Pass DECL argument to string_constant to get info on non
271 terminated strings. Update *NONSTR as needed.
272 (fold_builtin_strlen): Add argument to calls to c_strlen.
273 Warn for unterminated arrays.
274 (warn_string_no_null): Add prototype.
275 * expr.c (string_constant): Update arguments. Update recursive
276 calls appropriately. Detect missing NUL terminator and outermost
277 declaration its missing in.
278 Improve checks for arrays with nonzero lower bound or elements
279 that are not a single byte. Simplify offset computation.
280 Simplify checks for non-NUL terminated strings.
281 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
282 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
284 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
286 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
288 * fold-const.c (c_getstr): Fix function comment. Remove unused third
289 argument. Fix range checks.
290 * fold-const.h (c_getstr): Adjust protoype.
291 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
292 string is constant but contains no NUL byte.
294 * expr.c (string_constant): Adjust function comment.
295 Remove bogus check for zero termination.
297 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
299 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
300 (get_constant_size): Don't make STRING_CSTs larger than they are.
301 (check_string_literal): New check function for STRING_CSTs.
302 (output_constant): Use it.
304 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
307 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
309 2018-09-13 Richard Biener <rguenther@suse.de>
311 PR tree-optimization/87263
312 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
313 (struct unwind_state): Add max_rpo field.
314 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
315 Compute max_rpo, the max RPO number a block can be backwards reached
316 from. Re-write non-iterating mode to a RPO ordered worklist approach,
317 separating it from the iterating mode.
319 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
321 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
322 (rfs_decision): New scheduling decision.
324 2018-09-13 Richard Biener <rguenther@suse.de>
327 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
328 (vn_nary_op_insert_pieces_predicated): Do not write useless
329 valid_dominated_by_p entry outside of the allocated storage.
331 2018-09-13 Omar Sandoval <osandov@osandov.com>
332 Tom de Vries <tdevries@suse.de>
335 * dwarf2out.c (is_c): New function.
336 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
338 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
341 * config/aarch64/aarch64.md (*aarch64_bfxil):
343 * config/aarch64/constraints.md (Ulc): Define.
344 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
346 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
349 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
351 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
352 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
353 aarch64_layout_frame call.
354 (aarch64_expand_epilogue): Likewise.
355 (aarch64_initial_elimination_offset): Likewise.
356 (aarch64_get_separate_components): Likewise.
357 (aarch64_use_return_insn_p): Likewise.
358 (aarch64_layout_frame): Remove unneeded check.
360 2018-09-13 Jakub Jelinek <jakub@redhat.com>
362 * configure.ac: Only append
363 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
364 gcc_config_arguments if it was never reconfigured or last reconfigure
365 was with different arguments.
366 * configure: Regenerated.
368 2018-09-13 Jakub Jelinek <jakub@redhat.com>
369 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
372 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
373 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
375 2018-09-13 Jakub Jelinek <jakub@redhat.com>
377 PR tree-optimization/87287
378 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
379 X % C == 0 to X % (unsigned) C == 0 optimization to ...
380 * match.pd (X % C == 0): ... here. New optimization.
382 2018-09-12 Jakub Jelinek <jakub@redhat.com>
385 * expr.h (maybe_optimize_mod_cmp): Declare.
386 * expr.c (mod_inv): New function.
387 (maybe_optimize_mod_cmp): New function.
388 (do_store_flag): Use it.
389 * cfgexpand.c (expand_gimple_cond): Likewise.
391 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
392 Julian Brown <julian@codesourcery.com>
395 * gimplify.c (gimplify_scan_omp_clauses): Set
396 target_firstprivatize_array_bases in OpenACC parallel and kernels
397 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
398 OpenACC data regions.
400 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
402 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
403 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
405 2018-09-12 Richard Biener <rguenther@suse.de>
407 PR tree-optimization/87280
408 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
409 edge but unreachable target.
410 (do_rpo_vn): For conservatively handling a PHI only mark
411 the backedge executable but not the block reachable.
413 2018-09-12 Richard Biener <rguenther@suse.de>
415 PR tree-optimization/87266
416 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
419 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
421 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
423 ("trunc<BFP:mode><DFP_ALL:mode>2")
424 ("trunc<DFP_ALL:mode><BFP:mode>2")
425 ("extend<BFP:mode><DFP_ALL:mode>2")
426 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
427 according to the target operand type.
429 2018-09-12 Jakub Jelinek <jakub@redhat.com>
430 Andreas Krebbel <krebbel@linux.ibm.com>
432 PR tree-optimization/86844
433 * gimple-ssa-store-merging.c
434 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
435 there are any overlapping stores in between them, make sure they are
436 also coalesced or we give up completely.
438 2018-09-12 Jakub Jelinek <jakub@redhat.com>
441 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
442 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
444 2018-09-12 Tom de Vries <tdevries@suse.de>
446 * common.opt (gdescribe-dies): Add option.
447 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
448 attribute for artifical and nameless decls.
449 (dwarf2out_register_external_die): Add description attribute to
450 external reference die.
451 (add_desc_attribute): New functions.
452 (gen_subprogram_die): Add description attribute to
453 DW_TAG_call_site_parameter.
454 * tree-pretty-print.c (print_generic_expr_to_str): New function.
455 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
456 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
458 (@item -gdescribe-dies): Add.
460 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
462 * tree-vrp.c (vrp_shift_undefined_p): Remove.
463 (extract_range_from_binary_expr_1: Call
464 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
465 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
468 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
470 * gimple-ssa-warn-alloca.c
471 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
472 field for ALLOCA_BOUND_*_LARGE.
474 2018-09-11 Nathan Sidwell <nathan@acm.org>
476 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
478 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
480 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
481 for clobbers. Remove obsolete comment.
483 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
485 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
486 mpxchk, mpxld and mpxst types.
487 (define_attr length_immediate): Remove all processing of mpx types.
488 (define_attr prefix_0f): Ditto.
489 (define_attr memory): Ditto.
491 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
493 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
494 (log<mode>2): Change operand 1 predicate to general_operand.
495 Extend operand 1 to XFmode and generate logxf3 insn.
496 (log10<mode>2): Change operand 1 predicate to general_operand.
497 Extend operand 1 to XFmode and generate log10xf3 insn.
498 (log2<mode>2): Change operand 1 predicate to general_operand.
499 Extend operand 1 to XFmode and generate log2xf3 insn.
500 (fyl2xp1_extend<mode>xf3_i387): Remove.
501 (log1p<mode>2): Change operand 1 predicate to general_operand.
502 Extend operand 1 to XFmode and generate log1pxf3 insn.
503 (fxtract_extend<mode>xf3_i387): Remove.
504 (logb<mode>2): Change operand 1 predicate to general_operand.
505 Extend operand 1 to XFmode and generate logbxf3 insn.
506 (ilogb<mode>2): Change operand 1 predicate to general_operand.
507 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
508 (significand<mode>2): Change operand 1 predicate to general_operand.
509 Extend operand 1 to XFmode and generate significandxf3 insn.
511 2018-09-11 Nathan Sidwell <nathan@acm.org>
513 * gcc.c (perror_with_name, pfatal_with_name): Delete.
514 (load_specs): Use fatal_error.
515 (DELETE_IF_ORDINARY, process_command): Use error.
516 (execute, run_attempt): Use fatal_error.
518 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
520 * diagnostic-core.h (sorry_at): New prototype.
521 * diagnostic.c (sorry_at): New function.
523 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
525 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
526 by zero as VR_UNDEFINED.
528 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
530 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
531 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
532 (<sincos>mode2): New expander.
533 (sincos_extend<mode>xf3_i387): Remove insn pattern.
534 (sincos -> sin, cos splitters): Remove splitter patterns.
535 (sincos<mode>3): Change operand 2 predicate to general_operand.
536 Extend operand 2 to XFmode and generate sincosxf3 insn.
537 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
538 Change operand 3 predicate to const1_operand.
539 (fptan_extend<mode>xf4_i387): Remove insn pattern.
540 (tanxf2): Update operands in the call to fptanxf4_i387.
541 (tan<mode>2): Change operand 1 predicate to general_operand.
542 Extend operand 1 to XFmode and generate tanxf3 insn.
543 (atan2xf3): Rename from *fpatanxf3_i387.
544 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
545 (atan2xf3): Remove expander.
546 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
547 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
548 (atan<mode>2): Change operand 1 predicate to general_operand.
549 Extend operand 1 to XFmode and generate atanxf3 insn.
551 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
553 * config/i386/i386.md (x87/SSE constant load splitter): Use
554 memory_operand instead of nonimmediate_operand for input operand
557 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
559 * config/i386/i386.md (float partial SSE register stall splitter): Move
560 splitter near its instruction pattern.
561 (float_extend partial SSE register stall splitter): Ditto.
562 (float_truncate partial SSE register stall splitter): Ditto.
564 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
567 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
568 to speculation_safe_value_not_needed.
571 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
572 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
573 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
574 leaf_function_p, instead use has_hard_reg_initial_val.
576 2018-09-09 Nathan Sidwell <nathan@acm.org>
578 * gcc.h (pfatal_with_name): Don't declare here.
579 * gcc.c (pfatal_with_name): Make static.
581 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
583 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
586 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
588 PR rtl-optimization/85458
589 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
590 priority hook to reduce the priority of EXPR.
592 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
594 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
595 DImode for x87 on 32bit targets. Conditionally disable x87 modes
596 with X87_ENABLE_FLOAT. Remove preparation code.
597 (*float<SWI48:mode><MODEF:mode>2): Rename from
598 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
599 math using "enabled" attribute.
600 (*floatdi<MODEF:mode>2_i387): Rename from
601 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
602 enable for 32bit targets only.
603 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
605 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
606 as operand 1 predicate. Rewrite as define_insn_and_split.
607 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
609 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
611 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
612 to fallthru to FLOAT case.
614 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
617 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
618 around folding of vec_sl to handle out of range shift values.
620 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
622 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
623 Update callers to gen_fix_trunc<mode>_i387_fisttp
624 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
625 nonimmediate_operand.
626 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
627 and corresponding splitters.
628 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
629 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
630 (fix_truncdi_i387_with_temp): Remove insn pattern
631 and corresponding splitters.
632 (fix_trunc<mode>_i387): Change operand 0 predicate to
633 nonimmediate_operand.
634 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
635 and corresponding splitters.
636 (*fistdi2_1): Remove.
638 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
639 (lrintxfdi2): Remove expander. Reimplement as define_insn.
640 (*fist<mode>2_1): Remove.
641 (fist<mode>2): Ditto.
642 (fist<mode>2_with_temp): Remove insn pattern and corresponding
644 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
645 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
646 (fistdi2_<rounding>): Change operand 0 predicate to
647 nonimmediate_operand.
648 (fistdi2_<rounding>_with_temp): Remove insn pattern
649 and corresponding splitters.
650 (fist<mode>2_<rounding>): Change operand 0 predicate to
651 nonimmediate_operand.
652 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
653 and corresponding splitters.
655 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
657 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
659 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
662 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
664 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
665 early gimple folding of vec_splat().
666 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
667 * gimple-fold.h: Add an extern define for tree_vec_extract().
669 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
671 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
672 wrappers around TREE_TYPE comparisons.
674 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
677 * config/s390/predicates.md: Add nonsym_memory_operand.
678 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
679 contains a SYMBOL_REF, load it into an intermediate pseudo.
680 (s390_emit_compare_and_swap): Legitimize operand.
681 * config/s390/s390.md: Use the new nonsym_memory_operand
682 with UNSPECV_CAS patterns.
684 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
687 * config/s390/s390-passes.def: New file.
688 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
690 (make_pass_s390_early_mach): Add declaration.
691 * config/s390/s390.c (make_pass_s390_early_mach):
692 (s390_option_override): Remove dynamic registration.
693 * config/s390/t-s390: Add s390-passes.def.
695 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
697 * config/s390/s390.c (s390_decompose_constant_pool_ref):
698 Remove UNSPEC_LTREL_BASE check.
699 (annotate_constant_pool_refs): Likewise.
700 (find_constant_pool_ref): Likewise.
701 (find_ltrel_base): Removed.
702 (replace_ltrel_base): Removed.
703 (s390_mainpool_finish): Remove replace_ltrel_base call.
704 (s390_chunkify_start): Remove pending LTREL_BASE logic.
705 (s390_chunkify_finish): Remove replace_ltrel_base call.
706 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
708 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
711 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
712 to speculation_safe_value_not_needed.
714 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
715 Bernd Schmidt <bernds_cb1@t-online.de>
717 * config/nvptx/nvptx-opts.h: New file.
718 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
719 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
721 (TARGET_SM35): New macro.
722 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
724 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
727 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
729 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
731 * config/i386/i386.md (truncdfsf2): Remove expander.
732 (truncdfsf2_with_temp): Ditto.
733 (truncxf<mode>2): Ditto.
734 (*truncdfsf_fast_mixed): Remove insn pattern.
735 (*truncdfsf_fast_i387): Ditto.
736 (*truncdfsf_mixed): Ditto.
737 (*truncdfsf_i387): Ditto.
738 (*truncdfsf2_i387_1): Ditto.
739 (*truncxfsf2_mixed): Ditto.
740 (*truncxfdf2_mixed): Ditto.
741 (*truncxf<mode>2_i387_noop): Ditto. Update callers
742 to call gen_truncxf<mode>2 instead.
743 (*truncxf<mode>2_i387): Remove.
744 (reg->reg splitters): Remove splitter pattern.
745 (reg->mem splitters): Ditto.
747 (truncdfsf2): New insn pattern.
748 (truncxf<mode>2): Ditto.
750 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
752 * tree-ssa-math-opts.c (is_mult_by): New function.
753 (is_square_of): Use the above.
754 (optimize_recip_sqrt): New function.
755 (pass_cse_reciprocals::execute): Use the above.
757 2018-09-05 Richard Biener <rguenther@suse.de>
760 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
761 to zero-init the emplaced vec.
763 2018-09-05 Martin Liska <mliska@suse.cz>
765 PR tree-optimization/87205
766 * tree-switch-conversion.c (pass_lower_switch::execute):
767 Group cases for switch statements.
769 2018-09-05 Richard Biener <rguenther@suse.de>
771 PR tree-optimization/87217
772 * tree-ssa-sccvn.c (vuse_valueize): New.
773 (vn_reference_lookup_pieces): Use it.
774 (vn_reference_lookup): Likewise.
776 2018-09-05 Nathan Sidwell <nathan@acm.org>
779 * stor-layout.c (place_field): Scan forwards to check last
780 bitfield when ms_bitfield_placement is in effect.
782 2018-09-05 Richard Biener <rguenther@suse.de>
785 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
788 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
789 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
791 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
792 * config.gcc (extra_objs): Build it.
793 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
795 * config/aarch64/aarch64-passes.def
796 (pass_tag_collision_avoidance): New pass.
797 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
798 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
799 (aarch64_classify_address): Remove static qualifier.
800 (aarch64_address_info, aarch64_address_type): Move to...
801 * config/aarch64/aarch64-protos.h: ... here.
802 (make_pass_tag_collision_avoidance): New function.
803 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
806 2018-09-05 Martin Liska <mliska@suse.cz>
808 * doc/gcov.texi: Update documentation of humar
810 * gcov.c (format_count): Print one decimal place, it provides
811 more fine number of situations like '1G' vs. '1.4G'.
813 2018-09-05 Martin Liska <mliska@suse.cz>
816 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
817 * optc-gen.awk: Allow 'Var' for Deprecated options in order
818 to generate a MASK value.
820 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
823 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
824 if frame pointer isn't used.
825 (compute_frame_pointer_to_fb_displacement): Likewise.
827 2018-09-04 Jakub Jelinek <jakub@redhat.com>
830 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
831 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
832 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
833 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
834 and OPTION_MASK_ISA_XSAVEC_UNSET.
836 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
838 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
839 XOR operations in NAND case.
841 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
843 * wide-int-range.cc (wide_int_range_convert): New.
844 * wide-int-range.h (wide_int_range_convert): New.
845 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
846 code into wide_int_range_convert.
847 (extract_range_into_wide_ints): Do not munge anti range constants
848 into the entire domain. Just return the range back.
850 2018-09-04 Martin Liska <mliska@suse.cz>
852 * genmatch.c (output_line_directive): Add new argument
854 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
856 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
858 * doc/invoke.texi (Option Summary): Add whitespace.
860 * doc/invoke.texi (Option Summary): Add -Waligned-new.
862 2018-09-04 Richard Biener <rguenther@suse.de>
864 PR tree-optimization/87211
865 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
866 backedge value we're supposed to treat as VARYING also number
867 the PHI to VARYING in case it got a different value-number already.
869 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
871 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
872 (extract_range_from_binary_expr_1): Do not call
873 vrp_can_optimize_bit_op.
874 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
876 (wide_int_range_get_mask_and_bounds): New.
877 (wide_int_range_optimize_bit_op): New.
878 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
879 (wide_int_range_bit_and): Same.
880 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
881 (wide_int_range_optimize_bit_op): New.
882 (wide_int_range_get_mask_and_bounds): New.
884 2018-09-04 Richard Biener <rguenther@suse.de>
886 PR tree-optimization/87176
887 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
888 variable. When value-numbering a virtual PHI node make sure
889 to not value-number to the backedge value.
891 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
893 * doc/extend.texi (Long Long, Hex Floats): Document support for
894 long long and hex floats in more recent versions of ISO C++.
896 2018-09-03 Richard Biener <rguenther@suse.de>
898 PR tree-optimization/87177
899 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
902 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
904 * bb-reorder.c (edge_order): Convert to C-qsort-style
905 tri-state comparator.
906 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
908 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
910 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
911 tri-state comparator.
912 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
914 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
916 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
917 (mergesort): ... here as maximum count for using netsort.
918 (gcc_qsort): Set nlim to 3 if stable sort is requested.
919 (gcc_stablesort): New.
920 * system.h (gcc_stablesort): Declare.
922 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
924 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
925 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
926 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
928 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
930 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
931 lxsdx and stxsdx alternatives.
932 (*mov<mode>_hardfloat64): Ditto.
933 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
935 2018-09-03 Richard Biener <rguenther@suse.de>
937 PR tree-optimization/87200
938 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
941 2018-09-03 Martin Liska <mliska@suse.cz>
943 PR tree-optimization/87201
944 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
945 Fix parenthesis in an expression.
947 2018-09-03 Richard Biener <rguenther@suse.de>
949 PR tree-optimization/87197
950 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
951 visited. CSE the VN_INFO hashtable lookup.
953 PR tree-optimization/87169
954 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
955 iterating make sure there's no extra backedges from irreducible
956 regions feeding the header. Mark the destination block
959 2018-09-03 Martin Liska <mliska@suse.cz>
962 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
963 * common/common-targhooks.c (default_get_valid_option_values):
965 * common/common-targhooks.h (default_get_valid_option_values):
967 * common/config/i386/i386-common.c: Move processor_target_table
969 (ix86_get_valid_option_values): New function.
970 (TARGET_GET_VALID_OPTION_VALUES): New macro.
971 * config/i386/i386.c (struct ptt): Move to i386-common.c.
972 (PTA_*): Move all defined masks into i386-common.c.
973 (ix86_function_specific_restore): Use new processor_cost_table.
974 * config/i386/i386.h (struct ptt): Moved from i386.c.
975 (struct pta): Likewise.
976 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
977 * doc/tm.texi.in: Likewise.
978 * opt-suggestions.c (option_proposer::suggest_option):
979 Pass prefix to build_option_suggestions.
980 (option_proposer::get_completions): Likewise.
981 (option_proposer::build_option_suggestions): Use the new target
983 * opts.c (struct option_help_tuple): New struct.
984 (print_filtered_help): Use the new target hook.
986 2018-09-03 Martin Liska <mliska@suse.cz>
989 * predict.c (set_even_probabilities): Add likely_edges
990 argument and handle cases where we have precisely one
992 (combine_predictions_for_bb): Catch also likely_edges.
993 (tree_predict_by_opcode): Handle gswitch statements.
994 * tree-cfg.h (find_case_label_for_value): New declaration.
995 (find_taken_edge_switch_expr): Likewise.
996 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
997 Find pivot in decision tree based on probabily, not by number of
1000 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
1002 * doc/standards.texi (Standards): Update Objective-C reference.
1004 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1006 * doc/install.texi (Prerequisites): Update link for MPC.
1008 2018-09-01 Michael Matz <matz@suse.de>
1010 PR tree-optimization/87074
1011 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1012 PHIs for outer-loop uses.
1014 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1016 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1017 * doc/invoke.texi (C Dialect Options): Ditto.
1019 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1021 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1023 2018-08-31 Richard Biener <rguenther@suse.de>
1025 PR tree-optimization/87168
1026 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1027 (rpo_elim::eliminate_avail): When OP was not visited it must
1030 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1032 * tree-vrp.c (copy_value_range): Convert param "from" from
1033 "value_range *" to "const value_range *".
1034 (range_is_null): Likewise for param "vr".
1035 (range_int_cst_p): Likewise.
1036 (range_int_cst_singleton_p): Likewise.
1037 (symbolic_range_p): Likewise.
1038 (value_ranges_intersect_p): Likewise for both params.
1039 (value_range_nonnegative_p): Likewise for param "vr".
1040 (value_range_constant_singleton): Likewise.
1041 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1042 (extract_range_into_wide_ints): Likewise for param "vr".
1043 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1045 (vrp_can_optimize_bit_op): Likewise.
1046 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1048 (extract_range_from_unary_expr): Likewise.
1049 (debug_value_range): Likewise for param "vr".
1050 (value_range::dump): Add "const" qualifier.
1051 (vrp_prop::check_array_ref): Convert local "vr" from
1052 "value_range *" to "const value_range *".
1053 (vrp_prop::check_mem_ref): Likewise.
1054 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1055 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1056 (vrp_intersect_ranges): Likewise.
1057 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1058 (vrp_prop::vrp_finalize): Likewise.
1059 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1060 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1061 (extract_range_from_unary_expr): Likewise.
1062 (value_range_constant_singleton): Likewise.
1063 (symbolic_range_p): Likewise.
1064 (copy_value_range): Likewise.
1065 (extract_range_from_binary_expr_1): Likewise.
1066 (range_int_cst_p): Likewise.
1067 (vrp_set_zero_nonzero_bits): Likewise.
1068 (range_int_cst_singleton_p): Likewise.
1070 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1072 * config/aarch64/arm_neon.h (vabsd_s64): New.
1073 (vnegd_s64): Likewise.
1075 2018-08-31 Martin Jambor <mjambor@suse.cz>
1077 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1079 2018-08-31 Martin Liska <mliska@suse.cz>
1081 * ipa-icf.c (sem_item::add_type): Use
1082 sem_item::m_type_hash_cache.
1083 * ipa-icf.h: Move the cache from sem_item_optimizer
1086 2018-08-31 Nathan Sidwell <nathan@acm.org>
1088 * doc/extend.texi (Backwards Compatibility): Remove implicit
1089 extern C leeway of () being (...).
1091 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1093 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1095 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1099 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1101 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1104 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1105 gen_int_mode. Formatting fixes.
1107 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1109 * target.def (custom_function_descriptors): Improve documentation.
1110 * doc/tm.texi.in (Trampolines): Expand discussion of function
1111 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1112 beginning of the section.
1113 * doc/tm.texi: Regenerated.
1115 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1117 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1118 base class in the initializer list. This is a workaround for
1119 building with older GCC.
1120 (class auto_bb_flag): Likewise.
1122 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1124 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1125 (altivec_vcmpequ<VI_char>_p): Remove star.
1126 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1128 (expand_strncmp_vec_sequence): New function.
1129 (emit_final_str_compare_vec): New function.
1130 (expand_strn_compare): Add support for vector strncmp.
1131 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1132 length specification to bytes.
1133 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1134 (vcmpnezb_p): New pattern.
1135 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1136 for option -mstring-compare-inline-limit.
1138 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1140 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1141 (PTA_SKYLAKE): Add PTA_AES.
1142 (PTA_GOLDMONT): Likewise.
1144 2018-08-29 Jan Hubicka <jh@suse.cz>
1147 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1148 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1150 2018-08-29 Jan Hubicka <jh@suse.cz>
1152 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1154 (hash_tree): Do not visit TYPE_STUB_DECL.
1155 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1156 stream TYPE_STUB_DECL.
1157 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1158 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1159 after free_lang_data.
1160 (type_in_anonymous_namespace_p): Likewise.
1162 2018-08-29 Jan Hubicka <jh@suse.cz>
1164 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1165 comment that it has to be even number.
1166 (class sreal): Change m_sig type to int32_t.
1167 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1168 int64_t for temporary calculations.
1169 (sreal_verify_basics): Drop one bit from minimum and maximum.
1171 2018-08-30 Richard Biener <rguenther@suse.de>
1173 PR tree-optimization/87147
1174 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1175 (visit_phi): When the degenerate result is from the backedge and
1176 we didn't visit its definition yet drop to VARYING.
1177 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1179 2018-08-29 Jan Hubicka <jh@suse.cz>
1181 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1183 (hash_tree): Likewise.
1185 2018-08-29 Jan Hubicka <jh@suse.cz>
1187 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1188 and TYPE_NEXT_REF_TO.
1190 2018-08-29 Jan Hubicka <jh@suse.cz>
1192 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1193 comment that it has to be even number.
1194 (class sreal): Change m_sig type to int32_t.
1195 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1196 int64_t for temporary calculations.
1197 (sreal_verify_basics): Drop one bit from minimum and maximum.
1199 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1201 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1203 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1206 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1207 if to_add is negative.
1209 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1212 * builtins.c (c_strlen): Improve range checks.
1214 2018-08-29 Martin Sebor <msebor@redhat.com>
1215 Jeff Law <law@redhat.com>
1217 PR tree-optimization/86714
1218 PR tree-optimization/86711
1219 * builtins.c (c_strlen): Add arguments to call to string_constant.
1220 * expr.c (string_constant): Add argument. Detect missing nul
1221 terminator and outermost declaration it's missing in.
1222 * expr.h (string_constant): Add argument.
1223 * fold-const.c (read_from_constant_string): Add arguments to call to
1225 (c_getstr): Likewise.
1226 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1228 * tree-ssa-strlen.c (get_stridx): Likewise.
1230 2018-08-29 Jan Hubicka <jh@suse.cz>
1232 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1233 Do not stream DECL_VINDEX.
1234 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1235 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1236 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1238 2018-08-29 Richard Biener <rguenther@suse.de>
1240 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1241 virtual operands that are not default defs to honor region
1243 (rpo_vn_valueize): Remove ineffective code here.
1245 2018-08-29 Richard Biener <rguenther@suse.de>
1247 PR tree-optimization/87132
1248 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1249 when skipping defs reachable over backedges.
1251 2018-08-29 Richard Biener <rguenther@suse.de>
1253 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1254 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1255 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1256 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1257 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1258 (vn_reference_lookup_3): Remove use of const_parms.
1259 (free_rpo_vn): Do not free const_parms.
1260 (do_rpo_vn): Do not call init_const_parms.
1261 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1262 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1263 (call_may_clobber_ref_p_1): Likewise.
1265 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1268 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1270 (-ftree-scev-cprop): Document.
1272 2018-08-29 Jan Hubicka <jh@suse.cz>
1274 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1276 (sreal constructor): Update.
1277 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1278 sreal:operator/): Update.
1280 2018-08-29 Martin Liska <mliska@suse.cz>
1282 * tree-switch-conversion.c (switch_conversion::expand):
1283 Strenghten assumption about gswitch statements.
1285 2018-08-29 Richard Biener <rguenther@suse.de>
1287 PR tree-optimization/87117
1288 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1289 re-value-number released SSA VDEFs.
1291 2018-08-29 Richard Biener <rguenther@suse.de>
1293 PR tree-optimization/87126
1294 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1296 2018-08-28 Jim Wilson <jimw@sifive.com>
1298 * config/riscv/pic.md: Rewrite.
1299 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1301 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1302 (SOFTF, default_load, softload, softstore): New.
1304 2018-08-28 Jeff Law <law@redhat.com>
1306 * fold-const.c (fold_binary_loc): Remove recently added assert.
1308 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1310 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1311 to OP parmeter of generated function.
1313 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1315 PR tree-optimization/87009
1316 * match.pd: Add boolean optimizations.
1318 2018-08-28 Martin Sebor <msebor@redhat.com>
1321 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1322 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1323 (pass_walloca::gate): Use it.
1324 (alloca_call_type): Same.
1325 (pass_walloca::execute): Same.
1326 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1328 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1330 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1331 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1334 2018-08-28 Richard Biener <rguenther@suse.de>
1336 PR tree-optimization/87124
1337 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1338 constants before looking up avail.
1340 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1343 * calls.c (maybe_warn_nonstring_arg): Punt early if
1344 warn_stringop_overflow is zero. Don't call get_range_strlen
1345 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1346 Swap comparison operands to have constants on rhs. Only use
1347 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1348 increment lenrng[0].
1350 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1352 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1353 use of tree_to_shwi. Remove duplicated test for the size being
1354 a whole number of bytes.
1356 2018-08-28 Richard Biener <rguenther@suse.de>
1358 PR tree-optimization/87117
1359 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1360 Handle removed stmt without LHS (GIMPLE_NOP).
1362 2018-08-28 Richard Biener <rguenther@suse.de>
1364 PR tree-optimization/87117
1365 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1366 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1368 2018-08-28 Richard Biener <rguenther@suse.de>
1370 PR tree-optimization/87117
1371 * tree-ssa-pre.c (compute_avail): Do not make expressions
1372 with predicated values available.
1373 (get_expr_value_id): Assert we do not run into predicated value
1376 2018-08-28 Richard Biener <rguenther@suse.de>
1378 PR tree-optimization/87117
1379 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1380 get virtual operands.
1381 (get_expr_operands): Handle STRING_CST like other decls.
1383 2018-08-28 Martin Liska <mliska@suse.cz>
1385 * tree.h: Update documentation of fndecl_built_in_p
1389 2018-08-27 Jeff Law <law@redhat.com>
1390 PR tree-optimization/87110
1391 * tree-ssa-dse.c (compute_trims): Handle non-constant
1394 2018-08-27 Martin Sebor <msebor@redhat.com>
1396 PR tree-optimization/86914
1397 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1399 2018-08-27 Martin Sebor <msebor@redhat.com>
1401 PR tree-optimization/87112
1402 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1403 the type of the bound argument.
1405 2018-08-27 Jeff Law <law@redhat.com>
1407 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1408 type does not have a TYPE_SIZE_UNIT.
1410 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1412 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1413 with include of backend.h.
1415 2018-08-27 Richard Biener <rguenther@suse.de>
1417 PR tree-optimization/86927
1418 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1419 use const cond reduction code.
1421 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1423 PR tree-optimization/85758
1424 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1426 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1429 * diagnostic-show-locus.c (class layout_range): Update for
1430 conversion of show_caret_p to a tri-state.
1431 (layout_range::layout_range): Likewise.
1432 (make_range): Likewise.
1433 (layout::maybe_add_location_range): Likewise.
1434 (layout::should_print_annotation_line_p): Don't show annotation
1435 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1436 (layout::get_state_at_point): Update for conversion of
1437 show_caret_p to a tri-state. Bail out early for
1438 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1439 underlining or source colorization.
1440 (gcc_rich_location::add_location_if_nearby): Update for conversion
1441 of show_caret_p to a tri-state.
1442 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1443 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1445 (selftest::test_one_liner_labels): Likewise.
1446 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1447 conversion of show_caret_p to a tri-state.
1448 * pretty-print.c (text_info::set_location): Likewise.
1449 * pretty-print.h (text_info::set_location): Likewise.
1450 * substring-locations.c (format_warning_n_va): Likewise.
1451 * tree-diagnostic.c (default_tree_printer): Likewise.
1452 * tree-pretty-print.c (newline_and_indent): Likewise.
1454 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1457 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1458 line above for line-insertion fix-it hints.
1459 (selftest::test_fixit_insert_containing_newline): Update the
1460 expected results, and add a test with line-numbering enabled.
1462 2018-08-27 Martin Liska <mliska@suse.cz>
1465 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1466 params with DECL_HAS_VALUE_EXPR_P.
1468 2018-08-27 Martin Liska <mliska@suse.cz>
1470 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1471 selected expansion strategy.
1473 2018-08-27 Martin Liska <mliska@suse.cz>
1475 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1476 * builtins.c (is_builtin_fn): Likewise.
1477 * attribs.c (diag_attr_exclusions): Use new function
1478 fndecl_built_in_p and remove check for FUNCTION_DECL if
1480 (builtin_mathfn_code): Likewise.
1481 (fold_builtin_expect): Likewise.
1482 (fold_call_expr): Likewise.
1483 (fold_builtin_call_array): Likewise.
1484 (fold_call_stmt): Likewise.
1485 (set_builtin_user_assembler_name): Likewise.
1486 (is_simple_builtin): Likewise.
1487 * calls.c (gimple_alloca_call_p): Likewise.
1488 (maybe_warn_nonstring_arg): Likewise.
1489 * cfgexpand.c (expand_call_stmt): Likewise.
1490 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1491 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1492 (cgraph_node::verify_node): Likewise.
1493 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1494 (cgraph_node::create_clone): Likewise.
1495 * config/arm/arm.c (arm_insert_attributes): Likewise.
1496 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1497 * dse.c (scan_insn): Likewise.
1498 * expr.c (expand_expr_real_1): Likewise.
1499 * fold-const.c (operand_equal_p): Likewise.
1500 (fold_binary_loc): Likewise.
1501 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1502 * gimple-low.c (lower_stmt): Likewise.
1503 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1504 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1505 * gimple.c (gimple_build_call_from_tree): Likewise.
1506 (gimple_call_builtin_p): Likewise.
1507 (gimple_call_combined_fn): Likewise.
1508 * gimplify.c (gimplify_call_expr): Likewise.
1509 (gimple_boolify): Likewise.
1510 (gimplify_modify_expr): Likewise.
1511 (gimplify_addr_expr): Likewise.
1512 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1513 * ipa-cp.c (determine_versionability): Likewise.
1514 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1515 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1516 * ipa-split.c (visit_bb): Likewise.
1517 (split_function): Likewise.
1518 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1519 * lto-cgraph.c (input_node): Likewise.
1520 * lto-streamer-out.c (write_symbol): Likewise.
1521 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1522 (lower_omp_1): Likewise.
1523 * predict.c (strip_predict_hints): Likewise.
1524 * print-tree.c (print_node): Likewise.
1525 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1526 * trans-mem.c (is_tm_irrevocable): Likewise.
1527 (is_tm_load): Likewise.
1528 (is_tm_simple_load): Likewise.
1529 (is_tm_store): Likewise.
1530 (is_tm_simple_store): Likewise.
1531 (is_tm_abort): Likewise.
1532 (tm_region_init_1): Likewise.
1533 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1534 * tree-cfg.c (verify_gimple_call): Likewise.
1535 (move_stmt_r): Likewise.
1536 (stmt_can_terminate_bb_p): Likewise.
1537 * tree-eh.c (lower_eh_constructs_2): Likewise.
1538 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1539 * tree-inline.c (remap_gimple_stmt): Likewise.
1540 (copy_bb): Likewise.
1541 (estimate_num_insns): Likewise.
1542 (fold_marked_statements): Likewise.
1543 * tree-sra.c (scan_function): Likewise.
1544 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1545 (optimize_stack_restore): Likewise.
1546 (pass_fold_builtins::execute): Likewise.
1547 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1548 (mark_all_reaching_defs_necessary_1): Likewise.
1549 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1550 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1551 (pass_forwprop::execute): Likewise.
1552 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1553 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1554 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1555 * tree-ssa-strlen.c (get_string_length): Likewise.
1556 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1557 (find_func_aliases_for_call): Likewise.
1558 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1559 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1560 * tree-tailcall.c (find_tail_calls): Likewise.
1561 * tree.c (need_assembler_name_p): Likewise.
1562 (free_lang_data_in_decl): Likewise.
1563 (get_call_combined_fn): Likewise.
1564 * ubsan.c (is_ubsan_builtin_p): Likewise.
1565 * varasm.c (incorporeal_function_p): Likewise.
1566 * tree.h (DECL_BUILT_IN): Remove and replace with
1568 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1569 (fndecl_built_in_p): New.
1571 2018-08-27 Martin Liska <mliska@suse.cz>
1573 PR tree-optimization/86847
1574 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1575 Dump also subtree probability.
1576 (switch_decision_tree::do_jump_if_equal): New function.
1577 (switch_decision_tree::emit_case_nodes): Handle special
1578 situations in balanced tree that can be emitted much simpler.
1579 Fix calculation of probabilities that happen in tree expansion.
1580 * tree-switch-conversion.h (struct cluster): Add
1582 (struct simple_cluster): Likewise.
1583 (struct case_tree_node): Add new function has_child.
1584 (do_jump_if_equal): New.
1586 2018-08-27 Martin Liska <mliska@suse.cz>
1588 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1589 Add new argument to bit_test_cluster constructor.
1590 (bit_test_cluster::emit): Set bits really number of values
1592 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1593 probability argument.
1594 * tree-switch-conversion.h (struct bit_test_cluster):
1595 Add m_handles_entire_switch.
1597 2018-08-27 Martin Liska <mliska@suse.cz>
1599 PR tree-optimization/86702
1600 * tree-switch-conversion.c (jump_table_cluster::emit):
1601 Make probabilities even for values in jump table
1602 according to number of cases handled.
1603 (switch_decision_tree::compute_cases_per_edge): Pass
1604 argument to reset_out_edges_aux function.
1605 (switch_decision_tree::analyze_switch_statement): Likewise.
1606 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1609 2018-08-27 Martin Liska <mliska@suse.cz>
1611 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1612 cfun argument explicitly.
1613 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1614 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1615 function gimple_switch_default_bb.
1616 (convert_switch_statements):
1618 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1619 * stmt.c (label_to_block_fn): Use label_to_block and pass
1620 cfun argument explicitly and use gimple_switch_label_bb.
1621 (expand_case): Likewise.
1622 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1623 cfun argument explicitly. Likewise.
1624 (make_edges_bb): Likewise.
1625 (make_cond_expr_edges): Likewise.
1626 (get_cases_for_edge): Likewise.
1627 (make_gimple_switch_edges): Likewise.
1628 (label_to_block_fn): Likewise.
1629 (label_to_block): Likewise.
1630 (make_goto_expr_edges): Likewise.
1631 (make_gimple_asm_edges): Likewise.
1632 (main_block_label): Likewise.
1633 (group_case_labels_stmt): Likewise.
1634 (find_taken_edge_computed_goto): Likewise.
1635 (find_taken_edge_switch_expr): Likewise.
1636 (gimple_verify_flow_info): Likewise.
1637 (gimple_redirect_edge_and_branch): Likewise.
1638 (gimple_switch_label_bb): New function.
1639 (gimple_switch_default_bb): Likewise.
1640 (gimple_switch_edge): Likewise.
1641 (gimple_switch_default_edge): Likewise.
1642 * tree-cfg.h (label_to_block_fn): Remove and replace ...
1643 (label_to_block): ... with this.
1644 (gimple_switch_label_bb): New.
1645 (gimple_switch_default_bb): Likewise.
1646 (gimple_switch_edge): Likewise.
1647 (gimple_switch_default_edge): Likewise.
1648 * tree-cfgcleanup.c (convert_single_case_switch): Use
1649 new gimple functions and pass new argument to label_to_block.
1650 (cleanup_control_flow_bb):
1651 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1652 cfun argument explicitly.
1653 (make_eh_edges): Likewise.
1654 (redirect_eh_dispatch_edge): Likewise.
1655 (lower_resx): Likewise.
1656 (lower_eh_dispatch): Likewise.
1657 (maybe_remove_unreachable_handlers): Likewise.
1658 (unsplit_eh): Likewise.
1659 (cleanup_empty_eh): Likewise.
1660 (verify_eh_edges): Likewise.
1661 (verify_eh_dispatch_edge): Likewise.
1662 * tree-ssa-dom.c (record_edge_info): Likewise.
1663 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1664 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1665 (thread_through_normal_block): Likewise.
1666 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1667 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
1668 * tree-switch-conversion.c (switch_conversion::collect): Use new
1670 (switch_conversion::check_final_bb): Likewise.
1671 (switch_conversion::gather_default_values): Pass new argument
1673 (switch_conversion::build_constructors): Likewise.
1674 (switch_decision_tree::compute_cases_per_edge): Use new
1675 gimple_switch_edge function.
1676 (switch_decision_tree::analyze_switch_statement): Pass new argument
1678 (switch_decision_tree::try_switch_expansion): Use
1679 gimple_switch_default_edge.
1680 * tree-vrp.c (find_switch_asserts): Pass new argument
1682 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
1683 (vr_values::simplify_switch_using_ranges): Likewise.
1685 2018-08-27 Richard Biener <rguenther@suse.de>
1687 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
1688 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
1690 * tree-ssa-sccvn.h (struct vn_pval): New structure.
1691 (struct vn_nary_op_s): Add unwind_to member. Add
1692 predicated_values flag and put result into a union together
1693 with a linked list of vn_pval.
1694 (struct vn_ssa_aux): Add name member to make maintaining
1695 a map of SSA name to vn_ssa_aux possible. Remove no longer
1696 needed info, dfsnum, low, visited, on_sccstack, use_processed
1697 and range_info_anti_range_p members.
1698 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
1699 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
1701 (vn_valueize): New global.
1702 (vn_context_bb): Likewise.
1703 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
1704 VN_INFO_PTR_INFO): Remove.
1705 * tree-ssa-sccvn.c: ... (rewrite)
1706 (pass_fre::execute): For -O2+ initialize loops and run
1707 RPO VN in optimistic mode (iterating). For -O1 and -Og
1708 run RPO VN in non-optimistic mode.
1709 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
1710 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
1711 * doc/invoke.texi (sccvn-max-scc-size): Remove.
1712 (rpo-vn-max-loop-depth): Document.
1713 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
1714 when valuezing the VUSE signals we walked out of the region.
1715 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
1716 (phi_translate): Set VN context block to use for availability
1718 (compute_avail): Likewise.
1719 (pre_valueize): New function.
1720 (pass_pre::execute): Adjust to the RPO VN API.
1722 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
1723 (propagate_constants_for_unrolling): Remove.
1724 (tree_unroll_loops_completely): Perform value-numbering
1725 on the unrolled bodies loop parent.
1727 2018-08-27 Richard Biener <rguenther@suse.de>
1729 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
1730 for partial antic compute.
1732 2018-08-27 Jakub Jelinek <jakub@redhat.com>
1734 PR rtl-optimization/87065
1735 * combine.c (simplify_if_then_else): Formatting fix.
1736 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
1738 (known_cond): Don't return const_true_rtx for vector modes. Use
1739 CONST0_RTX instead of const0_rtx. Formatting fixes.
1741 2018-08-27 Martin Liska <mliska@suse.cz>
1743 PR gcov-profile/87069
1744 * gcov.c (process_file): Record files already processed
1745 and warn about a file being processed multiple times.
1747 2018-08-27 Martin Liska <mliska@suse.cz>
1750 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1751 Set default values for x_aarch64_*_string strings.
1752 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
1753 prefix. For -mabi do not print '=ABI' in help and use
1754 <option_value> format for -msve-vector-bits and -moverride
1757 2018-08-26 Jeff Law <law@redhat.com>
1759 * config/mips/frame-header-opt.c: Include "backend.h" rather than
1762 2018-08-26 Marek Polacek <polacek@redhat.com>
1764 PR c++/87029, Implement -Wredundant-move.
1765 * doc/invoke.texi: Document -Wredundant-move.
1767 2018-08-25 Martin Sebor <msebor@redhat.com>
1769 PR tree-optimization/87059
1770 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
1771 to the same type as the other.
1772 * fold-const.c (fold_binary_loc): Assert expectation.
1774 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
1776 * config/darwin.c (machopic_legitimize_pic_address): Clean up
1777 extraneous parentheses, dead code section and formatting.
1779 2018-08-24 David Malcolm <dmalcolm@redhat.com>
1782 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1783 wide enough for jumps in the line-numbering to be visible.
1784 (layout::print_gap_in_line_numbering): New member function.
1785 (layout::calculate_line_spans): When using line numbering, merge
1786 line spans that are only 1 line apart.
1787 (diagnostic_show_locus): When printing line numbers, show gaps in
1788 line numbering directly, rather than printing headers.
1789 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1790 line-numbering with multiple line spans.
1791 (selftest::test_fixit_insert_containing_newline_2): Add test of
1792 line-numbering, in which the spans are close enough to be merged.
1794 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1796 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1797 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
1798 * tree-vrp.c (range_is_nonnull): Remove.
1799 (range_includes_zero_p): Accept value_range instead of min/max.
1800 (extract_range_from_binary_expr_1): Do not early bail on
1802 Use range_includes_zero_p instead of range_is_nonnull.
1803 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1804 of range_is_nonnull.
1805 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1806 special case VR_ANTI_RANGE.
1807 (vrp_finalize): Same.
1808 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1810 (range_is_nonnull): Remove.
1811 * vr-values.c (vrp_stmt_computes_nonzero): Use
1812 range_includes_zero_p instead of range_is_nonnull.
1813 (extract_range_basic): Pass value_range to range_includes_zero_p
1814 instead of range_is_nonnull.
1816 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1818 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1819 * emit-rtl.h (rtl_data): Remove return_bnd.
1820 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1821 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1822 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1823 (POINTER_BOUNDS_MODE): Remove definition.
1824 (make_pointer_bounds_mode): Remove.
1825 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1826 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1827 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1828 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1829 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1830 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1831 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1833 * config/i386/i386-modes.def (BND32, BND64): Remove.
1834 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1835 (dbx64_register_map): Ditto.
1836 (svr4_dbx_register_map): Ditto.
1837 (indirect_thunk_bnd_needed): Remove.
1838 (indirect_thunks_bnd_used): Ditto.
1839 (indirect_return_bnd_needed): Ditto.
1840 (indirect_return_via_cx_bnd): Ditto.
1841 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1842 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1843 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1844 (output_indirect_thunk_function): Remove handling of
1845 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1846 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1847 (ix86_save_reg): Remove handling of crtl->return_bnd.
1848 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1849 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1850 and UNSPEC_BNDLX_ADDR.
1851 (ix86_output_indirect_branch_via_reg): Remove handling of
1852 indirect_thunk_prefix_bnd.
1853 (ix86_output_indirect_branch_via_push): Ditto.
1854 (ix86_output_function_return): Ditto.
1855 (ix86_output_indirect_function_return): Ditto.
1856 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1857 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1858 (CALL_USED_REGISTERS): Ditto.
1859 (REG_ALLOC_ORDER): Update for removal of bound registers.
1860 (HI_REGISTER_NAMES): Ditto.
1861 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1862 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1863 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1864 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1865 (FIRST_PSEUDO_REG): Update.
1866 (BND): Remove mode iterator.
1867 * config/i386/predicates.md (bnd_mem_operator): Remove.
1869 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1871 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1874 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1876 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1877 the case in which the permute needs only a single element and
1878 repeats for every vector of the result. Extend that case to
1879 handle variable-length vectors.
1880 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1882 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1885 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1886 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1888 2018-08-24 Richard Biener <rguenther@suse.de>
1890 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1891 bb_flags_allocated members.
1892 (auto_flag): New RAII class for allocating flags.
1893 (auto_edge_flag): New RAII class for allocating edge flags.
1894 (auto_bb_flag): New RAII class for allocating bb flags.
1895 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1897 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1898 in favor of temporarily allocated BB flag.
1899 * hsa-brig.c: Re-order includes.
1900 * hsa-dump.c: Likewise.
1901 * hsa-regalloc.c: Likewise.
1902 * print-rtl.c: Likewise.
1903 * profile-count.c: Likewise.
1905 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1908 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1911 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1914 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1915 from wide_int_range_multiplicative_op.
1917 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1919 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1922 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1924 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1927 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1929 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1930 full GENERIC expression used for replacement.
1932 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1934 * tree-vrp.c (abs_extent_range): Remove.
1935 (extract_range_into_wide_ints): Pass wide ints by reference.
1936 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1937 Pass wide ints by reference in all calls to
1938 extract_range_into_wide_ints.
1939 * wide-int-range.cc (wide_int_range_div): New.
1940 * wide-int-range.h (wide_int_range_div): New.
1941 (wide_int_range_includes_zero_p): New.
1942 (wide_int_range_zero_p): New.
1944 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1946 * config/aarch64/aarch64.md (arches): New enum.
1947 (arch): New enum attr.
1948 (arch_enabled): New attr.
1949 (enabled): Now uses arch_enabled only.
1950 (simd, sve, fp16): Removed attribute.
1951 (fp): Attr now defined in terms of 'arch'.
1952 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1953 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1954 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1955 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1956 attributes into 'arch'.
1957 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1958 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1959 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1960 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1961 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1962 'simd' attribute into 'arch'.
1963 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1964 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1965 Convert use of 'fp' attribute to 'arch'.
1966 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1967 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1969 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1970 (different modes) Merge 'fp' and 'simd' into 'arch'.
1971 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1974 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1976 PR rtl-optimization/87026
1977 * expmed.c (canonicalize_comparison): If we can no longer create
1978 pseudoregisters, don't.
1980 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1983 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1985 * config/arm/arm.c (speculation_barrier_libfunc): New static
1987 (arm_init_libfuncs): Initialize it.
1988 (arm_emit_speculation_barrier): New function.
1989 * config/arm/arm.md (speculation_barrier): Call
1990 arm_emit_speculation_barrier for architectures that do not have
1992 (speculation_barrier_insn): Only match on Armv7 or later.
1994 2018-08-23 Richard Biener <rguenther@suse.de>
1997 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2000 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2002 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2003 of single-vector TBLs.
2004 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2007 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2010 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2011 aarch64_evpc_tbl guard.
2013 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2015 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2018 2018-08-22 Martin Sebor <msebor@redhat.com>
2021 * tree-pretty-print.c (pretty_print_string): Add argument.
2022 (dump_generic_node): Call to pretty_print_string with string size.
2024 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2026 PR rtl-optimization/86771
2027 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2028 of two SETs into those two SETs, one to be placed at i2, if that SETs
2029 destination is modified between i2 and i3.
2031 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2033 PR tree-optimization/86725
2034 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2036 (vect_analyze_scalar_cycles_1): Check it.
2038 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2040 PR tree-optimization/86725
2041 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2042 an outer loop phi as a double reduction, make sure that the
2043 single user of the phi result is an inner loop phi.
2045 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2047 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2048 grouped stores with gaps to a strided group.
2050 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2052 * tree-vect-stmts.c (get_group_load_store_type)
2053 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2054 first statement in a group.
2056 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2058 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2059 the sequence used in gcc/gcc.c.
2061 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2064 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2067 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2069 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2070 Darwin10-specific unwinder-shim.
2071 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2072 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2073 New to cater for Darwin10 Rosetta.
2075 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2077 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2078 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2080 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2085 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2086 (dwarf2out_switch_text_section): Generate a local label for the
2087 second function sub-section and apply it as the second FDE start
2089 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2090 second sub-section start.
2092 2018-08-22 Richard Biener <rguenther@suse.de>
2094 PR tree-optimization/86988
2095 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2097 2018-08-22 Richard Biener <rguenther@suse.de>
2099 PR tree-optimization/86945
2100 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2102 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2104 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2105 a comment about how uses of r2 for .sdata2 come about.
2107 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2109 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2111 2018-08-21 Marek Polacek <polacek@redhat.com>
2113 PR c++/86981, Implement -Wpessimizing-move.
2114 * doc/invoke.texi: Document -Wpessimizing-move.
2116 2018-08-21 Jan Hubicka <jh@suse.cz>
2118 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2119 * tree.h (is_redundant_typedef): Remove.
2120 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2122 2018-08-21 Jan Hubicka <jh@suse.cz>
2124 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2127 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2129 * expmed.c (extract_low_bits): Reject invalid subregs early.
2131 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2134 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2137 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2139 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2140 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2141 * config/vxworks.c: Likewise.
2143 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2145 * config/vxworks.c: Set targetm.have_ctors_dtors
2146 if HAVE_INITFINI_ARRAY_SUPPORT.
2147 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2148 if HAVE_INITFINI_ARRAY_SUPPORT.
2150 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2152 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2153 default search path for VxWorks < 7.
2155 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2157 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2158 (get_format_string): Refer to c_getstr.
2160 2018-08-21 Tom de Vries <tdevries@suse.de>
2162 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2163 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2164 (debuginfo_early_stop): Declare.
2165 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2166 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2167 (debuginfo_early_stop): New function.
2168 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2169 and debuginfo_early_stop.
2170 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2171 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2172 (general_init): Call debuginfo_early_init.
2173 (finalize): Call debuginfo_fini.
2174 (do_compile): Call debuginfo_init.
2175 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2177 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2179 2018-08-21 Tom de Vries <tdevries@suse.de>
2181 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2182 flag_dump_noaddr and flag_dump_unnumbered.
2184 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2186 * wide-int-range.cc (wide_int_range_abs): New.
2187 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2188 * wide-int-range.h (wide_int_range_abs): New.
2189 (wide_int_range_min_max): New.
2190 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2191 case to call wide_int_range_abs.
2192 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2193 (extract_range_from_abs_expr): Delete.
2195 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2198 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2199 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2202 2018-08-20 Nathan Sidwell <nathan@acm.org>
2203 Jeff Law <law@redhat.com>
2205 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2206 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2208 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2211 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2212 (decl_attributes): Likewise.
2213 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2215 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2216 * diagnostic-core.h (class auto_diagnostic_group): New class.
2217 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2218 (diagnostic_report_diagnostic): Handle the first diagnostics within
2220 (emit_diagnostic): Add auto_diagnostic_group instance.
2222 (inform_n): Likewise.
2223 (warning): Likewise.
2224 (warning_at): Likewise.
2225 (warning_n): Likewise.
2226 (pedwarn): Likewise.
2227 (permerror): Likewise.
2229 (error_n): Likewise.
2230 (error_at): Likewise.
2232 (fatal_error): Likewise.
2233 (internal_error): Likewise.
2234 (internal_error_no_backtrace): Likewise.
2235 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2236 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2237 * diagnostic.h (struct diagnostic_context): Add fields
2238 "diagnostic_group_nesting_depth",
2239 "diagnostic_group_emission_count", "begin_group_cb",
2241 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2242 Add auto_diagnostic_group instance(s).
2243 (find_explicit_erroneous_behavior): Likewise.
2244 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2245 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2246 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2247 (gimplify_va_arg_expr): Likewise.
2248 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2249 (HSA_SORRY_AT): Likewise.
2250 * ipa-devirt.c (compare_virtual_tables): Likewise.
2251 (warn_odr): Likewise.
2252 * multiple_target.c (expand_target_clones): Likewise.
2253 * opts-common.c (cmdline_handle_error): Likewise.
2254 * reginfo.c (globalize_reg): Likewise.
2255 * substring-locations.c (format_warning_n_va): Likewise.
2256 * tree-inline.c (expand_call_inline): Likewise.
2257 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2258 * tree-ssa-loop-niter.c
2259 (do_warn_aggressive_loop_optimizations): Likewise.
2260 * tree-ssa-uninit.c (warn_uninit): Likewise.
2261 * tree.c (warn_deprecated_use): Likewise.
2263 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2266 * config/i386/i386.md (eh_return): Always update EH return
2267 address in word_mode.
2269 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2271 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2272 TARGET_SPLIT_COMPLEX_ARG is defined.
2274 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2276 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2278 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2281 * expr.c (expand_assignment): Assert that bitpos is positive.
2282 (store_field): Likewise
2283 (expand_expr_real_1): Make sure that bitpos is positive.
2284 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2287 2018-08-20 Nathan Sidwell <nathan@acm.org>
2289 * Makefile.in (CPP_ID_DATA_H): Delete.
2290 (CPP_INTERNAL_H): Don't add it.
2291 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2292 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2294 2018-08-20 Richard Biener <rguenther@suse.de>
2296 PR tree-optimization/78655
2297 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2298 pointer + offset nonnull if either operand is nonnull work.
2300 2018-08-20 Tom de Vries <tdevries@suse.de>
2302 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2303 unless the referenced die describes the added property using
2304 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2305 Otherwise, add a DW_AT_location to the referenced die.
2307 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2310 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2311 register_operand when calling ix86_set_reg_reg_cost.
2312 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2313 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2314 predicate and to 1 otherwise.
2316 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2318 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2319 emit a diagnostic that it is not supported and reset the option.
2320 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2321 supported and consume the option. (ASM_FINAL_SPEC): New.
2323 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2325 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2328 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2330 C-SKY port: Documentation
2332 * doc/extend.texi (C-SKY Function Attributes): New section.
2333 * doc/invoke.texi (Option Summary): Add C-SKY options.
2334 (C-SKY Options): New section.
2335 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2337 2018-08-17 Jojo <jijie_rong@c-sky.com>
2338 Huibin Wang <huibin_wang@c-sky.com>
2339 Sandra Loosemore <sandra@codesourcery.com>
2340 Chung-Lin Tang <cltang@codesourcery.com>
2342 C-SKY port: Backend implementation
2344 * config/csky/*: New.
2345 * common/config/csky/*: New.
2347 2018-08-17 Jojo <jijie_rong@c-sky.com>
2348 Huibin Wang <huibin_wang@c-sky.com>
2349 Sandra Loosemore <sandra@codesourcery.com>
2350 Chung-Lin Tang <cltang@codesourcery.com>
2351 Andrew Jenner <andrew@codesourcery.com>
2353 C-SKY port: Configury
2355 * config.gcc (csky-*-*): New.
2356 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2357 * configure: Regenerated.
2359 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2361 * dump-context.h: Include "dumpfile.h".
2362 (dump_context::dump_printf_va): Convert final param from va_list
2363 to va_list *. Convert from ATTRIBUTE_PRINTF to
2364 ATTRIBUTE_GCC_DUMP_PRINTF.
2365 (dump_context::dump_printf_loc_va): Likewise.
2366 * dumpfile.c: Include "stringpool.h".
2367 (make_item_for_dump_printf_va): Delete.
2368 (make_item_for_dump_printf): Delete.
2369 (class dump_pretty_printer): New class.
2370 (dump_pretty_printer::dump_pretty_printer): New ctor.
2371 (dump_pretty_printer::emit_items): New member function.
2372 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2374 (dump_pretty_printer::emit_item): New member function.
2375 (dump_pretty_printer::stash_item): New member function.
2376 (dump_pretty_printer::format_decoder_cb): New member function.
2377 (dump_pretty_printer::decode_format): New member function.
2378 (dump_context::dump_printf_va): Reimplement in terms of
2379 dump_pretty_printer.
2380 (dump_context::dump_printf_loc_va): Convert final param from va_list
2382 (dump_context::begin_scope): Reimplement call to
2383 make_item_for_dump_printf.
2384 (dump_printf): Update for change to dump_printf_va.
2385 (dump_printf_loc): Likewise.
2386 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2387 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2388 with %T, %E, and %G.
2389 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2390 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2391 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2392 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2393 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2394 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2395 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2396 within a dump_printf_loc call to "%wu".
2397 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2398 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2399 missing space after "=".
2400 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2401 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2402 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2403 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2404 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2405 duplicate "vectorized" from message.
2407 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2409 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2410 polyNxK_t element's TYPE_STRING_FLAG.
2412 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2414 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2415 (they were unnamed before). Fix comments.
2417 2018-08-17 Nathan Sidwell <nathan@acm.org>
2419 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2421 2018-08-17 Richard Biener <rguenther@suse.de>
2423 PR tree-optimization/86841
2424 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2426 2018-08-17 Martin Liska <mliska@suse.cz>
2428 * common.opt: Remove Warn, Init and Report for options with
2429 Ignore/Deprecated flag. Warning is done automatically for
2431 * config/i386/i386.opt: Likewise.
2432 * config/ia64/ia64.opt: Likewise.
2433 * config/rs6000/rs6000.opt: Likewise.
2434 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2435 Remove usage of flag_check_pointer_bounds.
2436 * lto-wrapper.c (merge_and_complain): Do not handle
2437 OPT_fcheck_pointer_bounds.
2438 (append_compiler_options): Likewise.
2439 * opt-functions.awk: Do not handle Deprecated.
2440 * optc-gen.awk: Check that Var, Report and Init are not
2441 used for an option with Ignore/Deprecated flag.
2442 * opts-common.c (decode_cmdline_option): Do not report
2444 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2446 * opts.h (struct cl_option): Remove cl_deprecated flag.
2447 (CL_ERR_DEPRECATED): Remove error enum value.
2449 2018-08-17 Richard Biener <rguenther@suse.de>
2452 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2453 across a va-arg-pack using call adjust its return value accordingly.
2455 2018-08-16 Martin Sebor <msebor@redhat.com>
2457 PR tree-optimization/86853
2458 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2459 (struct fmtresult): Add member and initialize it in ctors.
2460 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2461 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2462 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2463 (parse_directive): Handle %C same as %c.
2464 (sprintf_dom_walker::compute_format_length): Adjust.
2465 (is_call_safe): Adjust.
2467 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2469 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2470 for determining how to count the elements.
2471 * builtins.h (c_strlen): Adjust prototype.
2472 * expr.c (string_constant): Add new parameter mem_size.
2473 Set *mem_size appropriately.
2474 * expr.h (string_constant): Adjust protoype.
2475 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2476 * gimple-fold.h (get_range_strlen): Adjust prototype.
2477 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2478 (format_string): Call get_string_length with eltsize.
2480 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2482 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2483 to emit the span, rather than setting it as the prefix.
2485 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2487 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2488 "margin_char" parameter, defaulting to space. Use it in place
2489 of pp_space for the initial part of the margin.
2490 (layout::print_leading_fixits): Use '+' when filling the margin
2491 of line-insertion fix-it hints.
2493 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2495 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2498 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2500 * config/rs6000/altivec.md: Don't set length attribute to the default
2502 * config/rs6000/darwin.md: Ditto.
2503 * config/rs6000/dfp.md: Ditto.
2504 * config/rs6000/htm.md: Ditto.
2505 * config/rs6000/rs6000.md: Ditto.
2506 * config/rs6000/sync.md: Ditto.
2507 * config/rs6000/vsx.md: Ditto.
2509 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2511 * config/rs6000/altivec.md: Don't set length attribute to the default
2512 value, for branch instructions.
2513 * config/rs6000/darwin.md: Ditto.
2514 * config/rs6000/rs6000.md: Ditto.
2516 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2518 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2519 (unnamed conditional branch define_insn): Set length to 4 or 8
2520 depending on offset.
2521 (<bd>_<mode>): Similar, for alternative 0.
2522 (<bd>tf_<mode>): Ditto.
2524 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2526 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2528 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2530 * doc/rtl.texi: Replace old RTX class names with new names.
2533 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2535 * expmed.h (canonicalize_comparison): New declaration.
2536 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2537 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2538 * optabs.c (prepare_cmp_insn): Likewise.
2539 * rtl.h (unsigned_condition_p): New function which checks if a
2540 comparison operator is unsigned.
2542 2018-08-16 Nathan Sidwell <nathan@acm.org>
2544 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2545 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2547 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2550 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2551 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2552 (mov<mov>): ..this and enable unconditionally.
2554 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2556 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2558 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2560 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2561 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2562 "Common" with "Target".
2564 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2566 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2567 * doc/invoke.texi (mmitigate-rop): Remove.
2568 * config/i386/i386.c: Do not include "regrename.h".
2569 (ix86_rop_should_change_byte_p, reg_encoded_number)
2570 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2572 (ix86_reorg): Remove call to ix86_mitigate_rop.
2573 * config/i386/i386.md (attr "modrm_class"): Remove.
2574 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2575 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2576 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2578 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2580 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2581 allow folding of mergeh() and mergel() for the float and double types.
2582 (fold_mergehl_helper): Rework to handle building a permute tree
2585 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2587 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2590 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2592 * common.opt (fdiagnostics-show-labels): New option.
2593 * diagnostic-show-locus.c (class layout_range): Add field
2595 (class layout): Add field "m_show_labels_p".
2596 (layout_range::layout_range): Add param "label" and use it to
2598 (make_range): Pass in NULL for new "label" param of layout_range's
2600 (layout::layout): Initialize m_show_labels_p.
2601 (layout::maybe_add_location_range): Pass in loc_range->m_label
2602 when constructing layout_range instances.
2603 (struct line_label): New struct.
2604 (layout::print_any_labels): New member function.
2605 (layout::print_line): Call it if label-printing is enabled.
2606 (selftest::test_one_liner_labels): New test.
2607 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2608 * diagnostic.c (diagnostic_initialize): Initialize
2609 context->show_labels_p.
2610 * diagnostic.h (struct diagnostic_context): Add field
2612 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2613 -fno-diagnostics-show-labels.
2614 * dwarf2out.c (gen_producer_string): Add
2615 OPT_fdiagnostics_show_labels to the ignored options.
2616 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2618 (gcc_rich_location::maybe_add_expr): Likewise.
2619 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2620 label" param, defaulting to NULL.
2621 (gcc_rich_location::add_expr): Add "label" param.
2622 (gcc_rich_location::maybe_add_expr): Likewise.
2623 (class text_range_label): New class.
2624 (class range_label_for_type_mismatch): New class.
2625 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2626 of format_warning_va.
2627 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2628 * lto-wrapper.c (merge_and_complain): Add
2629 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2630 (append_compiler_options): Likewise to the dropped options.
2631 (append_diag_options): Likewise to the passed-on options.
2632 * opts.c (common_handle_option): Handle the new option.
2633 * selftest-diagnostic.c
2634 (test_diagnostic_context::test_diagnostic_context): Enable
2636 * substring-locations.c: Include "gcc-rich-location.h".
2637 (format_warning_n_va): Add "fmt_label" and "param_label" params
2638 and use them as appropriate.
2639 (format_warning_va): Add "fmt_label" and "param_label" params,
2640 passing them on to format_warning_n_va.
2641 (format_warning_at_substring): Likewise.
2642 (format_warning_at_substring_n): Likewise.
2643 * substring-locations.h (format_warning_va): Add "fmt_label" and
2644 "param_label" params.
2645 (format_warning_n_va): Likewise.
2646 (format_warning_at_substring): Likewise.
2647 (format_warning_at_substring_n): Likewise.
2648 * toplev.c (general_init): Initialize global_dc->show_labels_p.
2650 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
2653 * builtins.c (expand_builtin_memcmp): Do not expand the call
2654 when overflow is detected.
2656 2018-08-15 Martin Sebor <msebor@redhat.com>
2658 PR tree-optimization/71625
2659 * config/aarch64/aarch64-builtins.c
2660 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2662 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
2664 * config/s390/s390.c (s390_reorg): Remove loop.
2666 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2669 (darwin_function_switched_text_sections): Delete.
2670 * gcc/config/darwin.h
2671 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2673 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2676 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2677 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2678 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2680 2018-08-15 Martin Liska <mliska@suse.cz>
2682 PR tree-optimization/86925
2683 * predict.c (expr_expected_value_1): When taking
2684 later predictor, assign also probability.
2685 Use fold_build2_initializer_loc in order to fold
2686 the expression in -frounding-math.
2688 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
2690 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
2692 (expand_vec_perm_1): Try the new method.
2694 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
2697 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
2698 Check whether lra_live_max_point is 0 before dividing.
2700 2018-08-14 Martin Sebor <msebor@redhat.com>
2702 PR tree-optimization/86650
2703 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
2704 (vrp_prop::check_mem_ref): Same.
2706 2018-08-13 Liu Hao <lh_mouse@126.com>
2708 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2709 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2710 and clear it thereafter, as it only works for DBCS.
2712 2018-08-13 Liu Hao <lh_mouse@126.com>
2714 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2715 handle returned by _get_osf_handle().
2717 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2719 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
2720 for folding vec_perm.
2722 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2724 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
2725 Add support for gimple-folding of vec_pack() and vec_unpack()
2728 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2730 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
2731 vec_xst variants to the list.
2732 (rs6000_gimple_fold_builtin): Add support for folding unaligned
2733 vector loads and stores.
2735 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
2737 * config.gcc (rs6000-ibm-aix4.x): Delete.
2738 (rs6000-ibm-aix5.1): Delete.
2739 (rs6000-ibm-aix5.2): Delete.
2740 (rs6000-ibm-aix5.3): Delete.
2741 * config/rs6000/aix43.h: Delete.
2742 * config/rs6000/aix51.h: Delete.
2743 * config/rs6000/aix52.h: Delete.
2744 * config/rs6000/t-aix43: Delete.
2746 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
2748 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2750 (s390_decompose_address): Factor out constant pool ref
2753 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2755 * config/nds32/nds32-predicates.c
2756 (nds32_can_use_bclr_p): Change return type as bool.
2757 (nds32_can_use_bset_p): Ditto.
2758 (nds32_can_use_btgl_p): Ditto.
2759 (nds32_can_use_bitci_p): Ditto.
2760 * config/nds32/nds32-protos.h
2761 (nds32_can_use_bclr_p): Change declaration.
2762 (nds32_can_use_bset_p): Ditto.
2763 (nds32_can_use_btgl_p): Ditto.
2764 (nds32_can_use_bitci_p): Ditto.
2766 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2768 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
2769 Support -msched-prolog-epilog option.
2770 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
2772 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2774 * common/config/nds32/nds32-common.c
2775 (nds32_option_optimization_table): Enalbe -malways-align.
2777 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2779 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2781 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2782 OPT_misr_secure_ case.
2783 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2784 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2785 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2787 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2788 (nds32_isr_info): New field security_level.
2789 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2790 * config/nds32/nds32.md (return_internal): Consider critical attribute.
2791 * config/nds32/nds32.opt (misr-secure): New option.
2792 * config/nds32/nds32_init.inc: New file.
2793 * config/nds32/nds32_isr.h: New file.
2795 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2797 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2798 Update comment for atomic instructions.
2799 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2800 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2802 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2803 load for __ATOMIC_SEQ_CST model.
2804 (atomic_loaddi_1): Remove float register target.
2805 (atomic_storedi): Handle CONST_INT values.
2806 (atomic_storedi_1): Remove float register source. Add special case
2808 (memory_barrier): New expander and insn.
2810 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2812 PR tree-optimization/86835
2813 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2814 new_stmt after def_gsi, make sure to insert new_square_stmt after
2815 that stmt, not 2 stmts before it.
2817 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2820 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2823 2018-08-10 Martin Liska <mliska@suse.cz>
2826 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2828 * builtins.c (expand_builtin_expect_with_probability):
2830 (expand_builtin_expect_with_probability): New function.
2831 (build_builtin_expect_predicate): Add new argumnet probability
2832 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2833 (fold_builtin_expect):
2836 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2837 * builtins.h (fold_builtin_expect): Set new argument.
2838 * doc/extend.texi: Document __builtin_expect_with_probability.
2839 * doc/invoke.texi: Likewise.
2840 * gimple-fold.c (gimple_fold_call): Pass new argument.
2841 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2842 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2843 * predict.c (get_predictor_value): New function.
2844 (expr_expected_value): Add new argument probability. Assume
2845 that predictor and probability are always non-null.
2846 (expr_expected_value_1): Likewise. For __builtin_expect and
2847 __builtin_expect_with_probability set probability. Handle
2848 combination in binary expressions.
2849 (tree_predict_by_opcode): Simplify code by simply calling
2850 get_predictor_value.
2851 (pass_strip_predict_hints::execute): Add handling of
2852 BUILT_IN_EXPECT_WITH_PROBABILITY.
2853 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2855 * tree.h (DECL_BUILT_IN_P): New function.
2857 2018-08-10 Martin Liska <mliska@suse.cz>
2859 PR tree-optimization/85799
2860 * passes.def: Add argument for pass_strip_predict_hints.
2861 * predict.c (class pass_strip_predict_hints): Add new argument
2863 (strip_predictor_early): New function.
2864 (pass_strip_predict_hints::execute): Call the function to
2866 (strip_predict_hints): New function.
2867 * predict.def: Fix comment.
2869 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2871 * Makefile.in: Clarify which tm.texi to copy over to assert the
2872 right to grant a GFDL license for all.
2874 2018-08-09 Jeff Law <law@redhat.com>
2876 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2879 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2881 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2884 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2886 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2887 reductions for variable-length vectors.
2889 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2892 * common.opt (fdiagnostics-show-line-numbers): New option.
2893 * diagnostic-show-locus.c (class layout): Add fields
2894 "m_show_line_numbers_p" and "m_linenum_width";
2895 (num_digits): New function.
2896 (test_num_digits): New function.
2897 (layout::layout): Initialize new fields. Update m_x_offset
2898 logic to handle any left margin.
2899 (layout::print_source_line): Print line number when requested.
2900 (layout::start_annotation_line): New member function.
2901 (layout::print_annotation_line): Call it.
2902 (layout::print_leading_fixits): Likewise.
2903 (layout::print_trailing_fixits): Likewise. Update calls to
2904 move_to_column for new parameter.
2905 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2906 it to potentially call start_annotation_line.
2907 (layout::show_ruler): Call start_annotation_line.
2908 (selftest::test_line_numbers_multiline_range): New selftest.
2909 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2910 and selftest::test_line_numbers_multiline_range.
2911 * diagnostic.c (diagnostic_initialize): Initialize
2912 show_line_numbers_p.
2913 * diagnostic.h (struct diagnostic_context): Add field
2914 "show_line_numbers_p".
2915 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2916 -fno-diagnostics-show-line-numbers.
2917 * dwarf2out.c (gen_producer_string): Add
2918 OPT_fdiagnostics_show_line_numbers to the ignored options.
2919 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2920 one setting" options.
2921 (append_compiler_options): Likewise to the dropped options.
2922 (append_diag_options): Likewise to the passed-on options.
2923 * opts.c (common_handle_option): Handle the new option.
2924 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2926 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2928 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2929 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2930 third argument of type "const signed char" to descriptions of
2931 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2932 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2933 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2934 __builtin_bcdsub_ov functions.
2936 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2938 PR tree-optimization/86858
2939 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2940 flow_bb_inside_loop_p calls.
2942 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2944 PR tree-optimization/86871
2945 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2946 instead of gimple_assign_lhs.
2948 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2951 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2952 register constraint to operand 0.
2953 (add<mode>3_carryinC): Likewise.
2954 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2956 2018-08-09 Martin Liska <mliska@suse.cz>
2959 * common.opt: Remove extra line.
2961 2018-08-09 Martin Liska <mliska@suse.cz>
2963 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2964 at the end of a line, make first letter capital and end up
2965 a sentence with a dot.
2966 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2967 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2968 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2969 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2970 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2971 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2972 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2973 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2974 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2975 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2976 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2977 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2978 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2979 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2980 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2981 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2982 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2983 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2985 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2988 * config/s390/s390.c (s390_option_override_internal): Reduce the
2989 stack-clash-protection-probe-interval param if it would be too big
2992 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2995 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2996 * config/m68k/m68k.c (handle_move_double): Don't call
2997 m68k_final_prescan_insn.
2998 (m68k_adjust_decorated_operand): Renamed from
2999 m68k_final_prescan_insn, remove first and third operand and
3001 (print_operand): Call it.
3002 (print_operand_address): Call it.
3004 2018-08-08 Nathan Sidwell <nathan@acm.org>
3006 * diagnostic.c (diagnostic_report_current_module): Use
3007 linemap_included_from & linemap_included_from_linemap.
3009 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3011 * config/aarch64/aarch64-cores.def: Add phecda core.
3012 * config/aarch64/aarch64-tune.md: Regenerate.
3013 * doc/invoke.texi: Add phecda core.
3015 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3018 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3020 * config/s390/s390.md ("movti"): Add more alternatives for
3021 constant to GPR copies.
3023 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3025 * config/s390/s390.c: Fix whitespace damage throughout the file.
3026 * config/s390/s390.h: Likewise.
3027 * config/s390/tpf.h: Likewise.
3029 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3031 * config/s390/s390.c (s390_loadrelative_operand_p):
3032 Remove TARGET_CPU_ZARCH usages.
3033 (s390_rtx_costs): Likewise.
3034 (s390_legitimate_constant_p): Likewise.
3035 (s390_cannot_force_const_mem): Likewise.
3036 (legitimate_reload_constant_p): Likewise.
3037 (s390_preferred_reload_class): Likewise.
3038 (legitimize_pic_address): Likewise.
3039 (legitimize_tls_address): Likewise.
3040 (s390_split_branches): Removed.
3041 (s390_add_execute): Removed.
3042 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3043 (s390_mainpool_start): Likewise.
3044 (s390_mainpool_finish): Likewise.
3045 (s390_mainpool_cancel): Removed.
3046 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3047 (s390_chunkify_cancel): Likewise.
3048 (s390_return_addr_rtx): Likewise.
3049 (s390_register_info): Remove split_branches_pending_p uages.
3050 (s390_optimize_register_info): Likewise.
3051 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3052 split_branches_pending_p usages.
3053 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3054 (s390_load_got): Likewise.
3055 (s390_expand_split_stack_prologue): Likewise.
3056 (output_asm_nops): Likewise.
3057 (s390_function_profiler): Likewise.
3058 (s390_emit_call): Likewise.
3059 (s390_conditional_register_usage): Likewise.
3060 (s390_optimize_prologue): Likewise.
3061 (s390_reorg): Remove TARGET_CPU_ZARCH and
3062 split_branches_pending_p usages.
3063 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3065 (s390_output_indirect_thunk_function): Likewise.
3066 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3067 (TARGET_CPU_ZARCH_P): Removed.
3068 (struct machine_function): Remove split_branches_pending_p.
3069 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3071 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3073 * common/config/s390/s390-common.c (processor_flags_table):
3075 * config.gcc: Remove with_arch/with_tune support.
3076 * config/s390/2064.md: Remove cpu attribute comparisons.
3077 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3079 * config/s390/linux.h (ASM_SPEC):
3080 Remove -march support.
3081 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3082 Use a table to get an arch level.
3083 * config/s390/s390-opts.h (enum processor_type):
3085 * config/s390/s390.c
3086 (processor_table): Remove entries, add arch_level values.
3087 (s390_issue_rate): Remove cases.
3088 (s390_option_override): Adjust
3089 s390_option_override_internal() call.
3090 (s390_option_override_internal): Remove deprecation warning.
3091 (s390_valid_target_attribute_tree): Adjust
3092 s390_option_override_internal() call.
3093 * config/s390/s390.h (struct s390_processor):
3094 Share with s390-c.c, add arch_level field.
3095 * config/s390/s390.md:
3096 Remove occurrences in cpu attribute.
3097 * config/s390/s390.opt: Remove -march/-mtune support.
3098 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3099 * doc/invoke.texi: Remove deprecation warning.
3101 2018-08-08 Luis Machado <luis.machado@linaro.org>
3103 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3105 (qdf24xx_tunings): Set vector cost structure to
3106 qdf24xx_vector_cost.
3108 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3109 <register_sextend>: Set to 3.
3111 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3114 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3115 * config/aarch64/aarch64-simd.md
3116 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3117 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3118 (aarch64_frecpx<mode>): ...this new pattern.
3119 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3120 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3122 2018-08-07 Martin Liska <mliska@suse.cz>
3125 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3126 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3127 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3128 * doc/extend.texi: Document that malloc attribute adds
3131 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3134 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3135 Define to speculation_safe_value_not_needed.
3137 2018-08-06 Jeff Law <law@redhat.com>
3139 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3140 the vr_values instance to cprop_into_stmt.
3141 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3142 (cprop_operand): Also query EVRP to determine if OP is a constant.
3144 2018-08-06 Nathan Sidwell <nathan@acm.org>
3146 * diagnostic.c (diagnostic_report_current_module): Reroll
3147 included-at loop. Translate text.
3149 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3151 * function-tests.c (selftest::test_expansion_to_rtl): Call
3152 free_after_compilation.
3154 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3156 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3158 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3160 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3161 loops with memory block operations from getting unrolled.
3163 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3166 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3167 Define to speculation_safe_value_not_needed.
3169 2018-08-06 Jeff Law <law@redhat.com>
3171 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3174 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3177 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3178 with all enabled __intN types.
3180 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3182 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3184 * alias.c (record_set): Check for clobber high.
3185 * cfgexpand.c (expand_gimple_stmt): Likewise.
3186 * combine-stack-adj.c (single_set_for_csa): Likewise.
3187 * combine.c (find_single_use_1): Likewise.
3188 (set_nonzero_bits_and_sign_copies): Likewise.
3189 (get_combine_src_dest): Likewise.
3190 (is_parallel_of_n_reg_sets): Likewise.
3191 (try_combine): Likewise.
3192 (record_dead_and_set_regs_1): Likewise.
3193 (reg_dead_at_p_1): Likewise.
3194 (reg_dead_at_p): Likewise.
3195 * dce.c (deletable_insn_p): Likewise.
3196 (mark_nonreg_stores_1): Likewise.
3197 (mark_nonreg_stores_2): Likewise.
3198 * df-scan.c (df_find_hard_reg_defs): Likewise.
3199 (df_uses_record): Likewise.
3200 (df_get_call_refs): Likewise.
3201 * dwarf2out.c (mem_loc_descriptor): Likewise.
3202 * haifa-sched.c (haifa_classify_rtx): Likewise.
3203 * ira-build.c (create_insn_allocnos): Likewise.
3204 * ira-costs.c (scan_one_insn): Likewise.
3205 * ira.c (equiv_init_movable_p): Likewise.
3206 (rtx_moveable_p): Likewise.
3207 (interesting_dest_for_shprep): Likewise.
3208 * jump.c (mark_jump_label_1): Likewise.
3209 * postreload-gcse.c (record_opr_changes): Likewise.
3210 * postreload.c (reload_cse_simplify): Likewise.
3211 (struct reg_use): Add source expr.
3212 (reload_combine): Check for clobber high.
3213 (reload_combine_note_use): Likewise.
3214 (reload_cse_move2add): Likewise.
3215 (move2add_note_store): Likewise.
3216 * print-rtl.c (print_pattern): Likewise.
3217 * recog.c (decode_asm_operands): Likewise.
3218 (store_data_bypass_p): Likewise.
3219 (if_test_bypass_p): Likewise.
3220 * regcprop.c (kill_clobbered_value): Likewise.
3221 (kill_set_value): Likewise.
3222 * reginfo.c (reg_scan_mark_refs): Likewise.
3223 * reload1.c (maybe_fix_stack_asms): Likewise.
3224 (eliminate_regs_1): Likewise.
3225 (elimination_effects): Likewise.
3226 (mark_not_eliminable): Likewise.
3227 (scan_paradoxical_subregs): Likewise.
3228 (forget_old_reloads_1): Likewise.
3229 * reorg.c (find_end_label): Likewise.
3230 (try_merge_delay_insns): Likewise.
3231 (redundant_insn): Likewise.
3232 (own_thread_p): Likewise.
3233 (fill_simple_delay_slots): Likewise.
3234 (fill_slots_from_thread): Likewise.
3235 (dbr_schedule): Likewise.
3236 * resource.c (update_live_status): Likewise.
3237 (mark_referenced_resources): Likewise.
3238 (mark_set_resources): Likewise.
3239 * rtl.c (copy_rtx): Likewise.
3240 * rtlanal.c (reg_referenced_p): Likewise.
3241 (single_set_2): Likewise.
3242 (noop_move_p): Likewise.
3243 (note_stores): Likewise.
3244 * sched-deps.c (sched_analyze_reg): Likewise.
3245 (sched_analyze_insn): Likewise.
3247 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3249 * cse.c (invalidate_reg): New function extracted from...
3250 (invalidate): ...here.
3251 (canonicalize_insn): Check for clobber high.
3252 (invalidate_from_clobbers): invalidate clobber highs.
3253 (invalidate_from_sets_and_clobbers): Likewise.
3254 (count_reg_usage): Check for clobber high.
3255 (insn_live_p): Likewise.
3256 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3257 (cselib_invalidate_regno): Check for clobber in setter.
3258 (cselib_invalidate_rtx): Pass through setter.
3259 (cselib_invalidate_rtx_note_stores):
3260 (cselib_process_insn): Check for clobber high.
3261 * cselib.h (cselib_invalidate_rtx): Add operand.
3263 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3265 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3266 (mark_not_eliminable): Likewise.
3267 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3268 * lra-lives.c (process_bb_lives): Check for clobber high.
3269 * lra.c (new_insn_reg): Remember clobber highs.
3270 (collect_non_operand_hard_regs): Check for clobber high.
3271 (lra_set_insn_recog_data): Likewise.
3272 (add_regs_to_insn_regno_info): Likewise.
3273 (lra_update_insn_regno_info): Likewise.
3275 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3277 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3278 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3280 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3282 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3283 (copy_insn_1): Likewise.
3284 (gen_hard_reg_clobber_high): New gen function.
3285 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3286 * genemit.c (gen_exp): Likewise.
3287 (gen_emit_seq): Pass through info.
3288 (gen_insn): Check for CLOBBER_HIGH.
3289 (gen_expand): Pass through info.
3290 (gen_split): Likewise.
3291 (output_add_clobbers): Likewise.
3292 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3293 (remove_clobbers): Likewise.
3294 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3296 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3298 * doc/rtl.texi (clobber_high): Add.
3299 (parallel): Add in clobber high
3300 * rtl.c (rtl_check_failed_code3): Add function.
3301 * rtl.def (CLOBBER_HIGH): Add expression.
3302 * rtl.h (RTL_CHECKC3): Add macro.
3303 (rtl_check_failed_code3): Add declaration.
3304 (XC3EXP): Add macro.
3306 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3309 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3310 cfun->machine->max_used_stack_alignment if needed.
3312 2018-08-04 Martin Sebor <msebor@redhat.com>
3314 PR tree-optimization/86571
3315 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3318 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3321 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3324 2018-08-03 Jeff Law <law@redhat.com>
3327 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3328 Define to speculation_safe_value_not_needed.
3330 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3332 * doc/gcov.texi (-x): Remove duplicate "to".
3333 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3334 (-Wif-not-aligned): Remove duplicate "is".
3335 (-flto): Remove duplicate "the".
3336 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3337 duplicate "v5.00.b".
3338 (MSP430 Options): Remove duplicate "and" from the description
3339 of "-mgprel-sec=regexp".
3340 (x86 Options): Remove duplicate copies of "vmldLog102" and
3341 vmlsLog104 from description of "-mveclibabi=type".
3343 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3345 * internal-fn.h (first_commutative_argument): Declare.
3346 * internal-fn.c (first_commutative_argument): New function.
3347 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3348 restrictions for pattern statements. Use first_commutative_argument
3349 to look for commutative operands in calls to internal functions.
3351 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3353 * Makefile.in (wide-int-range.o): New.
3354 * tree-vrp.c: Move all the wide_int_* functions to...
3355 * wide-int-range.cc: ...here.
3356 * tree-vrp.h: Move all the wide_int_* prototypes to...
3357 * wide-int-range.h: ...here.
3359 2018-08-03 Tom de Vries <tdevries@suse.de>
3361 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3363 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3364 * except.c (output_function_exception_table): Do early exit if
3365 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3367 2018-08-03 Martin Liska <mliska@suse.cz>
3369 * predict.c (dump_prediction): Change to 2 digits
3372 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3374 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3376 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3378 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3380 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3381 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3382 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3383 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3384 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3386 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3388 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3389 pp_take_prefix when saving the existing prefix.
3390 * diagnostic.c (diagnostic_append_note): Likewise.
3391 * langhooks.c (lhd_print_error_function): Likewise.
3392 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3393 param's type. Free the existing prefix.
3394 (pp_take_prefix): New function.
3395 (pretty_printer::pretty_printer): Drop the prefix parameter.
3396 Rename the length parameter to match the comment.
3397 (pretty_printer::~pretty_printer): Free the prefix.
3398 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3400 (struct pretty_printer): Drop the "const" from "prefix" field's
3401 type and clarify memory management.
3402 (pp_set_prefix): Drop the "const" from the 2nd param.
3403 (pp_take_prefix): New decl.
3405 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3407 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3408 for word_mode_ok here instead of passing as argument.
3409 (expand_block_compare): Change select_block_compare_mode() call.
3410 (expand_strncmp_gpr_sequence): New function.
3411 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3413 2018-08-02 Jeff Law <law@redhat.com>
3416 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3417 Define to speculation_safe_value_not_needed.
3420 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3421 Define to speculation_safe_value_not_needed.
3423 2018-08-02 Tom de Vries <tdevries@suse.de>
3426 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3427 function. Return UI_TARGET unconditionally.
3428 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3429 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3431 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3433 * genemit.c (print_overload_test): Fix typo.
3435 2018-08-02 Richard Biener <rguenther@suse.de>
3437 PR tree-optimization/86816
3438 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3439 which checks for value availability before querying it.
3440 (gvn_uses_equal): Use it.
3441 (same_succ_hash): Likewise.
3442 (gimple_equal_p): Likewise.
3444 2018-08-02 Nick Clifton <nickc@redhat.com>
3447 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3448 Define to speculation_safe_value_not_needed.
3451 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3452 Define to speculation_safe_value_not_needed.
3455 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3456 Define to speculation_safe_value_not_needed.
3459 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3460 Define to speculation_safe_value_not_needed.
3463 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3464 Define to speculation_safe_value_not_needed.
3467 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3468 Define to speculation_safe_value_not_needed.
3471 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3472 Define to speculation_safe_value_not_needed.
3475 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3476 Define to speculation_safe_value_not_needed.
3479 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3480 speculation_safe_value_not_needed.
3483 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3484 to speculation_safe_value_not_needed.
3486 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3488 * doc/md.texi: Expand the documentation of instruction names
3489 to mention port-local uses. Document '@' in pattern names.
3490 * read-md.h (overloaded_instance, overloaded_name): New structs.
3492 (md_reader::handle_overloaded_name): New member function.
3493 (md_reader::get_overloads): Likewise.
3494 (md_reader::m_first_overload): New member variable.
3495 (md_reader::m_next_overload_ptr): Likewise.
3496 (md_reader::m_overloads_htab): Likewise.
3497 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3498 m_next_overload_ptr and m_overloads_htab.
3499 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3500 (get_mode_token, get_code_token, get_int_token): New functions.
3501 (map_attr_string): Add an optional argument that passes back
3502 the associated iterator.
3503 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3504 (md_reader::handle_overloaded_name, add_overload_instance): New
3506 (apply_iterators): Handle '@' names. Report an error if '@'
3507 is used without iterators.
3508 (initialize_iterators): Initialize the new iterator_group fields.
3509 * genopinit.c (handle_overloaded_code_for)
3510 (handle_overloaded_gen): New functions.
3511 (main): Use them to print declarations of maybe_code_for_* and
3512 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3513 * genemit.c (print_overload_arguments, print_overload_test)
3514 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3515 (main): Use it to print definitions of maybe_code_for_* and
3516 maybe_gen_* functions.
3517 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3518 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3519 instead of explicit mode checks.
3520 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3521 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3522 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3523 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3524 (aarch64_expand_compare_and_swap): Likewise
3525 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3526 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3527 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3528 (aarch64_constant_pool_reload_icode): Delete.
3529 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3530 instead of aarch64_constant_pool_reload_icode. Use
3531 code_for_aarch64_reload_mov instead of explicit mode checks.
3532 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3533 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3534 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3535 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3536 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3537 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3538 (aarch64_atomic_load_op_code): Delete.
3539 (aarch64_emit_atomic_load_op): Likewise.
3540 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3541 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3542 instead of aarch64_emit_atomic_load_op.
3543 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3544 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3545 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3546 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3547 character before the pattern name.
3548 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3549 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3550 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3551 (aarch64_frecps<mode>): Likewise.
3552 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3553 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3554 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3555 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3556 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3558 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3560 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3561 Allow HFmode constants if TARGET_FP_F16INST.
3563 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3566 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3567 No longer check last store for clobber of address register.
3569 2018-08-02 Martin Liska <mliska@suse.cz>
3571 PR gcov-profile/86817
3572 * gcov.c (process_all_functions): New function.
3574 (process_file): Move functions processing to
3575 process_all_functions.
3577 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3579 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3580 "const" to the "gimple *" and "rtx_insn *" parameters.
3581 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3583 (dump_location_t::dump_location_t): Likewise.
3585 2018-08-01 Martin Sebor <msebor@redhat.com>
3587 PR tree-optimization/86650
3588 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3589 rather than a "gcall *". Directly pass the data of interest
3590 to percent_K_format, rather than building a temporary CALL_EXPR
3592 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3593 (gimple_fold_builtin_strncat): Adjust.
3594 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3595 gcall* argument with gimple*.
3596 * gimple-ssa-warn-restrict.c (check_call): Same.
3597 (wrestrict_dom_walker::before_dom_children): Same.
3598 (builtin_access::builtin_access): Same.
3599 (check_bounds_or_overlap): Same
3600 (maybe_diag_overlap): Same.
3601 (maybe_diag_offset_bounds): Same.
3602 * tree-diagnostic.c (default_tree_printer): Move usage of
3603 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3605 * tree-pretty-print.c (percent_K_format): Add argument.
3606 * tree-pretty-print.h: Add argument.
3607 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3608 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3609 (maybe_diag_stxncpy_trunc): Same.
3610 (handle_builtin_stxncpy): Same.
3611 (handle_builtin_strcat): Same.
3613 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3615 * match.pd: Optimise pointer range checks.
3617 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3619 PR tree-optimization/86758
3620 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3621 to remove pattern statements.
3623 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3625 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3626 result of dfs_enumerate_from when constructing stmt_vec_infos,
3627 instead of additionally calling get_loop_body.
3629 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3631 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3633 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3634 When creating an iv, assert that the step is not known to be zero.
3635 (vect_setup_realignment): Update call accordingly.
3636 * tree-vect-stmts.c (vectorizable_store): Likewise.
3637 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
3639 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3641 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3642 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3643 (vectorizable_reduction): Likewise.
3644 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3645 (vect_detect_hybrid_slp_stmts): Likewise.
3646 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3648 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
3650 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3651 (wide_int_set_zero_nonzero_bits): ...this.
3652 (zero_nonzero_bits_from_vr): Rename to...
3653 (vrp_set_zero_nonzero_bits): ...this.
3654 (extract_range_from_multiplicative_op_1): Abstract wide int
3656 (wide_int_range_multiplicative_op): ...here.
3657 (extract_range_from_binary_expr_1): Extract wide int binary
3658 operations into their own functions.
3659 (wide_int_range_lshift): New.
3660 (wide_int_range_can_optimize_bit_op): New.
3661 (wide_int_range_shift_undefined_p): New.
3662 (wide_int_range_bit_xor): New.
3663 (wide_int_range_bit_ior): New.
3664 (wide_int_range_bit_and): New.
3665 (wide_int_range_trunc_mod): New.
3666 (extract_range_into_wide_ints): New.
3667 (vrp_shift_undefined_p): New.
3668 (extract_range_from_multiplicative_op): New.
3669 (vrp_can_optimize_bit_op): New.
3670 * tree-vrp.h (value_range::dump): New.
3671 (wide_int_range_multiplicative_op): New.
3672 (wide_int_range_lshift):New.
3673 (wide_int_range_shift_undefined_p): New.
3674 (wide_int_range_bit_xor): New.
3675 (wide_int_range_bit_ior): New.
3676 (wide_int_range_bit_and): New.
3677 (wide_int_range_trunc_mod): New.
3678 (zero_nonzero_bits_from_bounds): Rename to...
3679 (wide_int_set_zero_nonzero_bits): ...this.
3680 (zero_nonzero_bits_from_vr): Rename to...
3681 (vrp_set_zero_nonzero_bits): ...this.
3682 (range_easy_mask_min_max): Rename to...
3683 (wide_int_range_can_optimize_bit_op): this.
3684 * vr-values.c (simplify_bit_ops_using_ranges): Rename
3685 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
3687 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3689 * tree-vectorizer.h (vect_orig_stmt): New function.
3690 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
3691 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
3692 (vect_create_epilog_for_reduction): Likewise.
3693 (vectorizable_live_operation): Likewise.
3694 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3695 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
3696 * tree-vect-stmts.c (vectorizable_call): Likewise.
3697 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
3699 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3701 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
3703 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
3704 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
3705 (vect_transform_loop): Likewise.
3706 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3708 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3710 * tree-vectorizer.h (vect_schedule_slp): Return void.
3711 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3712 (vect_schedule_slp): Likewise.
3714 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3716 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
3718 (vect_transform_loop): Update calls accordingly. Schedule SLP
3719 instances before the main loop, if any exist.
3721 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3723 PR tree-optimization/86749
3724 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
3725 If the lhs is used in a COND_EXPR, check that it is being used
3726 as the "then" or "else" value.
3728 2018-08-01 Tom de Vries <tdevries@suse.de>
3731 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3732 speculation_safe_value_not_needed.
3734 2018-08-01 Richard Biener <rguenther@suse.de>
3736 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
3739 2018-08-01 Martin Liska <mliska@suse.cz>
3741 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
3742 format how successful transformation is dumped.
3743 (gimple_mod_pow2_value_transform): Likewise.
3744 (gimple_mod_subtract_transform): Likewise.
3745 (gimple_stringops_transform): Likewise.
3747 2018-08-01 Martin Liska <mliska@suse.cz>
3750 * value-prof.c (interesting_stringop_to_profile_p):
3751 Simplify the code and add BUILT_IN_MEMMOVE.
3752 (gimple_stringops_transform): Likewise.
3754 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
3756 * config/aarch64/aarch64-simd.md
3757 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
3758 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
3759 use GPI iterator instead of SI mode.
3761 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3763 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
3764 rs6000_speculation_barrier.
3765 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
3766 new barrier pattern name.
3768 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3770 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
3771 (speculation_barrier): New insn.
3773 2018-08-01 Richard Biener <rguenther@suse.de>
3776 * graphite.h: Include isl/id.h and isl/space.h to allow build
3779 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3782 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3783 mode for COFF targets.
3784 * defaults.h (TARGET_COFF): Define.
3785 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3786 TARGET_COFF): Define.
3787 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3788 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3789 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3791 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3793 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3794 * gimple-streamer-out.c (output_bb): Save it.
3795 * lto-streamer-in.c (input_struct_function_base): Restore
3796 instance discriminator if available. Create map on demand.
3797 * lto-streamer-out.c (output_struct_function_base): Save it if
3799 * final.c (decl_to_instance_map): Document LTO strategy.
3801 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3802 Olivier Hainque <hainque@adacore.com>
3804 * debug.h (decl_to_instance_map_t): New type.
3805 (decl_to_instance_map): Declare.
3806 (maybe_create_decl_to_instance_map): New inline function.
3807 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3808 to track basic block discriminators.
3809 (final_start_function_1): Initialize them.
3810 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3812 (decl_to_instance_map): New variable.
3813 (map_decl_to_instance, maybe_set_discriminator): New functions.
3814 (notice_source_line): Set discriminator.
3816 2018-07-31 Ian Lance Taylor <iant@golang.org>
3818 * targhooks.c (default_have_speculation_safe_value): Add
3821 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3823 * dump-context.h: Include "pretty-print.h".
3824 (dump_context::refresh_dumps_are_enabled): New decl.
3825 (dump_context::emit_item): New decl.
3826 (class dump_context): Add fields "m_test_pp" and
3828 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3829 (temp_dump_context::get_dumped_text): New decl.
3830 (class temp_dump_context): Add field "m_pp".
3831 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3832 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3834 (set_dump_file): Update for above change.
3835 (set_alt_dump_file): Likewise.
3836 (dump_loc): New overload, taking a pretty_printer *.
3837 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3838 to any test pretty-printer.
3839 (make_item_for_dump_gimple_stmt): New function, adapted from
3840 optinfo::add_gimple_stmt.
3841 (dump_context::dump_gimple_stmt): Call it, and use the result,
3842 eliminating the direct usage of dump_file and alt_dump_file in
3843 favor of indirectly using them via emit_item.
3844 (make_item_for_dump_gimple_expr): New function, adapted from
3845 optinfo::add_gimple_expr.
3846 (dump_context::dump_gimple_expr): Call it, and use the result,
3847 eliminating the direct usage of dump_file and alt_dump_file in
3848 favor of indirectly using them via emit_item.
3849 (make_item_for_dump_generic_expr): New function, adapted from
3851 (dump_context::dump_generic_expr): Call it, and use the result,
3852 eliminating the direct usage of dump_file and alt_dump_file in
3853 favor of indirectly using them via emit_item.
3854 (make_item_for_dump_printf_va): New function, adapted from
3855 optinfo::add_printf_va.
3856 (make_item_for_dump_printf): New function.
3857 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3858 and use the result, eliminating the direct usage of dump_file and
3859 alt_dump_file in favor of indirectly using them via emit_item.
3860 (make_item_for_dump_dec): New function.
3861 (dump_context::dump_dec): Call it, and use the result,
3862 eliminating the direct usage of dump_file and alt_dump_file in
3863 favor of indirectly using them via emit_item.
3864 (make_item_for_dump_symtab_node): New function, adapted from
3865 optinfo::add_symtab_node.
3866 (dump_context::dump_symtab_node): Call it, and use the result,
3867 eliminating the direct usage of dump_file and alt_dump_file in
3868 favor of indirectly using them via emit_item.
3869 (dump_context::begin_scope): Reimplement, avoiding direct usage
3870 of dump_file and alt_dump_file in favor of indirectly using them
3872 (dump_context::emit_item): New member function.
3873 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3874 Set up test pretty-printer on the underlying context. Call
3875 refresh_dumps_are_enabled.
3876 (temp_dump_context::~temp_dump_context): Call
3877 refresh_dumps_are_enabled.
3878 (temp_dump_context::get_dumped_text): New member function.
3879 (selftest::verify_dumped_text): New function.
3880 (ASSERT_DUMPED_TEXT_EQ): New macro.
3881 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3882 and then without optinfo enabled. Add uses of
3883 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3884 * dumpfile.h: Update comment for the dump_* API.
3885 * optinfo-emit-json.cc
3886 (selftest::test_building_json_from_dump_calls): Update for new
3887 param for temp_dump_context ctor.
3888 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3889 and "m_owned" field.
3890 (optinfo_item::~optinfo_item): Likewise.
3891 (optinfo::add_item): New member function.
3892 (optinfo::emit): Update comment.
3893 (optinfo::add_string): Delete.
3894 (optinfo::add_printf): Delete.
3895 (optinfo::add_printf_va): Delete.
3896 (optinfo::add_gimple_stmt): Delete.
3897 (optinfo::add_gimple_expr): Delete.
3898 (optinfo::add_tree): Delete.
3899 (optinfo::add_symtab_node): Delete.
3900 (optinfo::add_dec): Delete.
3901 * optinfo.h (class dump_context): New forward decl.
3902 (optinfo::add_item): New decl.
3903 (optinfo::add_string): Delete.
3904 (optinfo::add_printf): Delete.
3905 (optinfo::add_printf_va): Delete.
3906 (optinfo::add_gimple_stmt): Delete.
3907 (optinfo::add_gimple_expr): Delete.
3908 (optinfo::add_tree): Delete.
3909 (optinfo::add_symtab_node): Delete.
3910 (optinfo::add_dec): Delete.
3911 (optinfo::add_poly_int): Delete.
3912 (optinfo_item::optinfo_item): Remove "owned" param.
3913 (class optinfo_item): Remove field "m_owned".
3915 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3918 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3919 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3920 requested variable alignment.
3921 (expand_one_ssa_partition): Likewise.
3922 (expand_one_var): Likewise.
3924 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3926 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3927 to speculation_safe_value_not_needed.
3929 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3931 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3932 * targhooks.c (speculation_safe_value_not_needed): New function.
3933 * target.def (have_speculation_safe_value): Update documentation.
3934 * doc/tm.texi: Regenerated.
3936 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3938 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3939 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3941 (despeculate_copy<ALLI:mode>_insn): New insn.
3942 (despeculate_copyti_insn): New insn.
3943 (despeculate_simple<ALLI:mode>): New insn
3944 (despeculate_simpleti): New insn.
3945 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3947 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3948 aarch64_speculation_safe_value.
3949 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3951 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3953 * config/aarch64/aarch64-speculation.cc: New file.
3954 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3955 before pass_reorder_blocks.
3956 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3958 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3959 X14 and X15 when tracking speculation.
3960 * config/aarch64/aarch64.md (register name constants): Add
3961 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3962 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3963 (speculation_barrier): New insn attribute.
3964 (cmp<mode>): Allow SP in comparisons.
3965 (speculation_tracker): New insn.
3966 (speculation_barrier): Add speculation_barrier attribute.
3967 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3968 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3969 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3971 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3973 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3974 aarch64_track_speculation is true.
3975 (tb<optab><mode>1): Likewise.
3976 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3977 generate CB[N]Z when tracking speculation.
3978 (aarch64_split_compare_and_swap): Likewise.
3979 (aarch64_split_atomic_op): Likewise.
3981 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3983 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3985 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3987 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3988 (speculation_barrier): New insn.
3990 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3992 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3993 * config/arm/arm.md (speculation_barrier): New expand.
3994 (speculation_barrier_insn): New pattern.
3996 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3998 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3999 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4000 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4001 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4003 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4004 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4005 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4006 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4007 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4008 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4009 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4010 * builtins.c (expand_speculation_safe_value): New function.
4011 (expand_builtin): Call it.
4012 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4013 * doc/extend.texi: Document __builtin_speculation_safe_value.
4014 * doc/md.texi: Document "speculation_barrier" pattern.
4015 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4016 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4017 * doc/tm.texi: Regenerated.
4018 * target.def (have_speculation_safe_value, speculation_safe_value): New
4020 * targhooks.c (default_have_speculation_safe_value): New function.
4021 (default_speculation_safe_value): New function.
4022 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4023 (default_speculation_safe_value): Add prototype.
4025 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4027 * dump-context.h (dump_context::dump_loc): New decl.
4028 * dumpfile.c (dump_context::dump_loc): New member function.
4029 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4030 and dump_gimple_stmt.
4031 (dump_context::dump_gimple_expr_loc): Likewise, using
4033 (dump_context::dump_generic_expr_loc): Likewise, using
4035 (dump_context::dump_printf_loc_va): Likewise, using
4037 (dump_context::begin_scope): Explicitly using the global function
4038 "dump_loc", rather than the member function.
4040 2018-07-31 Martin Sebor <msebor@redhat.com>
4042 PR tree-optimization/86741
4043 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4045 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4047 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4048 depend on whether prefetch instructions will be emitted or not.
4049 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4050 will be emitted or not.
4051 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4052 (TARGET_SETMEM_PFD): New macros.
4054 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4056 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4057 (NULL_STMT_VEC_INFO): Delete.
4058 (stmt_vec_info::operator*): Likewise.
4059 (stmt_vec_info::operator gimple *): Likewise.
4060 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4061 of NULL_STMT_VEC_INFO.
4062 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4063 (vect_reassociating_reduction_p): Likewise.
4064 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4065 (vectorizable_store): Likewise.
4066 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4067 (vec_info::free_stmt_vec_infos): Likewise.
4069 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4071 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4072 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4074 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4076 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4077 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4078 (vec_info::free_stmt_vec_info): New private member functions.
4079 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4080 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4081 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4082 set_stmt_vec_info_vec and free_stmt_vec_infos.
4083 * tree-vect-loop.c (_loop_vec_info): Remove call to
4084 set_stmt_vec_info_vec.
4085 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4086 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4087 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4088 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4089 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4090 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4091 with the clearing in the xcalloc.
4092 (stmt_vec_info_vec): Delete.
4093 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4094 (vectorize_loops): Likewise.
4095 (vec_info::~vec_info): Remove argument from call to
4096 free_stmt_vec_infos.
4097 (vec_info::add_stmt): Remove vinfo argument from call to
4100 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4102 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4103 rather than a gimple stmt.
4104 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4105 information for pattern statements when passed the original
4106 statement; instead wait to be passed the pattern statement itself.
4107 Don't call set_vinfo_for_stmt here.
4108 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4109 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4110 stmt_vec_infos here.
4111 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4112 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4113 stmt_vec_infos entry.
4115 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4117 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4118 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4119 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4120 * tree-vect-stmts.c (vectorizable_call): Likewise.
4121 (vectorizable_simd_clone_call): Likewise.
4123 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4125 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4126 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4127 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4128 * tree-vect-loop.c (vect_transform_loop): Likewise.
4129 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4130 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4132 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4134 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4135 (vect_dr_stmt): Delete.
4136 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4137 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4139 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4140 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4141 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4142 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4143 (vect_analyze_data_refs): Likewise.
4144 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4146 (vect_find_same_alignment_drs): Likewise.
4147 (vect_slp_analyze_node_dependences): Update calls accordingly.
4148 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4149 instead of DR_VECT_AUX.
4150 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4151 of a vector data references. Use vec_info::lookup_dr instead of
4153 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4154 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4155 instead of DR_VECT_AUX.
4157 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4159 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4161 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4163 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4164 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4165 (vect_gen_prolog_loop_niters): Likewise.
4167 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4169 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4170 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4171 (vect_known_alignment_in_bytes, vect_dr_behavior)
4172 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4173 instead of data_references. Update calls to other routines for
4174 which the same change has been made.
4175 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4176 dr_vec_infos instead of stmt_vec_infos.
4177 (vect_analyze_data_ref_dependence): Update call accordingly.
4178 (vect_slp_analyze_data_ref_dependence)
4179 (vect_record_base_alignments): Use DR_VECT_AUX.
4180 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4181 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4182 (vector_alignment_reachable_p, vect_get_data_access_cost)
4183 (vect_peeling_supportable, vect_analyze_group_access_1)
4184 (vect_analyze_group_access, vect_analyze_data_ref_access)
4185 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4186 (vect_compile_time_alias, vect_small_gap_p)
4187 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4188 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4189 instead of data_references. Update calls to other routines for
4190 which the same change has been made.
4191 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4192 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4193 (vect_slp_analyze_and_verify_node_alignment)
4194 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4195 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4196 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4198 (_vect_peel_info::dr): Replace with...
4199 (_vect_peel_info::dr_info): ...this new field.
4200 (vect_peeling_hash_get_most_frequent)
4201 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4202 (vect_peeling_hash_get_lowest_cost):
4203 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4204 routines for which the same change has been made.
4205 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4207 * tree-vect-loop-manip.c (get_misalign_in_elems)
4208 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4210 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4211 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4212 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4213 (get_group_load_store_type, get_negative_load_store_type)
4214 (vect_get_data_ptr_increment, vectorizable_store)
4215 (vectorizable_load): Likewise.
4216 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4217 Update calls to other routines for which the same change has been made.
4219 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4221 * tree-vectorizer.h (vec_info::move_dr): New member function.
4222 (dataref_aux): Rename to...
4223 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4224 (_stmt_vec_info::dr_aux): Update accordingly.
4225 (_stmt_vec_info::data_ref_info): Delete.
4226 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4227 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4228 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4229 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4231 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4232 (STMT_VINFO_DR_INFO): New macro.
4233 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4234 (set_dr_misalignment): Update after rename of dataref_aux.
4235 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4236 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4237 initialization of STMT_VINFO_DATA_REF.
4238 * tree-vectorizer.c (vec_info::move_dr): New function.
4239 * tree-vect-patterns.c (vect_recog_bool_pattern)
4240 (vect_recog_mask_conversion_pattern)
4241 (vect_recog_gather_scatter_pattern): Use it.
4242 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4243 the "dr" and "stmt" fields of dr_vec_info instead of
4244 STMT_VINFO_DATA_REF.
4246 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4248 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4249 (is_pattern_stmt_p): Use it.
4250 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4251 on pattern statements.
4253 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4255 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4256 original stmt as a stmt_vec_info rather than a gimple stmt.
4257 (vect_pattern_recog_1): Take the statement directly as a
4258 stmt_vec_info, rather than via a gimple_stmt_iterator.
4259 Update call to vect_mark_pattern_stmts.
4260 (vect_pattern_recog): Update calls accordingly.
4262 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4264 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4265 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4266 a vect_def_type for the first argument.
4267 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4268 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4269 operand if it isn't defined by a vectorized statement.
4270 (vect_build_gather_load_calls): Remove the mask_dt argument and
4271 update calls to vect_get_vec_def_for_stmt_copy.
4272 (vectorizable_bswap): Likewise the dt argument.
4273 (vectorizable_call): Update calls to vectorizable_bswap and
4274 vect_get_vec_def_for_stmt_copy.
4275 (vectorizable_simd_clone_call, vectorizable_assignment)
4276 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4277 (vectorizable_comparison): Update calls to
4278 vect_get_vec_def_for_stmt_copy.
4279 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4281 (vect_get_loop_based_defs): Remove dt argument and update call
4282 to vect_get_vec_def_for_stmt_copy.
4283 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4284 and vect_get_vec_def_for_stmt_copy.
4285 (vectorizable_load): Update calls to vect_build_gather_load_calls
4286 and vect_get_vec_def_for_stmt_copy.
4287 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4288 (vectorizable_reduction, vectorizable_live_operation): Update calls
4289 to vect_get_vec_def_for_stmt_copy.
4291 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4293 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4294 and gimple stmt arguments with a stmt_vec_info.
4295 (vect_record_base_alignments): Update calls accordingly.
4296 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4297 and gimple stmt arguments with a stmt_vec_info.
4298 (vect_build_slp_tree_1): Remove vinfo argument and update call
4299 to vect_record_max_nunits.
4300 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4301 and vect_record_max_nunits.
4303 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4305 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4306 file and take a stmt_vec_info instead of a gimple stmt.
4307 (supportable_widening_operation, vect_finish_replace_stmt)
4308 (vect_finish_stmt_generation, vect_get_store_rhs)
4309 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4310 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4311 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4312 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4313 (vect_create_data_ref_ptr, bump_vector_ptr)
4314 (vect_permute_store_chain, vect_setup_realignment)
4315 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4316 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4317 (vectorizable_reduction, vectorizable_induction)
4318 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4319 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4320 than gimple stmts as arguments.
4321 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4322 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4323 (can_group_stmts_p, vect_check_gather_scatter)
4324 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4325 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4326 (vect_permute_load_chain, vect_shift_permute_load_chain)
4327 (vect_transform_grouped_load)
4328 (vect_record_grouped_load_vectors): Likewise.
4329 * tree-vect-loop.c (vect_fixup_reduc_chain)
4330 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4331 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4332 (vectorizable_reduction, vectorizable_induction)
4333 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4334 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4335 (vect_get_load_store_mask): Likewise.
4336 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4337 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4338 * tree-vect-stmts.c (vect_mark_relevant)
4339 (is_simple_and_all_uses_invariant)
4340 (exist_non_indexing_operands_for_use_p, process_use)
4341 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4342 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4343 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4344 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4345 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4346 (get_negative_load_store_type, get_load_store_type)
4347 (vect_check_load_store_mask, vect_check_store_rhs)
4348 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4349 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4350 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4351 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4352 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4353 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4354 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4355 (supportable_widening_operation): Likewise.
4357 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4359 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4360 a stmt_vec_info instead of a gcall.
4361 (vect_check_gather_scatter): Update call accordingly.
4362 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4364 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4365 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4366 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4367 instead of a gimple stmt.
4368 (vect_transform_loop): Update calls accordingly.
4369 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4370 stmt_vec_infos instead of gimple stmts.
4371 (vect_analyze_slp_instance): Update use accordingly.
4372 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4373 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4374 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4375 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4376 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4377 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4378 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4379 down to subroutines.
4381 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4383 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4384 of the worklist from a vector of gimple stmts to a vector of
4386 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4387 (vect_mark_stmts_to_be_vectorized): Likewise
4389 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4391 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4392 statement before passing it to vect_analyze_stmt.
4393 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4394 the chain of phi vector definitions. Track the exit phi via its
4396 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4397 STMT_VINFO_REDUC_DEF.
4398 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4399 stmt_vec_infos to handle the statement chains.
4400 (vect_get_slp_defs): Record the first statement in the node
4401 using a stmt_vec_info.
4402 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4403 statements here and pass their stmt_vec_info down to subroutines.
4404 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4405 down to vect_finish_stmt_generation.
4406 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4407 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4408 stmt_vec_infos to subroutines.
4409 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4412 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4414 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4415 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4416 (vect_permute_store_chain, vect_setup_realignment)
4417 (vect_permute_load_chain, vect_shift_permute_load_chain)
4418 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4419 stmts internally, and when passing values to other vectorizer routines.
4420 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4421 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4422 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4423 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4424 (vectorizable_reduction, vectorizable_induction)
4425 (vectorizable_live_operation, vect_transform_loop_stmt)
4426 (vect_transform_loop): Likewise.
4427 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4428 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4429 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4430 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4431 (vect_slp_analyze_node_operations_1): Likewise.
4432 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4433 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4434 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4435 (vect_finish_stmt_generation_1, get_group_load_store_type)
4436 (get_load_store_type, vect_build_gather_load_calls)
4437 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4438 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4439 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4440 (vectorizable_store, vectorizable_load, vectorizable_condition)
4441 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4442 (supportable_widening_operation): Likewise.
4443 (vect_get_vector_types_for_stmt): Likewise.
4444 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4446 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4448 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4449 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4450 (vect_permute_store_chain, vect_permute_load_chain)
4451 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4452 repeated stmt_vec_info lookups.
4453 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4454 (vect_update_ivs_after_vectorizer): Likewise.
4455 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4456 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4457 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4458 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4459 (vect_bb_slp_scalar_cost): Likewise.
4460 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4462 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4464 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4465 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4466 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4467 to get gassigns and gcalls, rather than operating on generc gimple
4469 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4470 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4471 (vectorizable_load, vect_analyze_stmt): Likewise.
4472 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4474 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4476 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4477 return stmt_vec_infos rather than gimple stmts. Do not accept
4479 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4481 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4482 Update use of get_later_stmt.
4483 (vect_get_constant_vectors): Update call accordingly.
4484 (vect_schedule_slp_instance): Likewise
4485 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4486 (vect_slp_analyze_instance_dependence): Likewise.
4487 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4489 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4491 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4492 (stmt_info_for_cost::stmt_info): ...this new field.
4493 (add_stmt_costs): Update accordingly.
4494 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4495 (vect_get_known_peeling_cost): Likewise.
4496 (vect_estimate_min_profitable_iters): Likewise.
4497 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4499 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4501 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4502 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4503 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4505 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4507 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4509 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4510 a gimple stmt to a stmt_vec_info.
4511 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4513 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4515 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4516 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4517 (_loop_vec_info::reduction_chains): Likewise.
4518 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4520 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4522 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4524 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4525 a gimple stmt to a stmt_vec_info.
4526 (_stmt_vec_info::next_element): Likewise.
4527 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4528 (vect_slp_analyze_and_verify_node_alignment)
4529 (vect_analyze_group_access_1, vect_analyze_group_access)
4530 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4531 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4532 (vect_supportable_dr_alignment): Update accordingly.
4533 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4534 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4535 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4536 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4537 * tree-vect-slp.c (vect_build_slp_tree_1)
4538 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4539 (vect_split_slp_store_group, vect_analyze_slp_instance)
4540 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4541 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4542 (get_group_load_store_type, get_load_store_type)
4543 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4544 (vect_transform_stmt, vect_remove_stores): Likewise.
4546 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4548 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4550 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4551 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4552 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4553 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4554 (vector_alignment_reachable_p, vect_get_data_access_cost)
4555 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4556 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4557 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4558 (vect_analyze_group_access_1, vect_analyze_group_access)
4559 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4560 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4561 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4562 result of vect_dr_stmt and use the stmt_vec_info instead of
4563 the associated gimple stmt.
4564 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4565 (vect_gen_prolog_loop_niters): Likewise.
4566 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4568 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4570 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4571 to a vec<stmt_vec_info>.
4572 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4573 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4575 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4576 to a vec<stmt_vec_info>.
4577 (bst_traits::value_type, bst_traits::value_type): Likewise.
4578 (bst_traits::hash): Update accordingly.
4579 (vect_get_and_check_slp_defs): Change the stmts parameter from
4580 a vec<gimple *> to a vec<stmt_vec_info>.
4581 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4582 (vect_build_slp_tree): Likewise.
4583 (vect_build_slp_tree_2): Likewise. Update uses of
4584 SLP_TREE_SCALAR_STMTS.
4585 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4586 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4587 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4588 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4589 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4590 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4591 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4592 (vect_get_constant_vectors, vect_get_slp_defs)
4593 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4594 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4595 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4596 instead of gimple stmts.
4597 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4598 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4599 (vect_slp_analyze_instance_dependence): Update uses of
4600 SLP_TREE_SCALAR_STMTS.
4601 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4602 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4603 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4604 (get_initial_defs_for_reduction): Likewise.
4605 (vect_create_epilog_for_reduction): Likewise.
4606 (vectorize_fold_left_reduction): Likewise.
4607 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4608 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4609 (can_vectorize_live_stmts): Likewise.
4611 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4613 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4614 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4615 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4617 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4619 (gather_scalar_reductions): Update after above interface changes.
4620 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4621 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4623 (vect_force_simple_reduction): Likewise.
4624 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4625 LOOP_VINFO_REDUCTIONS.
4626 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4628 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4630 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4631 a gimple stmt to a stmt_vec_info.
4632 * tree-vect-loop.c (vect_active_double_reduction_p)
4633 (vect_force_simple_reduction, vectorizable_reduction): Update
4636 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4638 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4639 vec<gimple *> to a vec<stmt_vec_info>.
4640 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4641 the reduction_phis argument from a vec<gimple *> to a
4643 (vectorizable_reduction): Likewise the phis local variable that
4644 is passed to vect_create_epilog_for_reduction. Update for new type
4645 of SLP_TREE_VEC_STMTS.
4646 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4647 (vectorizable_live_operation): Likewise.
4648 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4649 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4651 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4653 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4654 a gimple stmt to a stmt_vec_info.
4655 (vectorizable_condition, vectorizable_live_operation)
4656 (vectorizable_reduction, vectorizable_induction): Pass back the
4657 vectorized statement as a stmt_vec_info.
4658 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4659 use of STMT_VINFO_VEC_STMT.
4660 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4661 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4662 as stmt_vec_infos rather than gimple stmts.
4663 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4665 (vectorizable_live_operation): Likewise.
4666 (vectorizable_reduction, vectorizable_induction): Likewise,
4667 updating use of STMT_VINFO_VEC_STMT.
4668 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
4669 of STMT_VINFO_VEC_STMT.
4670 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
4671 (vectorizable_simd_clone_call, vectorizable_conversion)
4672 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4673 (vectorizable_store, vectorizable_load, vectorizable_condition)
4674 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
4675 from a gimple stmt to a stmt_vec_info.
4676 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
4677 pointer to a stmt_vec_info to the vectorizable_* routines.
4679 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4681 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
4682 a gimple stmt to a stmt_vec_info.
4683 (is_pattern_stmt_p): Update accordingly.
4684 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
4685 (vect_record_grouped_load_vectors): Likewise.
4686 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4687 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
4688 (vect_model_reduction_cost): Likewise.
4689 (vect_create_epilog_for_reduction): Likewise.
4690 (vectorizable_reduction, vectorizable_induction): Likewise.
4691 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4692 Return the stmt_vec_info for the pattern statement.
4693 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
4694 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
4695 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4696 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
4697 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4698 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
4699 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
4700 (free_stmt_vec_info, vect_is_simple_use): Likewise.
4702 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4704 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
4705 (vect_finish_stmt_generation): Likewise.
4706 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4707 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
4708 (vect_build_gather_load_calls): Use the return value of the above
4709 functions instead of a separate call to vinfo_for_stmt. Use narrow
4710 scopes for the input gimple stmt and wider scopes for the associated
4711 stmt_vec_info. Use vec_info::lookup_def when setting these
4712 stmt_vec_infos from an SSA_NAME definition.
4713 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4714 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4715 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4716 (vectorizable_store, vectorizable_load, vectorizable_condition)
4717 (vectorizable_comparison): Likewise.
4718 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
4719 (vectorizable_reduction): Likewise.
4721 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4723 * tree-vectorizer.h (vect_is_simple_use): Add an optional
4724 stmt_vec_info * parameter before the optional gimple **.
4725 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4726 (process_use, vect_get_vec_def_for_operand_1): Update callers.
4727 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
4728 * tree-vect-loop.c (vectorizable_reduction): Likewise.
4729 (vectorizable_live_operation): Likewise.
4730 * tree-vect-patterns.c (type_conversion_p): Likewise.
4731 (vect_look_through_possible_promotion): Likewise.
4732 (vect_recog_rotate_pattern): Likewise.
4733 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4735 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4737 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
4738 a typedef to a wrapper class.
4739 (NULL_STMT_VEC_INFO): New macro.
4740 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
4741 (stmt_vec_info::operator*): New function.
4742 (stmt_vec_info::operator gimple *): Likewise.
4743 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
4744 (add_stmt_costs): Likewise.
4745 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
4746 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4747 (vect_get_known_peeling_cost): Likewise.
4748 (vect_estimate_min_profitable_iters): Likewise.
4749 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4750 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
4751 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4752 (vectorizable_store, free_stmt_vec_infos): Likewise.
4753 (new_stmt_vec_info): Change return type of xcalloc to
4756 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4758 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
4759 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
4760 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
4761 a single_imm_use-based sequence.
4762 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
4764 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4766 * tree-vectorizer.h (vec_info::lookup_def): Declare.
4767 * tree-vectorizer.c (vec_info::lookup_def): New function.
4768 * tree-vect-patterns.c (vect_get_internal_def): Use it.
4769 (vect_widened_op_tree): Likewise.
4770 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4771 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
4772 (vectorizable_reduction): Likewise.
4773 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
4775 (vect_is_slp_reduction): Update calls accordingly. Use
4776 vec_info::lookup_def.
4777 (vect_is_simple_reduction): Likewise
4778 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
4780 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4782 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4783 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4784 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4786 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4787 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4788 (vect_update_vf_for_slp, vect_analyze_loop_operations)
4789 (vect_is_slp_reduction, vectorizable_induction)
4790 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4791 * tree-vect-patterns.c (vect_init_pattern_stmt):
4792 (vect_determine_min_output_precision_1, vect_determine_precisions)
4793 (vect_pattern_recog): Likewise.
4794 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4795 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4796 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4797 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4798 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4799 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
4800 info field from a loop to a loop_vec_info.
4802 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4804 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4805 (vec_info::add_stmt): Declare.
4806 * tree-vectorizer.c (vec_info::add_stmt): New function.
4807 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4808 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4809 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4810 (vectorizable_induction): Likewise.
4811 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4812 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4813 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4814 (vectorizable_load): Likewise.
4815 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4816 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4817 (vect_recog_gather_scatter_pattern): Likewise.
4818 (append_pattern_def_seq): Likewise. Remove a check that is
4819 performed by add_stmt itself.
4821 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4823 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4824 which make_ssa_name was called with new_stmt before new_stmt
4827 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4829 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4831 (vect_is_slp_reduction): ...here...
4832 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4833 that are already known to be false.
4835 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4837 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4838 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4839 STMT_VINFO_NUM_SLP_USES when it's true.
4840 (vect_free_slp_instance): Add a final_p parameter and pass it to
4842 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4843 (vect_analyze_slp_instance): Likewise.
4844 (vect_slp_analyze_operations): Likewise.
4845 (vect_slp_analyze_bb_1): Likewise.
4846 * tree-vectorizer.c (vec_info): Likewise.
4847 * tree-vect-loop.c (vect_transform_loop): Likewise.
4849 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4851 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4852 function is not called for second and subsequent members of
4855 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4857 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4858 cases for nested loops from here to ...
4859 (vect_create_epilog_for_reduction): ...here. Only call
4860 vect_is_simple_use for inner-loop reductions.
4862 2018-07-31 Martin Liska <mliska@suse.cz>
4864 PR gcov-profile/85338
4865 PR gcov-profile/85350
4866 PR gcov-profile/85372
4867 * profile.c (struct location_triplet): New.
4868 (struct location_triplet_hash): Likewise.
4869 (output_location): Do not output a BB that
4870 is already recorded for a line.
4871 (branch_prob): Use streamed_locations.
4873 2018-07-31 Martin Liska <mliska@suse.cz>
4875 PR gcov-profile/85370
4876 * coverage.c (coverage_begin_function): Do not mark target
4877 clones as artificial functions.
4879 2018-07-31 Martin Liska <mliska@suse.cz>
4881 PR gcov-profile/83813
4882 PR gcov-profile/84758
4883 PR gcov-profile/85217
4884 PR gcov-profile/85332
4885 * profile.c (branch_prob): Do not record GOTO expressions
4886 for GIMPLE statements which locations are already streamed.
4888 2018-07-31 Olivier Hainque <hainque@adacore.com>
4890 * gcc.c (handle_spec_function): Accept a soft_matched_part
4891 argument, as do_spec_1. Pass it down to ...
4892 (eval_spec_function): Accept a soft_matched_part argument,
4893 and pass it down to ...
4894 (do_spec_2): Accept a soft_matched_part argument, and pass
4895 it down to do_spec_1.
4896 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4897 (handle_braces): Update call to handle_spec_function.
4898 (driver::set_up_specs): Update calls to do_spec_2.
4899 (compare_debug_dump_opt_spec_function): Likewise.
4900 (compare_debug_self_opt_spec_function): Likewise.
4902 2018-07-31 Olivier Hainque <hainque@adacore.com>
4904 * common.opt (nolibc): New option.
4905 * doc/invoke.texi (Link Options): Document it.
4906 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4907 * config/alpha/linux.h: Likewise.
4908 * config/arc/elf.h: Likewise.
4909 * config/arm/uclinux-elf.h: Likewise.
4910 * config/arm/unknown-elf.h: Likewise.
4911 * config/avr/avrlibc.h: Likewise.
4912 * config/bfin/bfin.h: Likewise.
4913 * config/bfin/linux.h: Likewise.
4914 * config/bfin/uclinux.h: Likewise.
4915 * config/darwin.h: Likewise.
4916 * config/darwin10.h: Likewise.
4917 * config/darwin12.h: Likewise.
4918 * config/gnu-user.h: Likewise.
4919 * config/lm32/uclinux-elf.h: Likewise.
4920 * config/pa/pa-hpux11.h: Likewise.
4921 * config/pa/pa64-hpux.h: Likewise.
4922 * config/sparc/sparc.h: Likewise.
4924 2018-07-31 Olivier Hainque <hainque@adacore.com>
4926 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4927 undefined variables.
4929 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4932 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4935 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4937 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4938 terminated string literal.
4940 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4942 PR rtl-optimization/85160
4943 * combine.c (is_just_move): New function.
4944 (try_combine): Allow combining two instructions into two if neither of
4945 the original instructions was a move.
4947 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4950 * doc/extend.texi (Global Register Variables): Discourage use of type
4952 (Local Register Variables): Likewise.
4954 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4956 PR tree-optimization/86506
4957 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4959 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4962 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4963 hard_regno, make sure no insn between `from` and `to` clobbers it.
4965 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4966 Tom de Vries <tdevries@suse.de>
4968 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4969 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4970 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4971 PTX_DEFAULT_RUNTIME_DIM.
4972 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4974 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4976 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4977 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4978 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4980 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4982 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4984 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4985 predicate to reflect toc fusion has been deleted.
4986 (toc_fusion_mem_raw): Delete, no longer used.
4987 (toc_fusion_mem_wrapped): Likewise.
4988 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4990 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4991 Delete, no longer used.
4992 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4993 meant to be used for toc fusion.
4994 (rs6000_debug_print_mode): Delete toc fusion debugging.
4995 (rs6000_debug_reg_global): Likewise.
4996 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4997 fusion and secondary reload support that were never used.
4998 (rs6000_option_override_internal): Delete TOC fusion, that was only
4999 partially defined, and it did not work unless you also used the
5001 (rs6000_legitimate_address_p): Delete TOC fusion support.
5002 (rs6000_opt_masks): Likewise.
5003 (fusion_wrap_memory_address): Delete function, no longer used.
5004 (fusion_split_address); Delete TOC fusion support.
5005 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5006 longer used with toc fusion being deleted.
5007 (TARGET_TOC_FUSION_FP): Likewise.
5008 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5010 (toc fusion spliter): Delete TOC fusion support.
5011 (toc_fusionload_<mode>): Likewise.
5012 (toc_fusionload_di): Likewise.
5013 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5014 longer needs to be named. Rename predicate to delete TOC fusion.
5015 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5016 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5017 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5018 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5019 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5021 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5023 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5024 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5025 __int128 in built-in function prototypes.
5026 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5027 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5029 2018-07-27 Martin Sebor <msebor@redhat.com>
5031 PR tree-optimization/86696
5032 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5033 types, including enums.
5034 (handle_char_store): Be prepared for the above function to fail.
5036 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5038 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5039 when optimization level is lower than 2 or optimize for size.
5041 2018-07-26 Martin Sebor <msebor@redhat.com>
5043 PR tree-optimization/86043
5044 PR tree-optimization/86042
5045 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5046 (get_string_cst_length): Rename...
5047 (get_min_string_length): ...to this. Add argument.
5048 (handle_char_store): Extend to handle multi-character stores by
5050 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5051 * tree.h (initializer_zerop): Add argument.
5053 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5056 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5057 declare target to variables if they have always,{to,from,tofrom} map
5060 2018-07-26 Martin Liska <mliska@suse.cz>
5063 * lto-wrapper.c: Add linker_output as prefix
5064 for ltrans_output_file.
5066 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5068 PR rtl-optimization/85805
5069 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5070 value for hard registers if that was written in the same mode.
5072 2018-07-26 Martin Liska <mliska@suse.cz>
5074 PR gcov-profile/86536
5075 * gcov.c (format_gcov): Use printf format %.*f directly
5076 and do not handle special values.
5078 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5080 * common/config/arc/arc-common.c (arc_option_optimization_table):
5081 Update default optimizations for size.
5083 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5085 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5086 * config/arc/constraints.md (CfZ): New constraint.
5087 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5088 (subsf3_fpu): Likewise.
5089 (cmpsf_fpu): Likewise.
5090 (cmpsf_fpu_uneq): Likewise.
5092 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5094 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5096 (arc_is_uncached_mem_p): uncached applies to both the variable and
5099 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5101 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5104 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5106 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5107 field "m_scopes" from vec to auto_vec.
5109 2018-07-25 Martin Liska <mliska@suse.cz>
5111 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5114 2018-07-25 Richard Biener <rguenther@suse.de>
5117 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5118 special wrt context_die late.
5119 (gen_subprogram_die): Re-use DIEs in local scope.
5121 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5123 PR tree-optimization/86644
5124 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5126 2018-07-25 Martin Liska <mliska@suse.cz>
5129 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5130 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5132 2018-07-25 Martin Liska <mliska@suse.cz>
5135 * params.def: Explain ASan abbreviation and provide
5136 a documentation link.
5138 2018-07-24 Martin Sebor <msebor@redhat.com>
5140 PR tree-optimization/86622
5141 PR tree-optimization/86532
5142 * builtins.h (string_length): Declare.
5143 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5144 (check_access): Be prepared for non-constant length ranges.
5145 (string_length): Make extern.
5146 * expr.c (string_constant): Only handle the minor non-constant
5147 array index. Use string_constant to compute the length of
5148 a generic string constant.
5150 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5152 PR tree-optimization/86618
5153 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5154 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5156 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5158 PR tree-optimization/86636
5159 * json.cc (json::object::set): Fix comment. Add assertions.
5160 (json::array::append): Move here from json.h. Add comment and an
5162 (json::string::string): Likewise.
5163 * json.h (json::array::append): Move to json.cc.
5164 (json::string::string): Likewise.
5165 * optinfo-emit-json.cc
5166 (optrecord_json_writer::impl_location_to_json): Assert that we
5167 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5168 wrapper around it. Expand the location once, rather than three
5170 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5171 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5173 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5174 places. Fix some overlong lines.
5176 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5178 * config/aarch64/aarch64-simd.md
5179 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5180 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5181 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5182 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5183 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5184 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5185 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5186 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5187 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5189 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5192 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5193 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5194 and abs_d == d, do the power of two handling if profitable.
5196 2018-07-24 Richard Biener <rguenther@suse.de>
5198 * match.pd: Add BIT_FIELD_REF canonicalizations.
5200 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5203 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5205 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5207 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5208 terminated STRING_CST object.
5210 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5212 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5214 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5216 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5217 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5218 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5219 the elements into a register.
5220 (rs6000_split_v4si_init_di_reg): Delete.
5221 (rs6000_split_v4si_init): Delete.
5222 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5223 (vsx_init_v4si): Rewrite as a define_expand.
5225 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5227 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5228 zero_extend argument from memory): New.
5230 2018-07-22 Martin Sebor <msebor@redhat.com>
5233 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5234 diagnosing calls with unknown arguments unless -Walloca-larger-than
5235 is restricted to less than PTRDIFF_MAX bytes.
5237 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5239 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5241 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5243 * pretty-print.c (text_info::set_location): Remove redundant
5244 "line_table" parameter from call to rich_location::set_range.
5246 2018-07-20 Martin Sebor <msebor@redhat.com>
5249 * builtins.c (expand_builtin_alloca): Adjust.
5250 * calls.c (alloc_max_size): Simplify.
5251 * cgraphunit.c (cgraph_node::expand): Adjust.
5252 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5254 (frame_larger_than_size): Same.
5255 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5256 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5258 * doc/invoke.texi (GCC Command Options): Document option arguments.
5259 Explain byte-size arguments and suffixes.
5260 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5261 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5262 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5263 * doc/options.texi (UInteger): Expand.
5264 (Host_Wide_Int, ByteSize): Document new properties.
5265 * final.c (final_start_function_1): Include sizes in an error message.
5266 * function.c (frame_offset_overflow): Same.
5267 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5268 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5269 Diagnose unbounded alloca calls only for limits of less than
5271 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5272 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5274 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5275 only for limits of less than PTRDIFF_MAX.
5276 * langhooks-def.h (lhd_handle_option): Change function argument
5278 * langhooks.c (lhd_handle_option): Same.
5279 * langhooks.h (handle_option): Same.
5280 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5282 (var_type, var_type_struct): Same.
5283 (var_set): Handle ByteSize flag.
5284 * optc-gen.awk: Add comments to output to ease debugging. Make
5285 use of HOST_WIDE_INT where appropriate.
5286 * opts-gen-save.awk: Use %lx to format unsigned long.
5287 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5288 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5289 arguments. Parse bytes-size suffixes.
5290 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5291 (enum_value_to_arg): Same.
5292 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5293 (handle_option): Adjust.
5294 (generate_option): Change function argument to HOST_WIDE_INT.
5295 (cmdline_handle_error): Adjust.
5296 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5297 (set_option): Change function argument to HOST_WIDE_INT.
5298 (option_enabled): Handle cl_host_wide_int.
5299 (get_option_state): Handle CLVC_SIZE.
5300 (control_warning_option): Same.
5301 * opts.c (common_handle_option): Change function argument to
5302 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5303 OPT_Wvla_larger_than_.
5304 * opts.h (enum cl_var_type): Add an enumerator.
5305 * stor-layout.c (layout_decl): Print a more meaningful warning.
5306 * toplev.c (output_stack_usage): Adjust.
5308 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5310 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5311 call to inline_expand_builtin_string_cmp.
5312 (expand_builtin_strcmp): Likewise.
5313 (expand_builtin_strncmp): Likewise.
5314 (inline_string_cmp): Delete the last parameter, change char_type_node
5315 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5317 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5318 the inlining expansion on target where the type of the call has same or
5319 narrower precision than unsigned char.
5321 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5323 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5324 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5325 * common.opt (fsave-optimization-record): New option.
5326 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5327 * doc/invoke.texi (-fsave-optimization-record): New option.
5328 * dumpfile.c: Include "optinfo-emit-json.h".
5329 (struct kv_pair): Move to coretypes.h.
5330 (optgroup_options): Make non-static.
5331 (dump_context::end_scope): Call
5332 optimization_records_maybe_pop_dump_scope.
5333 * dumpfile.h (optgroup_options): New decl.
5334 * json.cc: New file.
5336 * optinfo-emit-json.cc: New file.
5337 * optinfo-emit-json.h: New file.
5338 * optinfo.cc: Include "optinfo-emit-json.h".
5339 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5340 (optinfo_enabled_p): Check optimization_records_enabled_p.
5341 (optinfo_wants_inlining_info_p): Likewise.
5342 * optinfo.h: Update comment.
5343 * profile-count.c (profile_quality_as_string): New function.
5344 * profile-count.h (profile_quality_as_string): New decl.
5345 (profile_count::quality): New accessor.
5346 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5347 and optinfo_emit_json_cc_tests.
5348 * selftest.h (selftest::json_cc_tests): New decl.
5349 (selftest::optinfo_emit_json_cc_tests): New decl.
5350 * toplev.c: Include "optinfo-emit-json.h".
5351 (compile_file): Call optimization_records_finish.
5352 (do_compile): Call optimization_records_start.
5353 * tree-ssa-live.c: Include optinfo.h.
5354 (remove_unused_scope_block_p): Retain inlining information if
5355 optinfo_wants_inlining_info_p returns true.
5357 2018-07-20 Richard Biener <rguenther@suse.de>
5360 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5361 to cover -flto-partition=none.
5363 2018-07-20 Martin Liska <mliska@suse.cz>
5365 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5366 (get_decl_source_range): Remove unused function.
5368 2018-07-20 Richard Biener <rguenther@suse.de>
5370 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5371 (struct vn_phi_s): Likewise.
5372 (struct vn_reference_s): Likewise.
5373 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5374 for searching the slot of an entry known to be in the hash itself.
5375 (vn_phi_hasher::equal): Likewise.
5376 (vn_reference_hasher::equal): Likewise.
5377 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5379 (optimistic_info, current_info): Remove, keeping only valid_info.
5380 (vn_reference_lookup_1): Remove fallback lookup.
5381 (vn_reference_lookup_2): Likewise.
5382 (vn_nary_op_lookup_1): Likewise.
5383 (vn_phi_lookup): Likewise.
5384 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5386 (vn_reference_insert): Adjust, chain the inserted hash element
5387 at last_inserted_ref.
5388 (vn_reference_insert_pieces): Likewise.
5389 (visit_reference_op_call): Likewise.
5390 (vn_nary_op_insert_into): Chain the inserted hash element at
5392 (vn_nary_op_insert_pieces): Adjust.
5393 (vn_nary_op_insert): Likewise.
5394 (vn_nary_op_insert_stmt): Likewise.
5395 (vn_phi_insert): Adjust, chain the inserted hash element at
5397 (process_scc): Remove clearing and copying the optimistic
5398 table. Instead remove elements inserted during an optimistic
5399 iteration from the single table we maintain.
5400 (init_scc_vn): Adjust.
5401 (free_scc_vn): Likewise.
5402 (sccvn_dom_walker::record_cond): Likewise.
5403 (sccvn_dom_walker::after_dom_children): Likewise.
5405 2018-07-19 Martin Sebor <msebor@redhat.com>
5407 PR tree-optimization/84047
5408 PR tree-optimization/83776
5409 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5410 (check_array_bounds): Call it.
5412 2018-07-19 Martin Sebor <msebor@redhat.com>
5414 * align.h (align_flags): Use member initialization.
5416 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5418 * Makefile.in (OBJS): Add optinfo.o.
5419 * coretypes.h (class symtab_node): New forward decl.
5420 (struct cgraph_node): New forward decl.
5421 (class varpool_node): New forward decl.
5422 * dump-context.h: New file.
5423 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5425 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5426 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5427 (set_alt_dump_file): Likewise.
5428 (dump_context::~dump_context): New dtor.
5429 (dump_gimple_stmt): Move implementation to...
5430 (dump_context::dump_gimple_stmt): ...this new member function.
5431 Add the stmt to any pending optinfo, creating one if need be.
5432 (dump_gimple_stmt_loc): Move implementation to...
5433 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5434 Start a new optinfo and add the stmt to it.
5435 (dump_gimple_expr): Move implementation to...
5436 (dump_context::dump_gimple_expr): ...this new member function.
5437 Add the stmt to any pending optinfo, creating one if need be.
5438 (dump_gimple_expr_loc): Move implementation to...
5439 (dump_context::dump_gimple_expr_loc): ...this new member function.
5440 Start a new optinfo and add the stmt to it.
5441 (dump_generic_expr): Move implementation to...
5442 (dump_context::dump_generic_expr): ...this new member function.
5443 Add the tree to any pending optinfo, creating one if need be.
5444 (dump_generic_expr_loc): Move implementation to...
5445 (dump_context::dump_generic_expr_loc): ...this new member
5446 function. Add the tree to any pending optinfo, creating one if
5448 (dump_printf): Move implementation to...
5449 (dump_context::dump_printf_va): ...this new member function. Add
5450 the text to any pending optinfo, creating one if need be.
5451 (dump_printf_loc): Move implementation to...
5452 (dump_context::dump_printf_loc_va): ...this new member function.
5453 Start a new optinfo and add the stmt to it.
5454 (dump_dec): Move implementation to...
5455 (dump_context::dump_dec): ...this new member function. Add the
5456 value to any pending optinfo, creating one if need be.
5457 (dump_context::dump_symtab_node): New member function.
5458 (dump_context::get_scope_depth): New member function.
5459 (dump_context::begin_scope): New member function.
5460 (dump_context::end_scope): New member function.
5461 (dump_context::ensure_pending_optinfo): New member function.
5462 (dump_context::begin_next_optinfo): New member function.
5463 (dump_context::end_any_optinfo): New member function.
5464 (dump_context::s_current): New global.
5465 (dump_context::s_default): New global.
5466 (dump_scope_depth): Delete global.
5467 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5468 (dump_symtab_node): New function.
5469 (get_dump_scope_depth): Reimplement in terms of dump_context.
5470 (dump_begin_scope): Likewise.
5471 (dump_end_scope): Likewise.
5472 (selftest::temp_dump_context::temp_dump_context): New ctor.
5473 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5474 (selftest::verify_item): New function.
5475 (ASSERT_IS_TEXT): New macro.
5476 (ASSERT_IS_TREE): New macro.
5477 (ASSERT_IS_GIMPLE): New macro.
5478 (selftest::test_capture_of_dump_calls): New test.
5479 (selftest::dumpfile_c_tests): Call it.
5480 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5481 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5482 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5483 descriptive comment.
5484 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5485 (dump_node, dump_bb): Move these unrelated decls.
5486 (class dump_manager): Add leading comment.
5487 * optinfo.cc: New file.
5488 * optinfo.h: New file.
5490 2018-07-19 Michael Collison <michael.collison@arm.com>
5491 Richard Henderson <rth@redhat.com>
5493 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5494 (subti): Handle op1 zero.
5495 (subvti4, usub4ti4): New.
5496 (*sub<GPI>3_compare1_imm): New.
5497 (sub<GPI>3_carryinCV): New.
5498 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5499 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5501 2018-07-19 Michael Collison <michael.collison@arm.com>
5502 Richard Henderson <rth@redhat.com>
5504 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5505 (addti3): Create simpler code if low part is already known to be 0.
5506 (addvti4, uaddvti4): New.
5507 (*add<GPI>3_compareC_cconly_imm): New.
5508 (*add<GPI>3_compareC_cconly): New.
5509 (*add<GPI>3_compareC_imm): New.
5510 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5511 handle constants within this pattern..
5512 (*add<GPI>3_compareV_cconly_imm): New.
5513 (*add<GPI>3_compareV_cconly): New.
5514 (*add<GPI>3_compareV_imm): New.
5515 (add<GPI>3_compareV): New.
5516 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5517 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5518 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5519 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5520 with 'comparison' operator.
5521 (*add<GPI>3_compareV_cconly_imm): Ditto.
5522 (*add<GPI>3_compareV_cconly): Ditto.
5523 (*add<GPI>3_compareV_imm): Ditto.
5524 (add<GPI>3_compareV): Ditto.
5525 (add<mode>3_carryinC): Ditto.
5526 (*add<mode>3_carryinC_zero): Ditto.
5527 (*add<mode>3_carryinC): Ditto.
5528 (add<mode>3_carryinV): Ditto.
5529 (*add<mode>3_carryinV_zero): Ditto.
5530 (*add<mode>3_carryinV): Ditto.
5532 2018-07-19 Michael Collison <michael.collison@arm.com>
5533 Richard Henderson <rth@redhat.com>
5535 * config/aarch64/aarch64-modes.def (CC_V): New.
5536 * config/aarch64/aarch64-protos.h
5537 (aarch64_addti_scratch_regs): Declare
5538 (aarch64_subvti_scratch_regs): Declare.
5539 (aarch64_expand_subvti): Declare.
5540 (aarch64_gen_unlikely_cbranch): Declare
5541 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5542 for signed overflow using CC_Vmode.
5543 (aarch64_get_condition_code_1): Handle CC_Vmode.
5544 (aarch64_gen_unlikely_cbranch): New function.
5545 (aarch64_addti_scratch_regs): New function.
5546 (aarch64_subvti_scratch_regs): New function.
5547 (aarch64_expand_subvti): New function.
5549 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5551 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5553 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5554 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5557 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5560 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5561 address check not strict.
5563 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5565 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5567 (store_pair_lanes<mode>): Likewise.
5568 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5569 enum value 'ADDR_QUERY_LDP_STP_N'.
5570 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5571 (aarch64_print_address_internal): Add declaration.
5572 (aarch64_print_ldpstp_address): Remove.
5573 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5574 (aarch64_print_operand): Change printing of 'y'.
5575 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5576 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5577 'true' rather than '1'.
5578 * config/aarch64/constraints.md (Uml): Likewise.
5579 (Uml): Rename to Umn.
5582 2018-07-19 Richard Biener <rguenther@suse.de>
5584 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5586 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5587 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5588 (vn_reference_hasher): Likewise.
5589 (struct vn_tables_s): Remove obstack and alloc-pool members.
5590 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5591 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5592 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5593 (vn_reference_insert_pieces): Likewise.
5594 (alloc_vn_nary_op_noinit): Adjust.
5595 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5596 (vn_phi_eq): Adjust.
5597 (shared_lookup_phiargs): Remove.
5598 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5599 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5600 (visit_reference_op_call): Likewise.
5601 (copy_nary, copy_phi, copy_reference): Remove.
5602 (process_scc): Rewind the obstack when iterating. Do not
5603 copy the elements to valid_info but just move them from one
5604 hashtable to the other.
5605 (allocate_vn_table): Adjust.
5606 (free_vn_table): Likewise.
5607 (init_scc_vn): Likewise.
5608 (free_scc_vn): Likewise.
5610 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5613 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5614 indirect_return as function type attribute.
5615 (ix86_attribute_table): Change indirect_return to function
5617 * doc/extend.texi: Update indirect_return attribute.
5619 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5621 * wide-int.h (widest2_int): New.
5622 * gimple-fold.c (arith_overflowed_p): Use it.
5623 * tree.h (widest2_int_cst): New.
5624 * tree-vrp.c (wide_int_binop_overflow): Rename from
5625 vrp_int_const_binop.
5626 Rewrite to work on trees.
5627 (extract_range_from_multiplicative_op_1): Abstract code to...
5628 (wide_int_range_min_max): ...here.
5629 (wide_int_range_cross_product): ...and here.
5630 (extract_range_from_binary_expr_1): Abstract overflow code to...
5631 (wide_int_range_mult_wrapping): ...here.
5632 * tree-vrp.h (wide_int_range_cross_product): New.
5633 (wide_int_range_mult_wrapping): New.
5635 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
5636 Julia Koval <julia.koval@intel.com>
5638 * config/i386/x86-tune-costs.h (skylake_memcpy,
5639 skylake_memset): Replace rep_prefix with unrolling for size 512.
5641 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
5644 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5645 comparision with EQ_EXPR in last stmt.
5647 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
5649 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5650 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5651 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5652 previously known as "PowerPC AltiVec Built-in Functions". Move
5653 some material to new subsubsections "PowerPC AltiVec Built-in
5654 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5656 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5657 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5658 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5659 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5661 2018-07-18 Richard Biener <rguenther@suse.de>
5663 PR tree-optimization/86557
5664 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5667 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
5669 * config/s390/s390.c (s390_function_profiler): Generate CFI.
5671 2018-07-17 Jeff Law <law@redhat.com>
5673 * config/arm/arm.c (get_label_padding): Update for recent
5674 changes to label_to_alignment.
5676 PR tree-optimization/86010
5677 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
5679 * config/mips/mips.c (vr4130_align_insns): Update for recent
5680 changes to label_to_alignment.
5682 * config/frv/frv.c (frv_label_align): Update for recent changes
5683 to label_to_alignment.
5685 * config/nios2/nios2.c (nios2_label_align): Update for recent
5686 changes which dropped ALIGN_LABELS_LOG.
5688 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
5690 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
5691 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
5693 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
5695 * config/arc/arc.c (arc_label_align): Use align_labels instead of
5696 deprecated align_labels_log.
5698 2018-07-17 Richard Biener <rguenther@suse.de>
5701 * dwarf2out.c (init_sections_and_labels): Always generate
5702 a debug_line_str_section for early LTO debug.
5703 (dwarf2out_finish): Reset debug_line_str_hash output early.
5704 Bump counter for extra dwarf5 .debug_loc labels to not conflict
5706 (dwarf2out_early_finish): Output debug_line_str.
5708 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5710 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
5711 index register on z196 or later.
5713 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5715 * config/s390/s390.c (s390_default_align): Set default function
5717 (s390_override_options_after_change): Call s390_default align.
5718 (s390_option_override_internal): Call s390_default_align.
5719 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
5721 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5724 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
5725 to _looptemp_ clauses, other than the first two.
5727 2018-07-17 Martin Liska <mliska@suse.cz>
5729 * opts.c: Do not enable OPT_falign_* for -Os.
5731 2018-07-17 Martin Liska <mliska@suse.cz>
5733 * align.h (MAX_CODE_ALIGN): New.
5734 (MAX_CODE_ALIGN_VALUE): New.
5735 * common/config/i386/i386-common.c (ix86_handle_option):
5736 (MAX_CODE_ALIGN): Moved to align.h.
5737 * final.c (MAX_CODE_ALIGN): Likewise.
5738 * opts.c (parse_and_check_align_values):
5739 (MAX_CODE_ALIGN): Likewise.
5740 (MAX_CODE_ALIGN_VALUE): Likewise.
5742 2018-07-17 Martin Liska <mliska@suse.cz>
5744 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
5745 in order to fulfil coding style.
5746 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
5747 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5748 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5749 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
5750 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
5751 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
5752 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
5753 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
5754 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5756 2018-07-17 Martin Liska <mliska@suse.cz>
5758 * align.h: New file.
5759 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
5761 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
5762 align_flags of label_to_alignment.
5763 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
5765 * config/m68k/m68k.c: Do not use removed align_labels_value and
5767 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
5768 (LOOP_ALIGN): Likewise.
5769 (LABEL_ALIGN): Likewise.
5770 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5771 Remove not used macro.
5772 (rs6000_loop_align): Change return type to align_flags.
5773 (rs6000_loop_align_max_skip): Remove.
5774 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
5775 Change return type to align_flags.
5776 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5777 Remove not used macro.
5778 (rs6000_loop_align): Change return type to align_flags.
5779 (rs6000_loop_align_max_skip): Remove.
5780 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5781 * config/rx/rx-protos.h (rx_align_for_label): Make it
5783 * config/rx/rx.c (rx_align_for_label): Change return type
5785 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5787 into align_flags class.
5788 (LABEL_ALIGN): Likewise.
5789 (LOOP_ALIGN): Likewise.
5790 * config/s390/s390.c (s390_label_align): Use align_flags
5792 (s390_asm_output_function_label): Likewise.
5793 * config/sh/sh.c (sh_override_options_after_change):
5794 Use align_flags class directly without macros.
5795 (find_barrier): Likewise.
5796 (barrier_align): Likewise.
5797 (sh_loop_align): Likewise.
5798 * config/spu/spu.c (spu_option_override):
5799 Use align_flags_tuple::get_value instead of removed macros.
5800 (spu_sched_init): Likewise.
5801 * config/spu/spu.h (GTY): Likewise.
5802 * config/visium/visium.c (visium_option_override):
5803 Set "8" as default secondary alignment.
5804 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5805 in order to guarantee secondary alignment of 8.
5806 * coretypes.h: Include align.h header file.
5807 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5808 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5809 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5810 * doc/tm.texi.in: Likewise.
5811 * final.c (struct label_alignment): Remove not used structure.
5812 (LABEL_ALIGN): Change type to align_flags.
5813 (LOOP_ALIGN): Likewise.
5814 (JUMP_ALIGN): Likewise.
5815 (default_loop_align_max_skip): Remove.
5816 (default_label_align_max_skip): Likewise.
5817 (default_jump_align_max_skip): Likewise.
5818 (default_label_align_after_barrier_max_skip):
5819 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5820 (LABEL_TO_MAX_SKIP): Remove.
5821 (label_to_alignment): Return align_flags type instead of integer.
5822 (label_to_max_skip): Remove.
5823 (align_fuzz): Use align_flags type.
5824 (compute_alignments): Use align_flags type and use align_flags::max
5825 to combine multiple alignments.
5826 (grow_label_align): Grow vec instead of C array.
5827 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5828 (shorten_branches): Use align_flags type and use align_flags::max
5829 to combine multiple alignments.
5830 (final_scan_insn_1): Remove usage of secondary alignment that comes
5831 from label alignment, but instead use proper secondary alignment
5832 which is computed in grow_label_align.
5833 * flags.h (struct align_flags_tuple): Move to align.h.
5834 (struct align_flags): Likewise.
5835 (state_align_loops): Rename to align_loops.
5836 (state_align_jumps): Rename to align_jumps.
5837 (state_align_labels): Rename to align_labels.
5838 (state_align_functions): Rename to align_functions.
5839 (align_loops_log): Remove.
5840 (align_jumps_log): Remove.
5841 (align_labels_log): Remove.
5842 (align_functions_log): Remove.
5843 (align_loops_max_skip): Remove.
5844 (align_jumps_max_skip): Remove.
5845 (align_labels_max_skip): Remove.
5846 (align_functions_max_skip): Remove.
5847 (align_loops_value): Remove.
5848 (align_jumps_value): Remove.
5849 (align_labels_value): Remove.
5850 (align_functions_value): Remove.
5851 * output.h (label_to_alignment): Change return type to align_flags.
5852 (label_to_max_skip): Remove.
5853 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5854 jump_align_max_skip macros.
5855 * targhooks.h (default_loop_align_max_skip): Remove.
5856 (default_label_align_max_skip): Likewise.
5857 (default_jump_align_max_skip): Likewise.
5858 (default_label_align_after_barrier_max_skip): Remove.
5859 * toplev.c (read_log_maxskip): Use ::normalize function.
5860 (parse_N_M): Remove not used argument and also call ::normalize.
5861 (parse_alignment_opts): Do not pass unused arguments.
5862 * varasm.c (assemble_start_function): Use directly align_functions
5863 instead of removed macros.
5864 * system.h: Do not poison removed macros.
5866 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5869 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5870 and cond temporaries don't have reference type if iterator has
5871 pointer type. For init use &for_pre_body instead of pre_p if
5872 for_pre_body is non-empty.
5874 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5876 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5877 double-double modes to SFmode directly directly.
5878 (trunc<mode>sf2_fprs): Delete.
5880 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5882 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5883 for conversions between IFmode and the decimal floating point modes.
5884 (init_float128_ieee): Use the correct names for conversions between
5885 KFmode and the decimal floating point modes.
5887 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5889 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5890 for the conversions between TDmode and IFmode.
5891 (init_float128_ieee): Use more correct names for the conversions
5892 between TDmode and KFmode.
5894 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5896 PR tree-optimization/86526
5897 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5898 (inline_expand_builtin_string_cmp): Likewise.
5899 (inline_string_cmp): Likewise. Use c_readstr instead of
5900 builtin_memcpy_read_str. Add unit_mode temporary.
5902 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5905 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5906 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5908 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5910 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5911 Alphabetize prototypes of built-in functions, separating out
5912 built-in functions that are listed in this section but should be
5913 described elsewhere.
5915 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5918 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5919 conditional move when trapping comparison was split to a
5920 non-trapping one (and vice versa).
5922 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5924 * config/s390/s390.c (s390_function_profiler): Generate nops
5925 instead of profiler call sequences.
5926 * config/s390/s390.opt: Add the new option.
5928 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5930 * config/s390/s390.c (s390_function_profiler): Generate
5931 __mcount_loc section.
5932 * config/s390/s390.opt: Add the new option.
5934 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5936 * common.opt: Add the new warning.
5937 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5938 %r0,__fentry__" when -mfentry is specified.
5939 (s390_option_override_internal): Disallow -mfentry for 31-bit
5941 * config/s390/s390.opt: Add the new option.
5943 2018-07-16 Richard Biener <rguenther@suse.de>
5946 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5947 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5949 2018-07-16 Martin Liska <mliska@suse.cz>
5952 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5955 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5957 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5959 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5961 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5963 * config/arc/arc-cpus.def (hs4x): New cpu.
5965 * config/arc/arc-tables.opt: Regenerate.
5966 * config/arc/arc.c (arc_sched_issue_rate): New function.
5967 (TARGET_SCHED_ISSUE_RATE): Define.
5968 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5969 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5971 (attr tune): Add ARCHS4x tune values.
5972 (attr tune_dspmpy): Define.
5973 (*tst): Correct instruction type.
5974 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5975 * config/arc/arcHS4x.md: New file.
5976 * config/arc/fpu.md: Update instruction type attributes.
5977 * config/arc/t-multilib: Regenerate.
5979 2018-07-16 Tom de Vries <tdevries@suse.de>
5982 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5984 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5986 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5989 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5991 PR tree-optimization/86514
5992 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5993 conversion to a boolean type from a type with greater precision.
5995 2018-07-16 Tom de Vries <tdevries@suse.de>
5997 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5998 dump_flags request TDF_SLIM.
6000 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
6002 * fold-const.c (int_const_binop_1): Abstract...
6003 (wide_int_binop): ...wide int code here.
6004 (poly_int_binop): ...poly int code here.
6005 Abstract the rest of int_const_binop_1 into int_const_binop.
6006 * fold-const.h (wide_int_binop): New.
6007 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6008 Remove useless PLUS/MINUS_EXPR case.
6009 (zero_nonzero_bits_from_vr): Move wide int code...
6010 (zero_nonzero_bits_from_bounds): ...here.
6011 (extract_range_from_binary_expr_1): Move mask optimization code...
6012 (range_easy_mask_min_max): ...here.
6013 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6014 (range_easy_mask_min_max): New.
6016 2018-07-15 Jeff Law <law@redhat.com>
6019 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6022 2018-07-14 Jim Wilson <jimw@sifive.com>
6024 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6026 2018-07-14 Paul Koning <ni1d@arrl.net>
6028 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6030 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6032 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6035 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6036 Steve Munroe <munroesj52@gmail.com>
6038 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6039 (_mm_andnot_si128): Likewise.
6040 (_mm_or_si128): Likewise.
6041 (_mm_xor_si128): Likewise.
6043 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6046 * builtins.c (expand_builtin_memcmp): Inline the calls first
6047 when result_eq is false.
6048 (expand_builtin_strcmp): Inline the calls first.
6049 (expand_builtin_strncmp): Likewise.
6050 (inline_string_cmp): New routine. Expand a string compare
6051 call by using a sequence of char comparison.
6052 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6053 a call to str(n)cmp/memcmp.
6054 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6056 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6058 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6060 * config/arm/driver-arm.c: Include arm-native.h.
6061 (host_detect_local_cpu): Use auto-generated data tables.
6062 (vendors, arm_cpu_table): Delete. Move part information to ...
6063 * config/arm/arm-cpus.in: ... here.
6064 * config/arm/parsecpu.awk (gen_native): New function.
6065 (vendor, part): New CPU fields.
6066 (END): Add support for building the native CPU detection tables.
6067 * config/arm/t-arm (arm-native.h): Add build rule.
6068 (driver-arm.o): Add dependency on arm-native.h.
6070 2018-07-13 Richard Biener <rguenther@suse.de>
6073 * match.pd (addr1 - addr2): Allow either of the operand to
6076 2018-07-13 Tom de Vries <tdevries@suse.de>
6078 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6081 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6083 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6084 arrays instead of numbered variables.
6086 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6088 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6089 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6091 2018-07-13 Richard Biener <rguenther@suse.de>
6094 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6095 instead of get_context_die.
6097 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6098 Richard Biener <rguenther@suse.de>
6101 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6102 that the loop latch destination where phi is defined.
6104 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6106 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6107 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6108 (riscv_set_current_function): Extract getting interrupt type to new
6110 (riscv_get_interrupt_type): New function.
6111 (riscv_merge_decl_attributes): New function, checking interrupt type is
6113 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6115 2018-07-12 Paul Koning <ni1d@arrl.net>
6117 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6120 2018-07-12 Paul Koning <ni1d@arrl.net>
6122 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6123 zero reference, add doloop_end instead.
6124 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6125 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6126 detail for doloop_end.
6128 2018-07-12 Martin Sebor <msebor@redhat.com>
6131 * attribs.c (decl_attributes): Reject conflicting attributes before
6132 calling attribute handlers.
6134 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6136 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6138 (gcc::dump_manager::get_dump_file_name): likewise.
6139 (dump_begin): Likewise.
6140 * dumpfile.h (dump_begin): Update prototype.
6141 (gcc::dump_manager::get_dump_file_name,
6142 gcc::dump_manager::get_dump_file_name): Update prototype.
6144 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6146 * internal-fn.h (vectorizable_internal_fn_p): New function.
6147 * tree-vect-slp.c (compatible_calls_p): Likewise.
6148 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6149 to internal functions.
6150 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6152 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6154 * fold-const.h (inverse_conditions_p): Declare.
6155 * fold-const.c (inverse_conditions_p): New function.
6156 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6157 that test the inverse condition of a conditional internal function.
6158 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6159 * internal-fn.c (internal_fn_mask_index): Handle conditional
6161 (vectorized_internal_fn_supported_p): New function.
6162 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6163 (any_pred_load_store): Replace with...
6164 (need_to_predicate): ...this new variable.
6165 (redundant_ssa_names): New variable.
6166 (ifcvt_can_use_mask_load_store): Move initial checks to...
6167 (ifcvt_can_predicate): ...this new function. Handle tree codes
6168 for which a conditional internal function exists.
6169 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6170 instead of ifcvt_can_use_mask_load_store. Update after variable
6172 (predicate_load_or_store): New function, split out from
6173 predicate_mem_writes.
6174 (check_redundant_cond_expr): New function.
6175 (value_available_p): Likewise.
6176 (predicate_rhs_code): Likewise.
6177 (predicate_mem_writes): Rename to...
6178 (predicate_statements): ...this. Use predicate_load_or_store
6179 and predicate_rhs_code.
6180 (combine_blocks, tree_if_conversion): Update after above name changes.
6181 (ifcvt_local_dce): Handle redundant_ssa_names.
6182 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6183 general conditional functions.
6184 * tree-vect-stmts.c (vectorizable_call): Likewise.
6186 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6187 Alan Hayward <alan.hayward@arm.com>
6188 David Sherwood <david.sherwood@arm.com>
6190 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6191 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6192 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6193 plus and minus and convert them into IFN_COND_FMA-based sequences.
6194 (convert_mult_to_fma): Handle conditional plus and minus.
6196 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6198 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6199 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6200 (cond_fnms_optab): New optabs.
6201 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6203 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6204 * internal-fn.h (get_conditional_internal_fn): Declare.
6205 (get_unconditional_internal_fn): Likewise.
6206 * internal-fn.c (cond_ternary_direct): New macro.
6207 (expand_cond_ternary_optab_fn): Likewise.
6208 (direct_cond_ternary_optab_supported_p): Likewise.
6209 (FOR_EACH_COND_FN_PAIR): Likewise.
6210 (get_conditional_internal_fn): New function.
6211 (get_unconditional_internal_fn): Likewise.
6212 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6213 (gimple_match_op::gimple_match_op): Add a new overload for 5
6215 (gimple_match_op::set_op): Likewise.
6216 (gimple_resimplify5): Declare.
6217 * genmatch.c (decision_tree::gen): Generate simplifications for
6219 * gimple-match-head.c (gimple_simplify): Define an overload for
6220 5 operands. Handle calls with 5 arguments in the top-level overload.
6221 (convert_conditional_op): Handle conversions from unconditional
6222 internal functions to conditional ones.
6223 (gimple_resimplify5): New function.
6224 (build_call_internal): Pass a fifth operand.
6225 (maybe_push_res_to_seq): Likewise.
6226 (try_conditional_simplification): Try converting conditional
6227 internal functions to unconditional internal functions.
6228 Handle 3-operand unconditional forms.
6229 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6230 Define ternary equivalents of the current rules for binary conditional
6232 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6234 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6235 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6236 (optab): Handle them.
6237 (SVE_COND_FP_TERNARY): New int iterator.
6238 (sve_fmla_op, sve_fmad_op): New int attributes.
6239 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6240 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6241 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6243 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6245 * target.def (preferred_else_value): New target hook.
6246 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6247 * doc/tm.texi: Regenerate.
6248 * targhooks.h (default_preferred_else_value): Declare.
6249 * targhooks.c (default_preferred_else_value): New function.
6250 * internal-fn.h (conditional_internal_fn_code): Declare.
6251 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6252 (get_conditional_internal_fn): Use it.
6253 (conditional_internal_fn_code): New function.
6254 * gimple-match.h (gimple_match_cond): New struct.
6255 (gimple_match_op): Add a cond member function.
6256 (gimple_match_op::gimple_match_op): Update all forms to take a
6258 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6259 for the suboperation, but don't specify a particular else_value.
6260 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6261 (visit_nary_op, visit_reference_op_load): Pass
6262 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6263 * gimple-match-head.c: Include tree-eh.h
6264 (convert_conditional_op): New function.
6265 (maybe_resimplify_conditional_op): Likewise.
6266 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6267 (gimple_resimplify2): Likewise.
6268 (gimple_resimplify3): Likewise.
6269 (gimple_resimplify4): Likewise.
6270 (maybe_push_res_to_seq): Return null for conditional operations.
6271 (try_conditional_simplification): New function.
6272 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6274 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6276 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6278 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6280 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6282 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6284 (hash_tree): Do not hash DECL_FCONTEXT
6285 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6286 Do not stream DECL_FCONTEXT.
6287 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6288 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6290 2018-07-12 Richard Biener <rguenther@suse.de>
6293 * dwarf2out.c (gen_block_die): Only output blocks when they have
6294 at least one !DECL_IGNORED_P variable.
6296 2018-07-12 Richard Biener <rguenther@suse.de>
6299 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6300 Remove -mieee-fp handling.
6302 2018-07-12 Richard Biener <rguenther@suse.de>
6304 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6305 left-over from last patch.
6307 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6309 PR tree-optimization/86492
6310 * gimple-ssa-store-merging.c
6311 (imm_store_chain_info::coalesce_immediate_stores): Call
6312 check_no_overlap even for the merge_overlapping case. Formatting fix.
6314 2018-07-12 Richard Biener <rguenther@suse.de>
6317 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6318 move possibly trapping operations into the conditional.
6320 2018-07-12 Richard Biener <rguenther@suse.de>
6322 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6323 (vn_lookup_simplify_result): Remove recursion limit applied
6325 (vn_nary_build_or_lookup_1): Adjust.
6326 (try_to_simplify): Likewise.
6327 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6329 (gimple_resimplify2): Likewise.
6330 (gimple_resimplify3): Likewise.
6331 (gimple_resimplify4): Likewise.
6333 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6335 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6336 Use __mmask64 type instead of __mmask8 for __M argument.
6337 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6338 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6340 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6341 __mmask16 for __M argument.
6342 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6343 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6344 to __mmask16 instead of __mmask8.
6345 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6346 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6347 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6348 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6349 instead of __mmask16 for __U argument.
6350 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6351 __mmask16 instead of __mmask8 for __U argument.
6352 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6354 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6356 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6358 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6360 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6362 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6364 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6366 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6368 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6370 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6372 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6373 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6374 return type as well as __M argument type and all casts from __mmask8
6376 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6377 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6378 return type as well as __M argument type and all casts from __mmask8
6380 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6381 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6382 return type as well as __M argument type and all casts from __mmask8
6384 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6385 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6386 return type as well as __M argument type and all casts from __mmask8
6388 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6389 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6392 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6394 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6395 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6398 2018-07-11 Paul Koning <ni1d@arrl.net>
6400 * doc/md.texi (define_subst): Document how multiple occurrences of
6401 the same argument in the replacement pattern are handled.
6403 2018-07-11 Paul Koning <ni1d@arrl.net>
6405 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6407 (Common Type Attributes): Add "mode" attribute.
6409 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6411 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6412 stream DECL_ORIGINAL_TYPE.
6413 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6414 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6415 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6416 Do not walk original type.
6417 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6419 (write_ts_decl_non_common_tree_pointers): Do not stream
6421 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6422 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6424 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6426 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6427 thread twice from the same starting edge.
6429 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6431 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6432 * gimple.c (gimple_call_nonnull_result_p): ...here...
6433 (gimple_call_nonnull_arg): ...and here.
6434 * gimple.h (gimple_call_nonnull_result_p): New.
6435 (gimple_call_nonnull_arg): New.
6437 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6439 * config/arm/arm-cpus.in: Move information from fpu field of each
6440 cpu definition to the isa field.
6441 * config/arm/parsecpu.awk (fpu): Delete match rule.
6442 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6444 2018-07-11 Richard Biener <rguenther@suse.de>
6447 * dwarf2out.c (init_sections_and_labels): Use
6448 output_asm_line_debug_info consistently.
6449 (dwarf2out_early_finish): Likewise.
6450 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6453 2018-07-11 Richard Biener <rguenther@suse.de>
6455 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6456 Rework father_bb setting in a way to avoid propagating constants
6457 multiple times on a loop body.
6459 2018-07-10 Mark Wielaard <mark@klomp.org>
6462 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6464 2018-07-10 Richard Biener <rguenther@suse.de>
6466 * hash-map.h (hash_map::iterator::operator*): Return
6467 references to key and value.
6469 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6472 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6473 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6474 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6475 (gimplify_omp_for): For composite loops, move outer
6476 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6477 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6478 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6479 TREE_LIST for both the original class iterator and the "last" helper
6480 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6481 loop, remember has_decl_expr from outer composite loops for the
6482 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6484 2018-07-09 Martin Sebor <msebor@redhat.com>
6488 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6489 accessing implicitly initialized array elements.
6490 * expr.c (string_constant): Handle string initializers of
6491 character arrays within aggregates.
6492 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6493 Store element offset. As a special case, handle zero size.
6494 (fold_nonarray_ctor_reference): Same.
6495 (fold_ctor_reference): Add argument. Store subobject offset.
6496 * gimple-fold.h (fold_ctor_reference): Add argument.
6498 2018-07-09 Paul Koning <ni1d@arrl.net>
6500 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6501 (pdp11_insn_cost): New function.
6502 (pdp11_md_asm_adjust): New function.
6503 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6504 (pdp11_rtx_costs): Update to match machine better.
6505 (output_addr_const_pdp11): Correct format mismatch warnings.
6506 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6507 * config/pdp11/pdp11.md: General change to add base_cost and/or
6508 length attributes for use by new pdp11_insn_cost function.
6509 (MIN_BRANCH): Correct definition.
6511 (doloop_end): Use standard pattern name for looping pattern.
6512 (doloop_end_nocc): New.
6513 (movsf): Add another constraint alternative.
6514 (zero_extendqihi2): Add constraint alternatives for not in place
6516 (zero_extendhisi2): Remove.
6517 (shift patterns): Add CC handling variants.
6521 (define_peephole2): New peephole to recognize mov that sets CC for
6524 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6526 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6527 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6530 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6533 * real.c (real_nextafter): Return true if result is denormal.
6535 2018-07-09 Martin Liska <mliska@suse.cz>
6537 * common.opt: Add back wrongly removed attribute.
6539 2018-07-09 Richard Biener <rguenther@suse.de>
6542 * dwarf2out.c (gen_block_die): For an early generated DIE
6543 always output high/low PC attributes.
6545 2018-07-09 Tom de Vries <tdevries@suse.de>
6547 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6548 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6549 onto VAR_DECL with abstract origin.
6551 2018-07-07 Jim Wilson <jimw@sifive.com>
6553 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6555 2018-07-07 Tom de Vries <tdevries@suse.de>
6557 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6559 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6561 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6563 (combine_bound): Use wide-int overflow calculation instead of
6565 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6567 * fold-const.c (int_const_binop_2): Same.
6568 (extract_muldiv_1): Same.
6569 (fold_div_compare): Same.
6570 (fold_abs_const): Same.
6572 * poly-int.h (add): Same.
6576 * predict.c (predict_iv_comparison): Same.
6577 * profile-count.c (slow_safe_scale_64bit): Same.
6578 * simplify-rtx.c (simplify_const_binary_operation): Same.
6579 * tree-chrec.c (tree_fold_binomial): Same.
6580 * tree-data-ref.c (split_constant_offset_1): Same.
6581 * tree-if-conv.c (idx_within_array_bound): Same.
6582 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6583 * tree-ssa-phiopt.c (minmax_replacement): Same.
6584 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6585 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6586 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6587 * wide-int.cc (wi::add_large): Same.
6588 (wi::mul_internal): Same.
6589 (wi::sub_large): Same.
6590 (wi::divmod_internal): Same.
6591 * wide-int.h: Change overflow type to overflow_type for neg, add,
6592 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6593 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6594 mul_internal, divmod_internal.
6595 (overflow_type): New enum.
6596 (accumulate_overflow): New.
6598 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6600 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6601 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6603 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6605 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6606 argument is checked for zero before entering loop, avoid checking again.
6608 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6610 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6611 * gimplify.h (generic_expr_could_trap_p): Likwise.
6612 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6614 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6616 PR tree-optimization/86401
6617 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6618 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6619 (fold_bit_and_mask): ... here. New helper function for match.pd.
6620 * fold-const.h (fold_bit_and_mask): Declare.
6621 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6623 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6626 * target.def (translate_mode_attribute): New hook.
6627 * targhooks.h (default_translate_mode_attribute): Declare.
6628 * targhooks.c (default_translate_mode_attribute): New function.
6629 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6630 * doc/tm.texi: Regenerate.
6631 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6632 (rs6000_translate_mode_attribute): New function.
6634 2018-07-06 Paul Koning <ni1d@arrl.net>
6636 * doc/md.texi (define_split): Document DONE and FAIL.
6637 (define_peephole2): Ditto.
6639 2018-07-05 Jeff Law <law@redhat.com>
6641 PR tree-optimization/86010
6642 * tree-ssa-dse.c (compute_trims): More aggressively trim at
6643 both the head and tail of mem* and str* calls.
6645 2018-07-05 Jim Wilson <jimw@sifive.com>
6647 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6649 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
6651 * config/aarch64/aarch64-simd.md: correct flags text for
6652 MIN_EXPR replacement.
6654 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6656 * configure: Regenerated.
6658 2018-07-05 Carl Love <cel@us.ibm.com>
6660 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6661 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6662 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6663 VSX_BUILTIN_DOUBLEL_V4SF.
6665 2018-07-05 Martin Sebor <msebor@redhat.com>
6668 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
6669 than its domain to compute its the upper bound of a char array.
6671 2018-07-05 Nathan Sidwell <nathan@acm.org>
6673 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
6674 * doc/cpp.texi: Update comment.
6675 * doc/tm.texi: Rebuilt.
6676 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
6677 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
6678 * doc/extend.texi (Backwards Compatibility): Clarify it is system
6679 headers affected by extern "C".
6680 * system.h: Poison NO_IMPLICIT_EXTERN_C.
6681 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
6682 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
6683 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
6684 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
6685 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
6686 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
6687 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
6688 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
6689 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
6690 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
6691 config/sparc/sp64-elf.h, config/spu/spu.h,
6692 config/stormy16/stormy16.h, config/v850/v850.h,
6693 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
6694 define NO_IMPLICIT_EXTERN_C.
6695 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
6697 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6700 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6701 instead of GET_MODE_SIZE when comparing Units.
6703 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6706 * rtlanal.c (set_noop_p): Constrain on mode change,
6707 include hard-reg-set.h
6709 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6711 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
6713 2018-07-05 Jakub Jelinek <jakub@redhat.com>
6716 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6719 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6721 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6723 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6726 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6728 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6730 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
6733 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6735 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
6737 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
6738 neg_*_op* variables.
6740 2018-07-04 Martin Liska <mliska@suse.cz>
6742 * tree-switch-conversion.c: Define
6743 max_ratio_for_speed and max_ratio_for_size constants.
6745 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6746 Martin Liska <mliska@suse.cz>
6751 * common.opt: Rename align options with 'str_' prefix.
6752 * common/config/i386/i386-common.c (set_malign_value): New
6754 (ix86_handle_option): Use it to set -falign-* options/
6755 * config/aarch64/aarch64-protos.h (struct tune_params): Change
6756 type from int to string.
6757 * config/aarch64/aarch64.c: Update default values from int
6759 * config/alpha/alpha.c (alpha_override_options_after_change):
6761 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
6762 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6763 max skip conditionally.
6764 * config/i386/freebsd.h (SUBALIGN_LOG): New.
6765 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6766 max skip conditionally.
6767 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6768 max skip conditionally.
6769 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
6770 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6771 max skip conditionally.
6772 * config/i386/i386.c (struct ptt): Change type from int to
6774 (ix86_default_align): Set default values.
6775 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
6776 max skip conditionally.
6777 * config/i386/iamcu.h (SUBALIGN_LOG): New.
6778 (ASM_OUTPUT_MAX_SKIP_ALIGN):
6779 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
6780 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6781 max skip conditionally.
6782 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6783 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6784 * config/i386/x86-64.h (SUBALIGN_LOG): New.
6785 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6786 max skip conditionally.
6787 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6788 * config/ia64/ia64.c (ia64_option_override): Set default values
6789 for alignment options.
6790 * config/m68k/m68k.c: Handle new str_align_* options.
6791 * config/mips/mips.c (mips_set_compression_mode): Change
6793 (mips_option_override): Set default values for options.
6794 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6796 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6798 * config/rx/rx.c (rx_option_override): Likewise.
6799 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6800 (LABEL_ALIGN): Use align_labels_log.
6801 (LOOP_ALIGN): Use align_loops_align.
6802 * config/s390/s390.c (s390_asm_output_function_label): Use new
6804 * config/sh/sh.c (sh_override_options_after_change):
6805 Change type of constants.
6806 * config/spu/spu.c (spu_sched_init): Likewise.
6807 * config/sparc/sparc.c (sparc_option_override): Set default
6809 * config/visium/visium.c (visium_option_override): Likewise.
6810 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6811 emit p2align format with last argument if it's not needed.
6812 * doc/invoke.texi: Document extended format of -falign-*.
6813 * final.c: Use align_labels alignment.
6814 * flags.h (struct target_flag_state): Change type to use
6816 (struct align_flags_tuple): New.
6817 (struct align_flags): Likewise.
6818 (align_loops_log): Redefine macro to use new types.
6819 (align_loops_max_skip): Redefine macro to use new types.
6820 (align_jumps_log): Redefine macro to use new types.
6821 (align_jumps_max_skip): Redefine macro to use new types.
6822 (align_labels_log): Redefine macro to use new types.
6823 (align_labels_max_skip): Redefine macro to use new types.
6824 (align_functions_log): Redefine macro to use new types.
6825 (align_loops): Redefine macro to use new types.
6826 (align_jumps): Redefine macro to use new types.
6827 (align_labels): Redefine macro to use new types.
6828 (align_functions): Redefine macro to use new types.
6829 (align_functions_max_skip): Redefine macro to use new types.
6830 (align_loops_value): New macro.
6831 (align_jumps_value): New macro.
6832 (align_labels_value): New macro.
6833 (align_functions_value): New macro.
6834 * function.c (invoke_set_current_function_hook): Propagate
6835 alignment values from flags to global variables default in
6837 * ipa-icf.c (sem_function::equals_wpa): Use
6838 cl_optimization_option_eq instead of memcmp.
6839 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6841 (cl_optimization_stream_in): Likewise.
6842 * optc-save-gen.awk: Support strings in cl_optimization.
6843 * opth-gen.awk: Likewise.
6844 * opts.c (finish_options): Remove error checking of invalid
6846 (MAX_CODE_ALIGN): Remove.
6847 (MAX_CODE_ALIGN_VALUE): Likewise.
6848 (parse_and_check_align_values): New function.
6849 (check_alignment_argument): Likewise.
6850 (common_handle_option): Use check_alignment_argument.
6851 * opts.h (parse_and_check_align_values): Declare.
6852 * toplev.c (init_alignments): Remove.
6853 (read_log_maxskip): New.
6854 (parse_N_M): Likewise.
6855 (parse_alignment_opts): Likewise.
6856 (backend_init_target): Remove usage of init_alignments.
6857 * toplev.h (parse_alignment_opts): Declare.
6858 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6860 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6861 * tree.c (cl_option_hasher::equal): New.
6862 * varasm.c: Use new global macros.
6864 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6866 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6867 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6868 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6869 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6870 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6871 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6872 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6873 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6874 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6875 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6876 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6877 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6878 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6880 2018-07-04 Martin Liska <mliska@suse.cz>
6881 Jonathan Wakely <jwakely@redhat.com>
6883 * coverage.c: Use correct type.
6884 * doc/invoke.texi: Language correction.
6886 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6889 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6890 ENDBRANCH for non-tail call which may return via indirect branch.
6891 * doc/extend.texi: Document indirect_return attribute.
6893 2018-07-03 Martin Sebor <msebor@redhat.com>
6895 PR tree-optimization/86274
6896 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6898 (format_floating): Correct handling of infinities and NaNs.
6900 2018-07-03 Martin Sebor <msebor@redhat.com>
6902 * print-tree.c (print_real_cst): New function.
6903 (print_node_brief): Call it.
6904 (print_node): Ditto.
6906 2018-07-03 Jeff Law <law@redhat.com>
6908 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6909 into a single pattern.
6911 * config/h8300/h8300.md (ors code_iterator): New.
6912 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6913 a single pattern and single splitter.
6914 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6915 (iorqi3_1, xorqi3_1): Likewise.
6916 (iorqi3, xorqi3 expanders): Similarly.
6918 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6919 (movmd_internal) into a single pattern using the P mode iterator.
6920 (movmd splitters): Similarly.
6921 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6922 (movsd splitters): Similarly.
6924 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6925 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6926 right table based on the size of the operand.
6927 * config/h8300/h8300.md (length_table): Corresponding changes. All
6928 references to "addb", "addw" and "addl" changed to "add".
6929 (btst patterns): Merge two variants into a single pattern.
6930 (tstqi, tsthi): Likewise.
6931 (addhi3_incdec, addsi3_incdec): Likewise.
6932 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6933 (mulhi3, mulsi3): Likewise.
6934 (udivhi3, udivsi3): Likewise.
6935 (divhi3, divsi3): Likewise.
6936 (andorqi3, andorhi3, andorsi3): Likewise.
6938 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6941 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6942 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6944 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6946 PR tree-optimization/85694
6947 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6948 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6950 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6951 (<u>avg<mode>3_ceil): New patterns.
6953 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6955 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6956 scan-tree-dump directive.
6957 * gcc.dg/vect/slp-perm-2.c: Likewise.
6958 * gcc.dg/vect/slp-perm-3.c: Likewise.
6959 * gcc.dg/vect/slp-perm-5.c: Likewise.
6960 * gcc.dg/vect/slp-perm-6.c: Likewise.
6961 * gcc.dg/vect/slp-perm-7.c: Likewise.
6962 * gcc.dg/vect/slp-perm-8.c: Likewise.
6964 2018-07-03 Marek Polacek <polacek@redhat.com>
6967 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6969 2018-07-03 Richard Biener <rguenther@suse.de>
6972 * tree-ssa-structalias.c (find_func_clobbers): Properly
6973 handle indirect calls.
6975 2018-07-03 Jeff Law <law@redhat.com>
6977 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6978 (shifts): New code iterator.
6979 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6980 expander. Fix HImode handling on H8/SX.
6981 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6982 (subqi3, subhi3, subsi3 expanders): Likewise.
6983 (andqi3, andhi3, andsi3 expanders): Likewise.
6984 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6985 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6986 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6987 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6988 (zero_extendqihi2, zero_extendqisi2): Likewise.
6989 (extendqihi2, extendqisi2): Likewise.
6990 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6991 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6992 (rotlqi3_1, rotlhi3_1): Likewise.
6993 (logicalhi3_sn, logicalsi3_sn): Likewise.
6994 (logicalhi3, logicalsi3): Likewise.
6996 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6998 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6999 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7000 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7001 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7002 type to append_pattern_def_seq instead of creating a stmt_vec_info
7004 (build_mask_conversion): Likewise. Remove vinfo argument.
7005 (vect_add_conversion_to_patterm): Likewise, renaming to...
7006 (vect_add_conversion_to_pattern): ...this.
7007 (vect_recog_mask_conversion_pattern): Update call to
7008 build_mask_conversion. Pass the vector type to
7009 append_pattern_def_seq here too.
7010 (vect_recog_gather_scatter_pattern): Update call to
7011 vect_add_conversion_to_pattern.
7013 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7015 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7016 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7017 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7018 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7019 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7020 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7021 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7022 append_pattern_def_seq instead of new_pattern_def_seq.
7023 (vect_recog_divmod_pattern): Do both of the above.
7024 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7027 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7029 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7030 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7031 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7032 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7033 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7034 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7035 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7036 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7037 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7038 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7039 parameter with a single stmt_vec_info.
7040 (vect_recog_func_ptr): Likewise.
7041 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7042 (vect_try_gather_scatter_pattern): ...this.
7043 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7044 the stmt_vec_info of the statement to be matched. Don't clear
7045 STMT_VINFO_RELATED_STMT.
7046 (vect_pattern_recog): Update call accordingly.
7048 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7050 PR tree-optimization/85694
7051 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7052 (uavgM3_ceil): Document new optabs.
7053 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7054 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7056 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7057 (savg_ceil_optab): New optabs.
7058 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7059 (vect_vect_recog_func_ptrs): Add it.
7060 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7061 constant directly from the associated lhs.
7063 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7065 * tree-vect-patterns.c (vect_split_statement): New function.
7066 (vect_convert_input): Use it to try to split an existing cast.
7068 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7070 * poly-int.h (print_hex): New function.
7071 * dumpfile.h (dump_dec, dump_hex): Declare.
7072 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7073 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7074 min_input_precision, operation_precision and operation_sign.
7075 * tree-vect-patterns.c (vect_get_range_info): New function.
7076 (vect_same_loop_or_bb_p, vect_single_imm_use)
7077 (vect_operation_fits_smaller_type): Delete.
7078 (vect_look_through_possible_promotion): Add an optional
7079 single_use_p parameter.
7080 (vect_recog_over_widening_pattern): Rewrite to use new
7081 stmt_vec_info infomration. Handle one operation at a time.
7082 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7083 (vect_truncatable_operation_p, vect_set_operation_type)
7084 (vect_set_min_input_precision): New functions.
7085 (vect_determine_min_output_precision_1): Likewise.
7086 (vect_determine_min_output_precision): Likewise.
7087 (vect_determine_precisions_from_range): Likewise.
7088 (vect_determine_precisions_from_users): Likewise.
7089 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7090 (vect_vect_recog_func_ptrs): Put over_widening first.
7092 (vect_pattern_recog): Call vect_determine_precisions.
7094 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7096 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7097 statements that have been replaced by further pattern statements.
7098 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7100 2018-07-03 Richard Biener <rguenther@suse.de>
7102 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7103 always set *dt. Dump vectype in vectype overload.
7104 * dumpfile.h (dump_gimple_expr): New function.
7105 (dump_gimple_expr_loc): Likewise.
7106 * dumpfile.c (dump_gimple_expr): New function.
7107 (dump_gimple_expr_loc): Likewise.
7109 2018-07-02 Jeff Law <law@redhat.com>
7111 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7112 the H8/300, H8/300H and H8/S variants into a single pattern.
7113 (movhi_h8300, movqi_h8300hs): Similarly.
7114 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7115 (QHI mode iterator): New.
7117 * config/h8300/h8300.md: Remove trailing whitespace.
7119 2018-07-02 Jim Wilson <jimw@sifive.com>
7121 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7122 instead of emit_insn for interrupt returns.
7123 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7124 (riscv_sret, riscv_uret): Likewise.
7126 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7128 * pretty-print.c (selftest::test_pp_format): Move save and restore
7129 of quotes to class auto_fix_quotes, and add an instance.
7130 * selftest.c: Include "intl.h".
7131 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7132 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7133 * selftest.h (selftest::auto_fix_quotes): New class.
7135 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7137 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7138 (aarch64_sve_prepare_conditional_op): Remove.
7139 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7140 Allow aarch64_simd_reg_or_zero as select operand; remove
7141 the aarch64_sve_prepare_conditional_op call.
7142 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7143 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7144 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7145 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7146 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7147 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7148 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7149 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7150 and a splitters to match all of the *_any patterns.
7151 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7153 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7154 (SVE_COND_FP_BINARY_REV): Remove.
7155 (sve_int_op_rev, sve_fp_op_rev): New.
7156 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7157 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7158 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7159 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7160 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7161 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7162 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7163 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7164 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7166 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7167 Remove match_dup 1 from the inner unspec.
7168 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7170 * config/aarch64/aarch64.md (movprfx): New attr.
7171 (length): Default movprfx to 8.
7172 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7173 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7174 (*<su>mul<SVE_I>3_highpart): Likewise.
7175 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7176 (*v<ASHIFT><SVE_I>3): Likewise.
7177 (*<su><MAXMIN><SVE_I>3): Likewise.
7178 (*<su><MAXMIN><SVE_F>3): Likewise.
7179 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7180 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7181 (*div<SVE_F>4): Likewise.
7183 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7185 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7188 2018-07-02 Richard Biener <rguenther@suse.de>
7190 PR tree-optimization/86363
7191 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7192 memset argument refers to a non-variable address.
7194 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7196 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7197 {PLUS,MINUS}_EXPR code to...
7198 (adjust_symbolic_bound): ...here,
7199 (combine_bound): ...here,
7200 (set_value_range_with_overflow): ...and here.
7202 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7204 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7206 (extract_range_from_abs_expr): ...here.
7208 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7210 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7211 -fno-omit-frame-pointer when not optimizing.
7213 2018-07-02 Martin Liska <mliska@suse.cz>
7216 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7217 (propagate_nothrow): Likewise.
7219 2018-07-02 Martin Liska <mliska@suse.cz>
7222 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7224 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7226 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7227 function in r262149, changing "loc" param from source_location to
7228 const dump_location_t &.
7229 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7230 declaration, as above.
7232 2018-07-01 Paul Koning <ni1d@arrl.net>
7234 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7235 -munit-asm, -mgnu-asm, -mdec-asm.
7236 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7237 (pdp11_output_labelref): New.
7238 (pdp11_output_def): New.
7239 (pdp11_output_addr_vec_elt): New.
7240 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7241 %# and %@ format codes.
7242 (pdp11_option_override): New.
7243 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7244 (pdp11_output_ident): New.
7245 (pdp11_asm_named_section): New.
7246 (pdp11_asm_init_sections): New.
7247 (pdp11_file_start): New.
7248 (pdp11_file_end): New.
7249 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7250 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7251 %o, like %c but octal.
7252 (pdp11_option_override): New.
7253 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7255 (DATA_SECTION_ASM_OP): Ditto.
7256 (READONLY_DATA_SECTION_ASM_OP): New.
7257 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7258 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7259 (ASM_OUTPUT_LABELREF): Ditto.
7260 (ASM_OUTPUT_DEF): Ditto.
7261 (ASM_OUTPUT_EXTERNAL): New.
7262 (ASM_OUTPUT_SOURCE_FILENAME): New.
7263 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7264 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7265 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7266 %# and %@ format codes.
7267 * config/pdp11/pdp11.opt (mgnu-asm): New.
7268 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7269 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7270 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7272 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7274 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7275 dereferencing path[] beyond its length.
7277 (debug_all_paths): New.
7278 (rewire_first_differing_edge): New.
7279 (adjust_paths_after_duplication): New.
7280 (duplicate_thread_path): Call adjust_paths_after_duplication.
7282 (thread_through_all_blocks): Add new argument to
7283 duplicate_thread_path.
7285 2018-06-30 Jim Wilson <jimw@sifive.com>
7287 * config/riscv/predicates.md (p2m1_shift_operand): New.
7288 (high_mask_shift_operand): New.
7289 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7290 pattern using p2m1_shift_operand.
7291 (lshsi3_zero_extend_3+2): New combiner pattern using
7292 high_mask_shift_operand.
7294 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7296 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7298 (vect_recog_rotate_pattern): ...here.
7299 (vect_convert_input): Try to insert casts of invariants in the
7301 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7302 preheader to be empty.
7304 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7306 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7307 vector type. If given, install it in the new statement's
7309 (vect_element_precision): New function.
7310 (vect_unpromoted_value): New struct.
7311 (vect_unpromoted_value::vect_unpromoted_value): New function.
7312 (vect_unpromoted_value::set_op): Likewise.
7313 (vect_look_through_possible_promotion): Likewise.
7314 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7315 (vect_widened_op_tree, vect_convert_input): Likewise.
7316 (vect_convert_inputs, vect_convert_output): Likewise.
7317 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7318 to handle the optional cast of the multiplication result and
7319 vect_widened_op_tree to detect the widened multiplication itself.
7320 Do not require the input and output of promotion casts to have
7321 the same sign, but base the signedness of the operation on the
7322 input rather than the result. If the pattern includes two
7323 promotions, check that those promotions have the same sign.
7324 Do not restrict the MULT_EXPR handling to a double-width result;
7325 handle quadruple-width results and wider. Use vect_convert_inputs
7326 to convert the inputs to the common type.
7327 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7328 to handle the optional cast of the ABS result. Also allow a sign
7329 change or a sign extension between the ABS and MINUS.
7330 Use vect_widened_op_tree to detect the widened subtraction and use
7331 vect_convert_inputs to convert the inputs to the common type.
7332 (vect_handle_widen_op_by_const): Delete.
7333 (vect_recog_widen_op_pattern): New function.
7334 (vect_recog_widen_mult_pattern): Use it.
7335 (vect_recog_widen_shift_pattern): Likewise.
7336 (vect_recog_widen_sum_pattern): Use
7337 vect_look_through_possible_promotion to handle the promoted
7340 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7342 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7343 the containing gimple_seq *.
7344 * gimple-iterator.h (gsi_for_stmt): Declare it.
7345 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7346 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7347 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7348 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7349 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7351 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7353 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7354 statement being replaced is part of an existing pattern
7355 definition sequence, inserting the new pattern statements before
7357 (vect_pattern_recog_1): Don't return a bool. If the statement
7358 is already part of a pattern, instead apply pattern matching
7359 to the pattern definition statements. Don't clear the
7360 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7361 (vect_pattern_recog): Don't break after the first match;
7362 continue processing the pattern definition statements instead.
7363 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7365 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7367 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7368 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7369 (vect_recog_widen_sum_pattern): Use it.
7371 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7373 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7374 phi is not a pattern statement and has not been replaced by
7375 a pattern statement.
7376 * tree-vect-patterns.c (type_conversion_p): Don't check
7377 STMT_VINFO_IN_PATTERN_P.
7378 (vect_recog_vector_vector_shift_pattern): Likewise.
7379 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7380 the pattern statement rather than the original statement; check
7381 directly for a WIDEN_MULT_EXPR here.
7382 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7383 vect_is_simple_use to return the pattern statement rather
7384 than the original statement; use is_pattern_stmt_p to check
7385 for such a pattern statement.
7386 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7387 to return the pattern statement rather than the original statement;
7388 don't do the same transformation here.
7389 (vect_is_simple_use): If the defining statement has been replaced
7390 by a pattern statement, return the pattern statement instead.
7391 Remove the corresponding (local) transformation from the vectype
7394 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7396 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7397 end and default to null.
7398 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7399 (vectorizable_reduction): Update calls accordingly, dropping the
7400 gimple ** argument if the passed-back statement isn't needed.
7401 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7402 (vect_recog_rotate_pattern): Likewise.
7403 (vect_recog_mask_conversion_pattern): Likewise.
7404 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7405 (vect_mask_constant_operand_p): Likewise.
7406 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7407 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7408 (get_group_load_store_type, get_load_store_type): Likewise.
7409 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7410 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7411 (vectorizable_conversion, vectorizable_assignment): Likewise.
7412 (vectorizable_shift, vectorizable_operation): Likewise.
7413 (vectorizable_store, vect_is_simple_cond): Likewise.
7414 (vectorizable_condition, vectorizable_comparison): Likewise.
7415 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7416 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7417 and move it to the end. Cope with null def_stmt_outs.
7419 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7421 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7423 2018-06-29 Jeff Law <law@redhat.com>
7425 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7426 displacements for TARGET_V850E2V3 and newer.
7427 (TARGET_LRA_P): Remove. Defaults to LRA now.
7428 * config/v850/v850.md (sign23byte_load): Remove.
7429 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7430 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7432 2018-06-29 Martin Liska <mliska@suse.cz>
7435 * coverage.c (coverage_init): Mangle full path name.
7436 * doc/invoke.texi: Document the change.
7437 * gcov-io.c (mangle_path): New.
7438 * gcov-io.h (mangle_path): Likewise.
7439 * gcov.c (mangle_name): Use mangle_path for path mangling.
7441 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7443 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7444 if starting source register is not even.
7446 2018-06-29 Martin Liska <mliska@suse.cz>
7448 PR tree-optimization/86263
7449 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7450 Make edge redirection.
7452 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7454 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7455 (dump_scope_depth): New variable.
7456 (get_dump_scope_depth): New function.
7457 (dump_begin_scope): New function.
7458 (dump_end_scope): New function.
7459 * dumpfile.h (get_dump_scope_depth): New declaration.
7460 (dump_begin_scope): New declaration.
7461 (dump_end_scope): New declaration.
7462 (class auto_dump_scope): New class.
7463 (AUTO_DUMP_SCOPE): New macro.
7464 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7467 2018-06-29 Richard Biener <rguenther@suse.de>
7469 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7470 compute_all_dependences succeeds.
7471 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7472 exceed --param loop-max-datarefs-for-datadeps.
7474 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7476 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7478 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7481 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7482 alternative 0 in preferred_for_speed attribute.
7484 2018-06-28 Paul Koning <ni1d@arrl.net>
7486 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7487 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7488 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7489 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7490 for shift insn patterns.
7492 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7494 * cgraph.c (cgraph_node::get_body): Replace assignments to
7495 "dump_file" with calls to set_dump_file.
7496 * dumpfile.c (alt_dump_file): Make static, and group with...
7497 (alt_flags): ...this definition.
7498 (dumps_are_enabled): New variable.
7499 (refresh_dumps_are_enabled): New function.
7500 (set_dump_file): New function.
7501 (set_alt_dump_file): New function.
7502 (gcc::dump_manager::dump_start): Replace assignments to
7503 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7505 (gcc::dump_manager::dump_finish): Likewise.
7506 * dumpfile.h (alt_dump_file): Delete decl.
7507 (dumps_are_enabled): New variable decl.
7508 (set_dump_file): New function decl.
7509 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7511 * tree-nested.c (lower_nested_functions): Replace assignments to
7512 "dump_file" with calls to set_dump_file.
7514 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7516 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7517 goto_locus of each outgoing edge of each basic block.
7519 2018-06-28 Richard Biener <rguenther@suse.de>
7521 * dwarf2out.c (decl_scope_table): Remove.
7522 (push_decl_scope): Likewise.
7523 (pop_decl_scope): Likewise.
7524 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7525 (gen_struct_or_union_type_die): Likewise.
7526 (gen_tagged_type_die): Likewise.
7527 (dwarf2out_init): Do not initialize decl_scope_table.
7528 (dwarf2out_c_finalize): Do not free it.
7530 2018-06-28 Richard Biener <rguenther@suse.de>
7532 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7533 deciding whether to not re-use a DIE.
7535 2018-06-28 Richard Biener <rguenther@suse.de>
7537 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7538 DW_AT_abstract_origin attribute.
7540 2018-06-28 Martin Liska <mliska@suse.cz>
7542 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7543 Use newly introduced constants.
7544 * tree-switch-conversion.h (struct jump_table_cluster):
7545 Define max_ratio_for_size and max_ratio_for_speed.
7547 2018-06-28 Martin Liska <mliska@suse.cz>
7549 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7550 Add new checking assert to catch invalid state.
7551 (jump_table_cluster::can_be_handled): Handle single case
7553 (jump_table_cluster::is_beneficial): Bail out for such case.
7554 (bit_test_cluster::find_bit_tests):
7555 Add new checking assert to catch invalid state.
7556 (bit_test_cluster::can_be_handled): Handle single case
7558 (bit_test_cluster::is_beneficial): Bail out for such case.
7559 (switch_decision_tree::analyze_switch_statement):
7562 2018-06-28 Martin Liska <mliska@suse.cz>
7564 * common.opt: Introduce -completion option.
7565 * gcc.c (driver_handle_option): Handle it.
7566 (driver::main): Print completions if completion
7568 * opt-suggestions.c (option_proposer::get_completions):
7570 (option_proposer::suggest_completion): Likewise.
7571 (option_proposer::find_param_completions): Likewise.
7572 (verify_autocompletions): Likewise.
7573 (test_completion_valid_options): Likewise.
7574 (test_completion_valid_params): Likewise.
7575 (in_completion_p): Likewise.
7576 (empty_completion_p): Likewise.
7577 (test_completion_partial_match): Likewise.
7578 (test_completion_garbage): Likewise.
7579 (opt_proposer_c_tests): Likewise.
7580 * opt-suggestions.h: Declare new functions.
7581 * opts.c (common_handle_option): Handle OPT__completion_.
7582 * selftest-run-tests.c (selftest::run_tests): Add
7583 opt_proposer_c_tests.
7584 * selftest.c (assert_str_startswith): New.
7585 * selftest.h (assert_str_startswith): Likewise.
7586 (opt_proposer_c_tests): New.
7587 (ASSERT_STR_STARTSWITH): Likewise.
7589 2018-06-28 Martin Liska <mliska@suse.cz>
7591 * Makefile.in: Add opt-suggestions.o.
7592 * gcc-main.c: Include opt-suggestions.h.
7593 * gcc.c (driver::driver): Likewise.
7594 (driver::~driver): Remove m_option_suggestions.
7595 (driver::build_option_suggestions): Moved to option_proposer.
7596 (driver::suggest_option): Likewise.
7597 (driver::handle_unrecognized_options): Use option_proposer.
7598 * gcc.h (class driver): Add new memver m_option_proposer.
7599 * opt-suggestions.c: New file.
7600 * opt-suggestions.h: New file.
7602 2018-06-28 Martin Liska <mliska@suse.cz>
7604 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7605 (auto_string_vec::~auto_string_vec): Likewise.
7607 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7609 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7610 prevent_decl_creation_for_types fields up and add reset_location field.
7611 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7612 statement if id->reset_location is true.
7613 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7614 id->reset_location is true.
7615 (copy_phis_for_bb): Force input_location on the arguments if
7616 id->reset_location is true.
7617 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7618 is set on the function to be inlined.
7620 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7622 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7624 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7626 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7627 registers for Pmode.
7628 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7629 hard registers for the clobbered pseudo.
7631 2018-06-27 Paul Koning <ni1d@arrl.net>
7633 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7634 mutually exclusive options.
7635 * config/pdp11/constraints.md (h): New constraint.
7636 (O): Update definition to match shift code generation.
7637 (D): New constraint.
7638 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7640 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7642 (output_jump): Change arguments.
7643 (pdp11_fixed_cc_regs): New function.
7644 (pdp11_cc_mode): Ditto.
7645 (pdp11_expand_shift): Ditto.
7646 (pdp11_assemble_shift): Ditto.
7647 (pdp11_small_shift): Ditto.
7648 (pdp11_branch_cost): Remove.
7649 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7651 (pdp11_register_move_cost): Update for CC registers.
7652 (pdp11_rtx_costs): Add case for LSHIFTRT.
7653 (pdp11_output_jump): Add CCNZ mode conditional branches.
7654 (notice_update_cc_on_set): Remove.
7655 (pdp11_cc_mode): New function.
7656 (simple_memory_operand): Correct pre/post decrement case.
7657 (no_side_effect_operand): New function.
7658 (pdp11_regno_reg_class): Add CC_REGS class.
7659 (pdp11_fixed_cc_regs): New function.
7660 (pdp11_small_shift): New function.
7661 (pdp11_expand_shift): New function to expand shift insns.
7662 (pdp11_assemble_shift): New function to output shifts.
7663 (pdp11_branch_cost): Remove.
7664 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7665 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7666 (WCHAR_TYPE): Ditto.
7667 (PTRDIFF_TYPE): Ditto.
7668 (ADJUST_INSN_LENGTH): New macro.
7669 (FIXED_REGISTERS): Add CC registers.
7670 (CALL_USED_REGISTERS): Ditto.
7672 (REG_CLASS_NAMES): Ditto.
7673 (REG_CLASS_CONTENTS): Ditto.
7674 (SELECT_CC_MODE): Use new function.
7675 (TARGET_FLAGS_REGNUM): New macro.
7676 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
7677 (cc0_reg_rtx): Remove.
7678 (CC_STATUS_MDEP): Remove.
7679 (CC_STATUS_MDEFP_INIT): Remove.
7680 (CC_IN_FPU): Remove.
7681 (NOTICE_UPDATE_CC): Remove.
7682 (REGISTER_NAMES): Add CC registers.
7683 (BRANCH_COST): Change to constant 1.
7684 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
7686 * config/pdp11/pdp11.opt (mbcopy): Remove.
7687 (mbcopy-builtin): Remove.
7688 (mbranch-cheap): Remove.
7689 (mbranch-expensive): Remove.
7690 * config/pdp11/predicates.md (expand_shift_operand): Update to
7691 match shift code generation.
7692 (ccnz_operator): New predicate.
7693 * doc/invoke.texi (PDP-11 Options): Remove deleted options
7694 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
7695 Remove non-existent option -mabshi, -mno-abshi. Document mutually
7697 * doc/md.texi (PDP-11): Document new D and h constraints. Update
7698 description of O constraint.
7700 2018-06-27 Jeff Law <law@redhat.com>
7701 Austin Law <austinklaw@gmail.com>
7703 * config/v850/v850.md (addsi3_set_flags): New pattern.
7704 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
7705 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
7706 (zero_extendhisi2_v850_set_flags): Likewise.
7707 (zero_extendqisi2_v850_set_flags): Likewise.
7708 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
7709 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
7710 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
7712 * config/v850/v850-protos.h (notice_update_cc): Remove.
7713 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
7714 (v850_print_operand): Handle 'D' and "d".
7715 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
7716 Add handling of arithmetic/logical operations compared against zero.
7717 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
7718 Do not look at v850_compare_op, instead get mode from last argument.
7719 (v850_gen_compare): Remove
7720 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
7721 after reload for prologue insns.
7722 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
7724 (construct_save_jarl): Likewise.
7725 (TARGET_FLAGS_REGNUM): Define.
7726 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
7727 (NOTICE_UPDATE_CC): Remove.
7728 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
7729 than cc0. Conditionalize on reload_completed.
7730 (cmpsi_insn, setfcc_insn): Likewise.
7731 (tst1 splitter): Turn into define_and_split which sets the flags
7733 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
7734 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
7735 (cstoresf4, cstoredf4): Clobber the flags.
7736 (cmpsi, cmpsf, cmpdf): Remove expanders.
7737 (setf_insn): Remove pattern.
7738 (addsi3): Turn into define_and_split which clobbers the flags after
7739 reload and a suitable pattern (addsi3_clobber_flags) for use after
7741 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
7742 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
7743 (ashrsi3, ashrsi3_v850e2): Likewise.
7744 (bins): Clobber the flags.
7745 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
7746 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
7747 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
7748 (call_value_internal_short, call_value_internal_long): Likewise.
7749 (callt_save_interrupt, callt_return_interrupt): Likewise.
7750 (save_interrupt, return_interrupt): Likewise.
7751 (callt_save_all_interrupt, save_all_interrupt): Likewise.
7752 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
7753 (restore_all_interrupt, _restore_all_interrupt): Likewise.
7754 (All FP comparisons): Only allow after reload has completed.
7756 (divh, divhu): Tweak output template.
7757 (branch_z_normal, branch_z_invert): Remove
7758 (branch_nz_normal, branch_nz_invert): Likewise.
7759 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
7761 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
7762 * config/v850/v850.c (notice_update_cc): Remove.
7763 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
7764 (CC_NO_CARRY): Likewise.
7765 (NOTICE_UPDATE_CC): Define to nothing.
7766 * config/v850/v850.md: Remove block comment on cc0 handling
7767 Remove "cc" attribute from all patterns. Remove cc_status handling
7768 from all patterns. Minor formatting fixes.
7770 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7772 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
7773 (cortex-a76.cortex-a55): Likewise.
7774 * config/aarch64/aarch64-tune.md: Regenerate.
7775 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
7776 cortex-a76.cortex-a55.
7778 2018-06-27 Jeff Law <law@redhat.com>
7780 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7781 (MULTILIB_DIRNAMES): Similarly.
7783 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
7785 * gimple.h (gimple_return_retbnd): Delete.
7786 (gimple_return_set_retbnd): Likewise.
7787 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7788 gimple_return_set_retbnd.
7789 * gimple-pretty-print.c (dump_gimple_return): Remove call to
7790 gimple_return_retbnd and adjust.
7791 * tree-inline.h (struct copy_body_data): Remove retbnd field.
7792 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7793 Explicitly return NULL in a couple more cases. Move assertion
7794 on debug statements and remove unreachable code.
7795 (reset_debug_binding): Do not test id->retbnd.
7796 (expand_call_inline): Do not set it.
7798 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7800 * configure.ac: Add --disable-gcov option.
7801 * configure: Regenerate.
7802 * Makefile.in: Honour @enable_gcov@.
7803 * doc/install.texi: Document --disable-gcov.
7805 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7807 * config/arm/arm-cpus.in (cortex-a76): New entry.
7808 (cortex-a76.cortex-a55): Likewise.
7809 * config/arm/arm-tables.opt: Regenerate.
7810 * config/arm/arm-tune.md: Likewise.
7811 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7812 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7813 cortex-a76.cortex-a55.
7815 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7818 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7820 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7822 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7824 (EPILOGUE_USES): Likewise.
7826 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7828 * tree-inline.c (remap_location): New function extracted from...
7829 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7830 (copy_phis_for_bb): ...here. Call remap_location.
7831 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7833 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7835 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7836 unaligned vsx for 16B memset.
7838 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7841 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7842 ieee128_float_type_node to long_double_type_node unless
7843 TARGET_LONG_DOUBLE_128 is set.
7845 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7847 * cfgloop.c (get_loop_location): Convert return type from
7848 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7849 by implicit construction from rtx_insn *, and using
7850 dump_user_location_t::from_function_decl for the fallback case.
7851 * cfgloop.h (get_loop_location): Convert return type from
7852 location_t to dump_user_location_t.
7853 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7854 dump_printf_loc to pass in a dump_location_t rather than a
7855 location_t, via the gimple stmt.
7856 * coverage.c (get_coverage_counts): Update calls to
7857 dump_printf_loc to pass in dump_location_t rather than a
7859 * doc/optinfo.texi (Dump types): Convert example of
7860 dump_printf_loc from taking "locus" to taking "insn". Update
7861 description of the "_loc" calls to cover dump_location_t.
7862 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7864 (dump_user_location_t::dump_user_location_t): New constructors,
7865 from gimple *stmt and rtx_insn *.
7866 (dump_user_location_t::from_function_decl): New function.
7867 (dump_loc): Make static.
7868 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7869 const dump_location_t &.
7870 (dump_generic_expr_loc): Delete.
7871 (dump_printf_loc): Convert param "loc" from location_t to
7872 const dump_location_t &.
7873 (selftest::test_impl_location): New function.
7874 (selftest::dumpfile_c_tests): New function.
7875 * dumpfile.h: Include "profile-count.h".
7876 (class dump_user_location_t): New class.
7877 (struct dump_impl_location_t): New struct.
7878 (class dump_location_t): New class.
7879 (dump_printf_loc): Convert 2nd param from source_location to
7880 const dump_location_t &.
7881 (dump_generic_expr_loc): Delete.
7882 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7883 const dump_location_t &.
7884 * gimple-fold.c (fold_gimple_assign): Update call to
7885 dump_printf_loc to pass in a dump_location_t rather than a
7886 location_t, via the gimple stmt.
7887 (gimple_fold_call): Likewise.
7888 * gimple-loop-interchange.cc
7889 (loop_cand::analyze_iloop_reduction_var): Update for change to
7890 check_reduction_path.
7891 (tree_loop_interchange::interchange): Update for change to
7893 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7894 change in return-type of find_loop_location.
7895 (graphite_regenerate_ast_isl): Likewise.
7896 * graphite-optimize-isl.c (optimize_isl): Likewise.
7897 * graphite.c (graphite_transform_loops): Likewise.
7898 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7899 pass in a dump_location_t rather than a location_t, via the
7901 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7902 * ipa.c (walk_polymorphic_call_targets): Likewise.
7903 * loop-unroll.c (report_unroll): Convert "locus" param from
7904 location_t to dump_location_t.
7905 (decide_unrolling): Update for change to get_loop_location's
7907 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7908 location_t to dump_user_location_t.
7909 (grid_find_single_omp_among_assignments_1): Updates calls to
7910 dump_printf_loc to pass in a dump_location_t rather than a
7911 location_t, via the gimple stmt.
7912 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7913 location_t to dump_location_t. Updates calls to dump_printf_loc
7914 to pass in a dump_location_t rather than a location_t, via the
7916 (grid_inner_loop_gridifiable_p): Likewise.
7917 (grid_dist_follows_simple_pattern): Likewise.
7918 (grid_gfor_follows_tiling_pattern): Likewise.
7919 (grid_target_follows_gridifiable_pattern): Likewise.
7920 (grid_attempt_target_gridification): Convert initialization
7921 of local "grid" from memset to zero-initialization; FIXME: does
7922 this require C++11? Update call to dump_printf_loc to pass in a
7923 optinfo_location rather than a location_t, via the gimple stmt.
7924 * profile.c (read_profile_edge_counts): Updates call to
7925 dump_printf_loc to pass in a dump_location_t rather than a
7927 (compute_branch_probabilities): Likewise.
7928 * selftest-run-tests.c (selftest::run_tests): Call
7930 * selftest.h (dumpfile_c_tests): New decl.
7931 * tree-loop-distribution.c (pass_loop_distribution::execute):
7932 Update for change in return type of find_loop_location.
7933 * tree-parloops.c (parallelize_loops): Likewise.
7934 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7935 "locus" from location_t to dump_user_location_t.
7936 (canonicalize_loop_induction_variables): Likewise.
7937 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7938 for change in return type of find_loop_location.
7939 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7940 to dump_printf_loc to pass in a dump_location_t rather than a
7941 location_t, via the stmt.
7942 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7944 * tree-vect-loop-manip.c (find_loop_location): Convert return
7945 type from source_location to dump_user_location_t.
7946 (vect_do_peeling): Update for above change.
7947 (vect_loop_versioning): Update for change in type of
7949 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7950 from location_t to dump_user_location_t.
7951 (vect_estimate_min_profitable_iters): Update for change in type
7953 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7954 location_t to dump_location_t.
7955 (vect_slp_bb): Update for change in type of vect_location.
7956 * tree-vectorizer.c (vect_location): Convert from source_location
7957 to dump_user_location_t.
7958 (try_vectorize_loop_1): Update for change in vect_location's type.
7959 (vectorize_loops): Likewise.
7960 (increase_alignment): Likewise.
7961 * tree-vectorizer.h (vect_location): Convert from source_location
7962 to dump_user_location_t.
7963 (find_loop_location): Convert return type from source_location to
7964 dump_user_location_t.
7965 (check_reduction_path): Convert 1st param from location_t to
7966 dump_user_location_t.
7967 * value-prof.c (check_counter): Update call to dump_printf_loc to
7968 pass in a dump_user_location_t rather than a location_t; update
7969 call to error_at for change in type of "locus".
7970 (check_ic_target): Update call to dump_printf_loc to
7971 pass in a dump_user_location_t rather than a location_t, via the
7974 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7976 * config/s390/s390.h (enum processor_flags): Do not use
7977 default tune parameter when -march was specified.
7979 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7982 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7983 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7986 2018-06-26 Richard Biener <rguenther@suse.de>
7988 PR tree-optimization/86287
7990 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7991 (vect_analyze_loop): Initialize n_stmts.
7993 2018-06-26 Richard Biener <rguenther@suse.de>
7996 * fold-const.c (fold_convertible_p): Pointer extension
7999 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8002 * dwarf2out.c (loc_list_has_views): Adjust comments.
8003 (dw_loc_list): Split single cross-partition range with
8006 2018-06-25 Jeff Law <law@redhat.com>
8008 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8009 on -mbig-switch by default.
8011 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8013 (const_float_0_operand): Remove unused predicate.
8014 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8015 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8016 (recipsf2): New expander. Original pattern now called
8018 (recipdf2, recipdf2_insn): Similarly.
8019 (rsqrtsf2, rsqrtsf2_insn): Similarly
8020 (rsqrtdf2, rsqrtdf2_insn): Similarly
8022 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8024 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8025 Simplify logic for FreeBSD (twice).
8027 2018-06-25 Martin Sebor <msebor@redhat.com>
8029 PR tree-optimization/86204
8030 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8031 a strnlen result if it's less than the length of the string.
8033 2018-06-25 Martin Sebor <msebor@redhat.com>
8035 PR tree-optimization/85700
8036 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8037 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8038 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8040 2018-06-25 Martin Sebor <msebor@redhat.com>
8042 * doc/extend.texi (Zero-length arrays): Update and clarify.
8044 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8046 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8047 added IEEE/IBM long double multilib support on PowerPC little
8048 endian Linux systems.
8049 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8050 (MULTILIB_DEFAULTS): Likewise.
8051 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8053 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8054 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8055 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8057 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8060 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8061 (REORDER_45): Likewise.
8063 2018-06-25 Jeff Law <law@redhat.com>
8065 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8066 dividend to 32 bits. Adjust length.
8067 (udivmodhi4): Cleanup output template. Fix length.
8069 2018-06-25 Carl Love <cel@us.ibm.com>
8071 * config/rs6000/vsx.md: Change word selector to prefered location.
8073 2018-06-25 Richard Biener <rguenther@suse.de>
8075 PR tree-optimization/86304
8076 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8077 epilogue-if-converted loops as well.
8079 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8081 * lto-section-out.c (lto_begin_section): Do not print section
8082 name for noaddr and unnumbered dumps.
8084 2018-06-25 Richard Biener <rguenther@suse.de>
8086 * tree-vectorizer.h (struct vec_info_shared): New structure
8087 with parts split out from struct vec_info and loop_nest from
8088 struct _loop_vec_info.
8089 (struct vec_info): Adjust accordingly.
8090 (struct _loop_vec_info): Likewise.
8091 (LOOP_VINFO_LOOP_NEST): Adjust.
8092 (LOOP_VINFO_DATAREFS): Likewise.
8093 (LOOP_VINFO_DDRS): Likewise.
8094 (struct _bb_vec_info): Likewise.
8095 (BB_VINFO_DATAREFS): Likewise.
8096 (BB_VINFO_DDRS): Likewise.
8097 (struct _stmt_vec_info): Add dr_aux member.
8098 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8099 (DR_MISALIGNMENT_UNINITIALIZED): New.
8100 (set_dr_misalignment): Adjust.
8101 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8102 (vect_analyze_loop): Adjust prototype.
8103 (vect_analyze_loop_form): Likewise.
8104 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8105 Compute dependences lazily.
8106 (vect_record_base_alignments): Use shared datarefs/ddrs.
8107 (vect_verify_datarefs_alignment): Likewise.
8108 (vect_analyze_data_refs_alignment): Likewise.
8109 (vect_analyze_data_ref_accesses): Likewise.
8110 (vect_analyze_data_refs): Likewise.
8111 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8112 constructor parameter for shared part.
8113 (vect_analyze_loop_form): Pass in shared part and adjust.
8114 (vect_analyze_loop_2): Pass in storage for the number of
8115 stmts. Move loop nest finding to the caller. Compute
8117 (vect_analyze_loop): Pass in shared part.
8118 (vect_transform_loop): Verify shared datarefs are unchanged.
8119 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8120 constructor parameter for shared part.
8121 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8122 (vect_slp_bb): Verify shared datarefs are unchanged before
8124 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8126 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8127 DR_MISALIGNMENT_UNINITIALIZED.
8128 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8129 parameter for shared part.
8130 (vec_info::~vec_info): Adjust.
8131 (vec_info_shared::vec_info_shared): New.
8132 (vec_info_shared::~vec_info_shared): Likewise.
8133 (vec_info_shared::save_datarefs): Likewise.
8134 (vec_info_shared::check_datarefs): Likewise.
8135 (try_vectorize_loop_1): Construct shared part live for analyses
8136 of a single loop for multiple vector sizes.
8137 * tree-parloops.c (gather_scalar_reductions): Adjust.
8139 2018-06-25 Richard Biener <rguenther@suse.de>
8141 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8142 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8143 (vect_analyze_data_refs): Remove similar code from here and
8144 simplify accordingly.
8146 2018-06-25 Richard Biener <rguenther@suse.de>
8148 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8149 for reverse storage order accesses rather than asserting
8150 they cannot happen here.
8152 2018-06-25 Tom de Vries <tdevries@suse.de>
8155 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8156 Use data16 instead of .byte for insn prefix.
8158 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8161 * parser.c (make_char_string_pack): Pass this literal chars
8162 through cpp_interpret_string.
8163 (cp_parser_userdef_numeric_literal): Check the result of
8164 make_char_string_pack.
8166 2018-06-24 Maya Rashish <coypu@sdf.org>
8168 * ginclude/stddef.h: Simplify conditions around avoiding
8169 re-definition of __size_t.
8171 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8173 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8174 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8176 2018-06-22 Maya Rashish <coypu@sdf.org>
8178 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8179 made to OpenBSD and NetBSD through the years.
8181 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8183 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8184 behavior of vec_pack (vector double, vector double) to match
8185 behavior of vec_float2 (vector double, vector double).
8187 2018-06-22 Olivier Hainque <hainque@adacore.com>
8189 * gimplify.c (gimplify_function_tree): Prevent creation
8190 of a trampoline for the address of the current function
8191 passed to entry/exit instrumentation hooks.
8193 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8196 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8199 2018-06-22 Martin Liska <mliska@suse.cz>
8201 PR tree-optimization/86263
8202 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8203 Bail out if is_enabled is false.
8204 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8206 (jump_table_cluster::is_enabled): New function.
8208 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8210 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8211 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8212 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8213 (lto_input_ts_binfo_tree_pointers): Likewise.
8214 * tree-streamer-out.c (streamer_write_tree_bitfields,
8215 write_ts_binfo_tree_pointers): Likewise.
8216 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8218 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8220 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8222 2018-06-22 Martin Liska <mliska@suse.cz>
8224 * symbol-summary.h (get): Make it pure and inline move
8225 functionality from ::get function.
8226 (get): Remove and inline into ::get and ::get_create.
8227 (get_create): Move code from ::get function.
8229 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8232 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8233 -x assembler-with-cpp.
8235 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8237 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8238 _FILE_OFFSET_BITS=64 for C++.
8240 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8242 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8243 conversion insn that shows up when pr85657-3.c is compiled using
8244 IEEE 128-bit long double.
8245 (neg<mode>2_internal): Use the correct mode to check whether the
8246 mode is IBM extended.
8247 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8248 multiply and divide external functions from being created more
8251 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8253 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8255 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8256 the edge can be forwarded.
8257 (cfg_layout_merge_blocks): Likewise.
8259 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8261 * except.c (finish_eh_generation): Commit edge insertions only after
8262 the EH edges have been redirected from post-landing to landing pads.
8264 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8266 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8267 create_tmp_var_for to create the FRAME decl.
8268 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8270 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8272 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8273 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8274 debug statement when resetting its value.
8275 (expand_call_inline): Copy the locus of the call onto the assignment
8276 of the return value, if any. Use local variable in more cases.
8278 2018-06-21 Martin Liska <mliska@suse.cz>
8280 * ipa-pure-const.c (propagate_nothrow): Use
8281 funct_state_summaries->get.
8282 (dump_malloc_lattice): Likewise.
8283 (propagate_malloc): Likewise.
8285 2018-06-21 Richard Biener <rguenther@suse.de>
8287 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8288 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8289 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8291 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8292 BLOCK_ABSTRACT_ORIGIN unconditionally.
8294 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8296 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8298 * ipa-reference.c (ipa_reference_c_finalize): Delete
8299 ipa_ref_opt_sum_summaries and set it to NULL.
8301 2018-06-21 Tom de Vries <tdevries@suse.de>
8303 PR tree-optimization/85859
8304 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8305 test with comment from bb_no_side_effects_p.
8307 2018-06-21 Richard Biener <rguenther@suse.de>
8309 PR tree-optimization/86232
8310 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8311 max for constant niter.
8313 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8315 * config/aarch64/aarch64-simd.md
8316 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8318 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8320 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8321 Make opernads of the unspec commutative.
8323 2018-06-21 Richard Biener <rguenther@suse.de>
8325 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8326 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8327 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8328 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8330 (vect_record_base_alignments): Do not record base alignment
8331 for gathers or scatters.
8332 (vect_compute_data_ref_alignment): Drop return value that is always
8333 true. Bail out early for gathers or scatters.
8334 (vect_enhance_data_refs_alignment): Bail out early for gathers
8336 (vect_find_same_alignment_drs): Likewise.
8337 (vect_analyze_data_refs_alignment): Remove dead code.
8338 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8339 (vect_analyze_data_refs): For possible gathers or scatters do
8340 not create an alternate DR, just check their possible validity
8341 and mark them. Adjust DECL_NONALIASED handling to not rely
8343 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8344 update inits of gathers or scatters.
8345 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8346 Also copy gather/scatter flag to pattern vinfo.
8348 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8350 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8351 behavior of vec_packsu (vector unsigned long long, vector unsigned
8352 long long) to match behavior of vec_packs with same signature.
8354 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8355 Thomas Schwinge <thomas@codesourcery.com>
8356 Cesar Philippidis <cesar@codesourcery.com>
8358 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8359 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8360 (gimplify_adjust_omp_clauses): Likewise.
8361 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8362 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8363 (gimplify_omp_target_update): Update handling of acc update and
8365 * omp-low.c (install_var_field): Remove unused parameter
8366 base_pointers_restrict.
8367 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8368 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8370 (omp_target_base_pointers_restrict_p): Delete.
8371 (scan_omp_target): Update call to scan_sharing_clauses.
8372 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8374 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8375 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8376 (convert_local_omp_clauses): Likewise.
8377 * tree-pretty-print.c (dump_omp_clause): Likewise.
8378 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8380 (omp_clause_code_name): Likewise.
8382 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8385 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8388 PR tree-optimization/86231
8389 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8390 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8392 2018-06-20 Tom de Vries <tdevries@suse.de>
8394 PR tree-optimization/86097
8395 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8396 iv type if signedness of iv type is not the same as that of *nit.
8398 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8400 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8401 EDGE_EH edges, verify they are all EDGE_EH.
8403 2018-06-20 Maya Rashish <coypu@sdf.org>
8405 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8407 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8409 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8410 * config/aarch64/aarch64.c (xgene1_tunings): Add
8411 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8412 (aarch64_mode_valid_for_sched_fusion_p):
8413 Allow 16-byte modes.
8414 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8415 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8417 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8419 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8420 * config/aarch64/iterators.md (VQ2): New mode iterator.
8422 2018-06-20 Martin Liska <mliska@suse.cz>
8424 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8425 Change default ratio from 10 to 8.
8427 2018-06-20 Martin Liska <mliska@suse.cz>
8429 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8431 (bit_test_cluster::find_bit_tests): Likewise.
8432 (switch_decision_tree::analyze_switch_statement): Find clusters.
8433 * tree-switch-conversion.h (struct jump_table_cluster): Document
8436 2018-06-20 Martin Liska <mliska@suse.cz>
8438 * tree-switch-conversion.c (switch_conversion::collect):
8439 Record m_uniq property.
8440 (switch_conversion::expand): Bail out for special conditions.
8441 (group_cluster::~group_cluster): New.
8442 (group_cluster::group_cluster): Likewise.
8443 (group_cluster::dump): Likewise.
8444 (jump_table_cluster::emit): New.
8445 (switch_decision_tree::fix_phi_operands_for_edges): New.
8446 (struct case_node): Remove struct.
8447 (jump_table_cluster::can_be_handled): New.
8448 (case_values_threshold): Moved to header.
8449 (reset_out_edges_aux): Likewise.
8450 (jump_table_cluster::is_beneficial): New.
8451 (bit_test_cluster::can_be_handled): Likewise.
8452 (add_case_node): Remove.
8453 (bit_test_cluster::is_beneficial): New.
8454 (case_bit_test::cmp): New.
8455 (bit_test_cluster::emit): New.
8456 (expand_switch_as_decision_tree_p): Remove.
8457 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8458 (fix_phi_operands_for_edge): Likewise.
8459 (switch_decision_tree::analyze_switch_statement): New.
8460 (compute_cases_per_edge): Move ...
8461 (switch_decision_tree::compute_cases_per_edge): ... here.
8462 (try_switch_expansion): Likewise.
8463 (switch_decision_tree::try_switch_expansion): Likewise.
8464 (record_phi_operand_mapping): Likewise.
8465 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8466 (emit_case_decision_tree): Likewise.
8467 (switch_decision_tree::emit): Likewise.
8468 (balance_case_nodes): Likewise.
8469 (switch_decision_tree::balance_case_nodes): Likewise.
8470 (dump_case_nodes): Likewise.
8471 (switch_decision_tree::dump_case_nodes): Likewise.
8472 (emit_jump): Likewise.
8473 (switch_decision_tree::emit_jump): Likewise.
8474 (emit_cmp_and_jump_insns): Likewise.
8475 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8476 (emit_case_nodes): Likewise.
8477 (switch_decision_tree::emit_case_nodes): Likewise.
8478 (conditional_probability): Remove.
8479 * tree-switch-conversion.h (enum cluster_type): New.
8481 (struct cluster): Likewise.
8482 (cluster::cluster): Likewise.
8483 (struct simple_cluster): Likewise.
8484 (simple_cluster::simple_cluster): Likewise.
8485 (struct group_cluster): Likewise.
8486 (struct jump_table_cluster): Likewise.
8487 (struct bit_test_cluster): Likewise.
8488 (struct min_cluster_item): Likewise.
8489 (struct case_tree_node): Likewise.
8490 (case_tree_node::case_tree_node): Likewise.
8491 (jump_table_cluster::case_values_threshold): Likewise.
8492 (struct case_bit_test): Likewise.
8493 (struct switch_decision_tree): Likewise.
8494 (struct switch_conversion): Likewise.
8495 (switch_decision_tree::reset_out_edges_aux): Likewise.
8497 2018-06-20 Martin Liska <mliska@suse.cz>
8499 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8500 (hoist_edge_and_branch_if_true): Likewise.
8501 (expand_switch_using_bit_tests_p): Likewise.
8502 (struct case_bit_test): Likewise.
8503 (case_bit_test_cmp): Likewise.
8504 (emit_case_bit_tests): Likewise.
8505 (switch_conversion::switch_conversion): New class.
8506 (struct switch_conv_info): Remove old struct.
8507 (collect_switch_conv_info): More to ...
8508 (switch_conversion::collect): ... this.
8509 (check_range): Likewise.
8510 (switch_conversion::check_range): Likewise.
8511 (check_all_empty_except_final): Likewise.
8512 (switch_conversion::check_all_empty_except_final): Likewise.
8513 (check_final_bb): Likewise.
8514 (switch_conversion::check_final_bb): Likewise.
8515 (create_temp_arrays): Likewise.
8516 (switch_conversion::create_temp_arrays): Likewise.
8517 (free_temp_arrays): Likewise.
8518 (gather_default_values): Likewise.
8519 (switch_conversion::gather_default_values): Likewise.
8520 (build_constructors): Likewise.
8521 (switch_conversion::build_constructors): Likewise.
8522 (constructor_contains_same_values_p): Likewise.
8523 (switch_conversion::contains_same_values_p): Likewise.
8524 (array_value_type): Likewise.
8525 (switch_conversion::array_value_type): Likewise.
8526 (build_one_array): Likewise.
8527 (switch_conversion::build_one_array): Likewise.
8528 (build_arrays): Likewise.
8529 (switch_conversion::build_arrays): Likewise.
8530 (gen_def_assigns): Likewise.
8531 (switch_conversion::gen_def_assigns): Likewise.
8532 (prune_bbs): Likewise.
8533 (switch_conversion::prune_bbs): Likewise.
8534 (fix_phi_nodes): Likewise.
8535 (switch_conversion::fix_phi_nodes): Likewise.
8536 (gen_inbound_check): Likewise.
8537 (switch_conversion::gen_inbound_check): Likewise.
8538 (process_switch): Use the newly created class.
8539 (switch_conversion::expand): New.
8540 (switch_conversion::~switch_conversion): New.
8541 * tree-switch-conversion.h: New file.
8543 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8545 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8546 tree-vect-patterns.c.
8547 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8548 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8549 (vect_recog_sad_pattern): Likewise.
8550 (vect_recog_widen_sum_pattern): Likewise.
8551 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8552 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8553 (vect_recog_rotate_pattern): Likewise.
8554 (vect_recog_mult_pattern): Likewise.
8555 (vect_recog_vector_vector_shift_pattern): Likewise.
8556 (vect_recog_divmod_pattern): Likewise.
8557 (vect_recog_mixed_size_cond_pattern): Likewise.
8558 (vect_recog_bool_pattern): Likewise.
8559 (vect_recog_mask_conversion_pattern): Likewise.
8560 (vect_try_gather_scatter_pattern): Likewise.
8561 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8562 (vect_recog_over_widening_pattern): Likewise.
8563 (vect_recog_gather_scatter_pattern): Likewise.
8564 (vect_recog_func_ptr): Move from tree-vectorizer.h
8565 (vect_vect_recog_func_ptrs): Move further down the file.
8566 (vect_recog_func): Likewise. Remove the third argument.
8567 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8568 (vect_pattern_recog_1): Expect the pattern function to do any
8569 necessary target tests. Also expect it to provide a vector type.
8570 Remove the type_in handling.
8572 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8574 * tree-vect-patterns.c (vect_pattern_detected): New function.
8575 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8576 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8577 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8578 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8579 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8580 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8581 (vect_recog_mask_conversion_pattern)
8582 (vect_try_gather_scatter_pattern): Likewise.
8584 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8586 * tree-vect-patterns.c (vect_get_internal_def): New function.
8587 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8588 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8589 (search_type_for_mask_1): Use it.
8591 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8593 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8594 redundant WIDEN_SUM_EXPR handling.
8595 (vect_recog_sad_pattern): Likewise.
8597 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8599 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8600 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8601 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8602 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8603 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8605 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8607 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8608 use the stmt_vec_info of the original bb statement for the
8609 new zero assignment, even if the call is part of a pattern.
8611 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8613 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8614 that the sequence is attached to the original statement rather
8615 than the pattern statement.
8616 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8617 PATTERN_DEF_SEQ from the original statement rather than
8618 the main pattern statement.
8619 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8620 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8621 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8623 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8625 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8626 definition statements before the early exit for statements that aren't
8628 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8630 (vect_transform_loop): ...here. Process pattern definition
8631 statements without first checking whether the main pattern
8632 statement is live or relevant.
8634 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8636 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8637 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8639 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
8641 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8642 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8643 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8644 (expand_block_compare): Change select_block_compare_mode call.
8645 (expand_strncmp_align_check): Use new functions, fix comment.
8646 (emit_final_str_compare_gpr): New function.
8647 (expand_strn_compare): Refactor and clean up code.
8648 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8650 2018-06-19 Tony Reix <tony.reix@atos.com>
8651 Damien Bergamini <damien.bergamini@atos.com>
8652 David Edelsohn <dje.gcc@gmail.com>
8654 * collect2.c (static_obj): New variable.
8655 (static_libs): New variable.
8656 (is_in_list): Uncomment declaration.
8657 (main): Track AIX libraries linked statically.
8658 (is_in_list): Uncomment definition.
8659 (scan_prog_file): Don't add AIX shared libraries initializer
8660 to constructor list if linking statically.
8662 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
8664 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8666 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
8668 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
8670 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
8673 2018-06-19 Martin Liska <mliska@suse.cz>
8675 * config/i386/i386.c (ix86_can_inline_p): Do not use
8676 ipa_fn_summaries::get_create.
8677 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
8679 (devirtualization_time_bonus): Likewise.
8680 (ipcp_propagate_stage): Likewise.
8681 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8682 (edge_set_predicate): Likewise.
8683 (evaluate_conditions_for_known_args): Likewise.
8684 (evaluate_properties_for_edge): Likewise.
8685 (ipa_call_summary::reset): Tranform to ...
8686 (ipa_call_summary::~ipa_call_summary): ... this.
8687 (ipa_fn_summary::reset): Transform to ...
8688 (ipa_fn_summary::~ipa_fn_summary): ... this.
8689 (ipa_fn_summary_t::remove): Rename to ...
8690 (ipa_fn_summary_t::remove_callees): ... this.
8691 (ipa_fn_summary_t::duplicate): Use placement new
8692 instead of memory copy.
8693 (ipa_call_summary_t::duplicate): Likewise.
8694 (ipa_call_summary_t::remove): Remove.
8695 (dump_ipa_call_summary): Change get_create to get.
8696 (ipa_dump_fn_summary): Dump only when summary exists.
8697 (analyze_function_body): Use symbol_summary::get instead
8699 (compute_fn_summary): Likewise.
8700 (estimate_edge_devirt_benefit): Likewise.
8701 (estimate_edge_size_and_time): Likewise.
8702 (inline_update_callee_summaries): Likewise.
8703 (remap_edge_change_prob): Likewise.
8704 (remap_edge_summaries): Likewise.
8705 (ipa_merge_fn_summary_after_inlining): Likewise.
8706 (write_ipa_call_summary): Likewise.
8707 (ipa_fn_summary_write): Likewise.
8708 (ipa_free_fn_summary): Likewise.
8709 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
8710 (struct ipa_call_summary): Likewise.
8711 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
8713 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
8714 (estimate_size_after_inlining): Likewise.
8715 (estimate_growth): Likewise.
8716 (growth_likely_positive): Likewise.
8717 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8718 (inline_call): Likewise.
8719 * ipa-inline.c (caller_growth_limits): Likewise.
8720 (can_inline_edge_p): Likewise.
8721 (can_inline_edge_by_limits_p): Likewise.
8722 (compute_uninlined_call_time): Likewise.
8723 (compute_inlined_call_time): Likewise.
8724 (want_inline_small_function_p): Likewise.
8725 (edge_badness): Likewise.
8726 (update_caller_keys): Likewise.
8727 (update_callee_keys): Likewise.
8728 (inline_small_functions): Likewise.
8729 (inline_to_all_callers_1): Likewise.
8730 (dump_overall_stats): Likewise.
8731 (early_inline_small_functions): Likewise.
8732 (early_inliner): Likewise.
8733 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8734 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8735 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8736 * ipa-split.c (execute_split_functions): Likewise.
8737 * symbol-summary.h: Likewise.
8738 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8740 2018-06-19 Richard Biener <rguenther@suse.de>
8742 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
8743 (vectorize_loops): ... here. Fix dbgcnt handling.
8744 (try_vectorize_loop): Wrap try_vectorize_loop_1.
8746 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
8749 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8750 ieee128 argument takes up only one (vector) register, not two (floating
8753 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8755 * gimplify.c (gimplify_init_constructor): Really never clear for an
8756 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
8758 2018-06-19 Richard Biener <rguenther@suse.de>
8760 PR tree-optimization/86179
8761 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
8762 after failed recognition.
8764 2018-06-18 Martin Sebor <msebor@redhat.com>
8767 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
8768 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
8769 Handle integer subtraction.
8770 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8771 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
8773 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8775 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
8776 param from rtx to rtx_insn *.
8777 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
8779 (frv_ifcvt_modify_insn): Likwise.
8780 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8781 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
8782 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8783 as_a <rtx_insn *> cast to local "unprotected_region" once
8784 it's been established that it's not NULL or pc_rtx.
8785 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8786 param "sethi" from rtx to rtx_insn *.
8787 (nds32_group_float_insns): Likewise for param "insn".
8788 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8790 (vax_output_int_subtract): Likewise.
8791 * config/vax/vax.c (vax_output_int_add): Likewise for param
8793 (vax_output_int_subtract): Likewise.
8794 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8795 (emit_pattern_after): Likewise for param "after".
8796 (emit_insn_after): Likewise.
8797 (emit_jump_insn_after): Likewise.
8798 (emit_call_insn_after): Likewise.
8799 (emit_debug_insn_after): Likewise.
8800 (emit_pattern_before): Likewise for param "before".
8801 (emit_insn_before): Likewise.
8802 (emit_jump_insn_before): Likewise.
8803 * final.c (get_insn_template): Likewise for param "insn", removing
8805 * output.h (get_insn_template): Likewise for 2nd param.
8806 * rtl.h (emit_insn_before): Likewise.
8807 (emit_jump_insn_before): Likewise.
8808 (emit_debug_insn_before_noloc): Likewise.
8809 (emit_insn_after): Likewise.
8810 (emit_jump_insn_after): Likewise.
8811 (emit_call_insn_after): Likewise.
8812 (emit_debug_insn_after): Likewise.
8813 (set_insn_deleted): Likewise for param.
8815 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8818 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8819 floating point modes, so that IFmode is numerically greater than
8820 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8821 to declare the ordering. This prevents IFmode from being
8822 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8823 machine. Include rs6000-modes.h to share the fractional values
8824 between genmodes* and the rest of the compiler.
8828 * config/rs6000/rs6000-modes.h: New file.
8829 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8830 meaning of rs6000_long_double_size so that 126..128 selects an
8831 appropriate 128-bit floating point type.
8832 (rs6000_option_override_internal): Likewise.
8833 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8834 (TARGET_LONG_DOUBLE_128): Change the meaning of
8835 rs6000_long_double_size so that 126..128 selects an appropriate
8836 128-bit floating point type.
8837 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8838 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8839 source and destination to match the standard usage.
8840 (truncifkf2): Likewise.
8841 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8842 ISA 2.07 to use an explicit clobber, instead of passing in a
8844 (copysign<mode>3_soft): Likewise.
8846 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8848 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8849 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8850 (vect_slp_analyze_instance_dependence): Likewise.
8851 (vect_enhance_data_refs_alignment): Likewise.
8852 (vect_analyze_data_refs_alignment): Likewise.
8853 (vect_slp_analyze_and_verify_instance_alignment
8854 (vect_analyze_data_ref_accesses): Likewise.
8855 (vect_prune_runtime_alias_test_list): Likewise.
8856 (vect_analyze_data_refs): Likewise.
8857 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8858 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8859 (vect_analyze_scalar_cycles_1): Likewise.
8860 (vect_get_loop_niters): Likewise.
8861 (vect_analyze_loop_form_1): Likewise.
8862 (vect_update_vf_for_slp): Likewise.
8863 (vect_analyze_loop_operations): Likewise.
8864 (vect_analyze_loop): Likewise.
8865 (vectorizable_induction): Likewise.
8866 (vect_transform_loop): Likewise.
8867 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8868 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8869 (vect_make_slp_decision): Likewise.
8870 (vect_detect_hybrid_slp): Likewise.
8871 (vect_slp_analyze_operations): Likewise.
8872 (vect_slp_bb): Likewise.
8873 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8874 (vectorizable_bswap): Likewise.
8875 (vectorizable_call): Likewise.
8876 (vectorizable_simd_clone_call): Likewise.
8877 (vectorizable_conversion): Likewise.
8878 (vectorizable_assignment): Likewise.
8879 (vectorizable_shift): Likewise.
8880 (vectorizable_operation): Likewise.
8881 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8883 2018-06-18 Martin Sebor <msebor@redhat.com>
8885 PR tree-optimization/81384
8886 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8887 * builtins.c (expand_builtin_strnlen): New function.
8888 (expand_builtin): Call it.
8889 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8890 * builtins.def (BUILT_IN_STRNLEN): New.
8891 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8892 Warn for bounds in excess of maximum object size.
8893 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8894 single-value ranges. Handle strnlen.
8895 (handle_builtin_strlen): Handle strnlen.
8896 (strlen_check_and_optimize_stmt): Same.
8897 * doc/extend.texi (Other Builtins): Document strnlen.
8899 2018-06-18 Maya Rashish <coypu@sdf.org>
8901 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8902 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8903 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8905 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8907 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8909 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8911 * tree.c (escaped_string::escape): Replace cast to char * by
8912 const_cast<char *> (unescaped).
8914 2018-06-18 Nick Clifton <nickc@redhat.com>
8917 * tree.c (escaped_string): New class. Converts an unescaped
8918 string into its escaped equivalent.
8919 (warn_deprecated_use): Use the new class to convert the
8920 deprecation message, if present.
8921 (test_escaped_strings): New self test.
8922 (test_c_tests): Add test_escaped_strings.
8923 * doc/extend.texi (deprecated): Add a note that the
8924 deprecation message is affected by the -fmessage-length
8925 option, and that control characters will be escaped.
8926 (#pragma GCC error): Document this pragma.
8927 (#pragma GCC warning): Likewise.
8928 * doc/invoke.texi (-fmessage-length): Document this option's
8929 effect on the #warning and #error preprocessor directives and
8930 the deprecated attribute.
8932 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8934 * tree.c (decl_value_expr_lookup): Revert latest change.
8935 (decl_value_expr_insert): Likewise.
8937 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8939 * gimplify.c (nonlocal_vlas): Delete.
8940 (nonlocal_vla_vars): Likewise.
8941 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8943 (gimplify_body): Do not create and destroy nonlocal_vlas.
8944 * tree-nested.c: Include diagnostic.h.
8945 (use_pointer_in_frame): Tweak.
8946 (lookup_field_for_decl): Add assertion and declare the transformation.
8947 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8948 internal error when the reference is in a wrong context. Do not
8949 create a debug decl by default.
8950 (note_nonlocal_block_vlas): Delete.
8951 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8952 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8953 create a debug decl by default.
8954 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8955 call is in a wrong context.
8956 (fixup_vla_decls): New function.
8957 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8958 debug variables were created.
8959 * tree.c (decl_value_expr_lookup): Add checking assertion.
8960 (decl_value_expr_insert): Likewise.
8962 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8965 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8966 * tree-scalar-evolution.c (interpret_expr): Likewise.
8967 (expression_expensive_p): Likewise.
8968 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8969 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8970 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8971 (ssa_defined_by_minus_one_stmt_p): New.
8973 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8976 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8977 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8978 * dojump.c (do_jump): Likewise.
8979 * expr.c (expand_expr_real_2): Check operand type's sign.
8980 * fold-const.c (const_unop): Handle ABSU_EXPR.
8981 (fold_abs_const): Likewise.
8982 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8983 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8984 (strip_sign_op_1): Likesise.
8985 * match.pd: Add new pattern to generate ABSU_EXPR.
8986 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8987 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8988 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8989 * tree-inline.c (estimate_operator_cost): Likewise.
8990 * tree-pretty-print.c (dump_generic_node): Likewise.
8991 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8992 * tree.def (ABSU_EXPR): New.
8994 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8997 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8998 documented as preserved for backward compatibility only.
8999 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9001 PR rtl-optimization/86108
9002 * bb-reorder.c (create_forwarder_block): Renamed to ...
9003 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9004 jump from new landing pad to the second part.
9005 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9008 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9011 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9012 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9013 Only call store_expr for halves if the mode is the same.
9016 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9017 Fix up comment formatting.
9019 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9021 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9022 (typed_splay_tree::closure,
9023 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9024 (typed_splay_tree::typed_splay_tree,
9025 typed_splay_tree::operator =): Declared private.
9026 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9027 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9028 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9029 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9030 typed_splay_tree::splay_tree_splay,
9031 typed_splay_tree::splay_tree_foreach_helper,
9032 typed_splay_tree::splay_tree_insert,
9033 typed_splay_tree::splay_tree_remove,
9034 typed_splay_tree::splay_tree_lookup,
9035 typed_splay_tree::splay_tree_predecessor,
9036 typed_splay_tree::splay_tree_successor,
9037 typed_splay_tree::splay_tree_min,
9038 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9039 (typed_splay_tree::root, typed_splay_tree::comp,
9040 typed_splay_tree::delete_key,
9041 typed_splay_tree::delete_value): New data members.
9042 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9043 typed_splay_tree::remove.
9045 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9047 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9048 -mginv and -mno-ginv to the assembler.
9049 * config/mips/mips.opt (-mcrc): New option.
9051 * doc/invoke.text (-mcrc): Document.
9054 2018-06-15 Nick Clifton <nickc@redhat.com>
9057 * tree.c (escaped_string): New class. Converts an unescaped
9058 string into its escaped equivalent.
9059 (warn_deprecated_use): Use the new class to convert the
9060 deprecation message, if present.
9061 (test_escaped_strings): New self test.
9062 (test_c_tests): Add test_escaped_strings.
9063 * doc/extend.texi (deprecated): Add a note that the
9064 deprecation message is affected by the -fmessage-length
9065 option, and that control characters will be escaped.
9066 (#pragma GCC error): Document this pragma.
9067 (#pragma GCC warning): Likewise.
9068 * doc/invoke.texi (-fmessage-length): Document this option's
9069 effect on the #warning and #error preprocessor directives and
9070 the deprecated attribute.
9072 2018-06-15 Richard Biener <rguenther@suse.de>
9074 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9075 here, also noting vector size used.
9076 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9077 size used in MSG_OPTIMIZED_LOCATIONS dump.
9078 (pass_slp_vectorize::execute): Adjust.
9080 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9083 * config/arc/arc.c (arc_return_address_register): Fix
9086 2018-06-15 Richard Biener <rguenther@suse.de>
9089 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9090 leave useless conversion stripping to force_gimple_operand_gsi.
9091 (gimplify_build2): Likewise.
9092 (gimplify_build1): Likewise.
9094 2018-06-15 Richard Biener <rguenther@suse.de>
9097 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9098 before adjusting their block.
9100 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9102 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9103 multilibs for *-*-rtems*.
9104 * config/riscv/t-rtems: New file.
9106 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9109 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9110 unsigned_type_for returns NULL.
9113 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9114 subregs of multi-word pseudos unless the float mode has word size.
9116 2018-06-14 Richard Biener <rguenther@suse.de>
9119 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9120 duplicate and harmful type_hash_canon.
9121 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9123 2018-06-14 Richard Biener <rguenther@suse.de>
9126 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9129 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9131 * config/rtems.h (STDINT_LONG32): Define.
9133 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9134 Prachi Godbole <prachi.godbole@imgtec.com>
9136 * config/mips/mips-cpus.def: Define P6600.
9137 * config/mips/mips-tables.opt: Regenerate.
9138 * config/mips/mips.c (mips_ucbranch_type): New enum.
9139 (mips_rtx_cost_data): Add support for P6600.
9140 (mips_issue_rate): Likewise.
9141 (mips_multipass_dfa_lookahead): Likewise.
9142 (mips_avoid_hazard): Likewise.
9143 (mips_reorg_process_insns): Likewise.
9144 (mips_classify_branch_p6600): New function.
9145 * config/mips/mips.h (TUNE_P6600): New define.
9146 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9147 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9148 * config/mips/mips.md: Include p6600.md.
9149 (processor): Add p6600.
9150 * config/mips/p6600.md: New file.
9151 * doc/invoke.texi: Add p6600 to supported architectures.
9153 2018-06-13 Martin Sebor <msebor@redhat.com>
9155 PR tree-optimization/86114
9156 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9158 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9160 2018-06-13 Richard Biener <rguenther@suse.de>
9162 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9163 Properly set vector type of the intermediate stmt.
9164 * tree-vect-stmts.c (vectorizable_operation): The destination
9165 var always has vectype_out type.
9167 2018-06-13 Jeff Law <law@redhat.com>
9169 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9170 integer 0 for argument to print_rtl_with_bb.
9171 (rl78_reorg): Likewise.
9173 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9175 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9176 from rtx to rtx_insn *.
9177 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9179 (add_sched_insns_for_speculation): Likewise for local "target",
9180 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9181 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9182 from rtx_insn ** to rtx_code_label **.
9183 (reorg_emit_nops): Likewise.
9184 (c6x_reorg): Likewise for local "call_labels".
9185 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9187 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9188 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9189 the loops over LABEL_REFS.
9190 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9192 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9193 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9194 (split_branches): Strengthen local "olabel" from rtx to
9195 rtx_insn *, adding a safe_as_a cast.
9196 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9198 (add_insn_after): Likewise for first two params.
9199 (add_insn_before): Likewise.
9200 (remove_insn): Likewise for param.
9201 (emit_pattern_before_noloc): Likewise for second and third params.
9202 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9203 (emit_call_insn_before_noloc): Likewise.
9204 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9206 (emit_barrier_before): Likewise.
9207 (emit_label_before): Strengthen "label" param from "rtx" to
9208 "rtx_code_label *". Strengthen "before" param from "rtx" to
9210 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9212 (emit_pattern_after_noloc): Likewise.
9213 (emit_insn_after_noloc): Likewise.
9214 (emit_jump_insn_after_noloc): Likewise.
9215 (emit_call_insn_after_noloc): Likewise.
9216 (emit_debug_insn_after_noloc): Likewise.
9217 (emit_barrier_after): Likewise.
9218 (emit_label_after): Likewise for both params.
9219 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9220 "loc" param from "int" to "location_t".
9221 (emit_insn_after_setloc): Likewise.
9222 (emit_jump_insn_after_setloc): Likewise.
9223 (emit_call_insn_after_setloc): Likewise.
9224 (emit_debug_insn_after_setloc): Likewise.
9225 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9226 "loc" param from "int" to "location_t".
9227 (emit_pattern_before): Convert NULL_RTX to NULL.
9228 (emit_insn_before_setloc): Convert "loc" param from "int" to
9230 (emit_jump_insn_before_setloc): Likewise.
9231 (emit_call_insn_before_setloc): Likewise.
9232 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9233 rtx_insn *. Convert "loc" param from "int" to "location_t".
9234 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9235 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9236 Convert 3rd param from "int" to "location_t".
9237 (emit_barrier_before, emit_barrier_after, next_real_insn):
9238 Strengthen param from rtx to rtx_insn *.
9239 (emit_label_before): Strengthen 1st param from "rtx" to
9240 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9242 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9243 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9244 Strengthen 2nd param from "rtx" to "rtx_insn *".
9245 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9246 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9247 Likewise. Convert 3rd param from "int" to "location_t".
9248 (emit_label_after): Strengthen 1st param from "rtx" to
9250 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9252 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9253 from "rtx" to "rtx_insn *".
9255 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9257 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9258 bodies streamed in with -Q.
9259 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9260 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9261 * gimple-streamer-out.c: Include gimple-pretty-print.h
9262 (output_bb): Dump stmts streamed.
9263 * lto-section-out.c: Include print-tree.h
9264 (lto_begin_section): Dump sections created.
9265 (lto_output_decl_index): Dump decl encoded.
9266 * lto-streamer-out.c: Include print-tree.h
9267 (create_output_block): Dump output block created.
9268 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9269 (output_function): Dump function output.
9270 (output_constructor): Dump constructor streamed.
9271 (write_global_stream): Output indexes encoded.
9272 (produce_asm_for_decls): Dump streams encoded.
9273 * lto-streamer.c (streamer_dump_file): New global var.
9274 * lto-streamer.h (streamer_dump_file): Declare.
9275 * passes.c (ipa_write_summaries): Initialize streamer dump.
9276 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9279 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9282 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9283 offsets for register save directives. Emit a second batch of save
9284 directives, if need be, when the function accesses prior frames.
9286 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9288 * config/arc/fpu.md (fmasf4): Force operand to register.
9289 (fnmasf4): Likewise.
9291 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9293 * config/arc/arc-protos.h (arc_pad_return): Remove.
9294 * config/arc/arc.c (machine_function): Remove force_short_suffix
9296 (arc_print_operand): Adjust printing of '&'.
9297 (arc_verify_short): Remove conditional printing of short suffix.
9298 (arc_final_prescan_insn): Remove reference to size_reason.
9299 (pad_return): New function.
9300 (arc_reorg): Call pad_return.
9301 (arc_pad_return): Remove.
9302 (arc_init_machine_status): Remove reference to force_short_suffix.
9303 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9304 (attr length): When attribute iscompact is true force to 2
9305 regardless; in the case of maybe check if we want to force the
9306 instruction to have 4 bytes length.
9307 (nopv): Change it to generate 4 byte long nop as well.
9308 (blockage): New pattern.
9309 (simple_return): Remove call to arc_pad_return.
9310 (p_return_i): Likewise.
9312 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9314 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9316 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9318 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9321 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9323 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9324 for ARC700 and ARCv2.
9326 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9329 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9330 operands[2] instead of operands[1].
9333 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9335 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9336 case, check whether the outer register overlaps an unallocatable
9337 register, not just whether it fits the required class.
9339 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9341 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9343 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9344 (memrefs_conflict_p): Likewise.
9345 (init_alias_analysis): Likewise.
9346 * cfgexpand.c (expand_debug_expr): Likewise.
9347 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9348 * cse.c (fold_rtx): Likewise.
9349 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9350 * expr.c (emit_block_move_hints): Likewise.
9351 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9352 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9353 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9354 (emit_group_store): Likewise.
9355 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9356 to read the PRE/POST_MODIFY increment.
9357 * calls.c (store_one_arg): Use strip_offset.
9358 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9360 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9362 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9363 (simplify_binary_operation_1): Extend CONST_INT handling to
9365 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9366 than a HOST_WIDE_INT.
9367 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9369 (adjust_mems, add_stores): Update accodingly.
9370 (vt_canonicalize_addr): Track polynomial offsets.
9371 (emit_note_insn_var_location): Likewise.
9372 (vt_add_function_parameter): Likewise.
9373 (vt_initialize): Likewise.
9375 2018-06-12 Jeff Law <law@redhat.com>
9377 * config.gcc (alpha*-*-freebsd*): Remove.
9378 * config/alpha/freebsd.h: Remove.
9380 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9383 * spellcheck-tree.c (levenshtein_distance): Rename to...
9384 (get_edit_distance): ...this, and update for underlying renaming.
9385 * spellcheck-tree.h (levenshtein_distance): Rename to...
9386 (get_edit_distance): ...this.
9387 * spellcheck.c (levenshtein_distance): Rename to...
9388 (get_edit_distance): ...this. Convert from Levenshtein distance
9389 to Damerau-Levenshtein distance by supporting transpositions of
9390 adjacent characters. Rename "v1" to "v_next" and "v0" to
9392 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9393 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9394 update for underlying renaming.
9395 (selftest::levenshtein_distance_unit_test): Rename to...
9396 (selftest::test_get_edit_distance_unit): ...this, and update for
9397 underlying renaming.
9398 (selftest::test_find_closest_string): Add example from PR 69968
9399 where transposition helps
9400 (selftest::test_metric_conditions): Update for renaming.
9401 (selftest::test_metric_conditions): Likewise.
9402 (selftest::spellcheck_c_tests): Likewise.
9403 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9404 (get_edit_distance): ...this.
9405 (best_match::consider): Update for renaming.
9407 2018-06-12 Martin Sebor <msebor@redhat.com>
9409 PR tree-optimization/85259
9410 * builtins.c (compute_objsize): Handle constant offsets.
9411 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9412 true iff a warning has been issued.
9413 * gimple.h (gimple_nonartificial_location): New function.
9414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9415 gimple_nonartificial_location and handle -Wno-system-headers.
9416 (handle_builtin_stxncpy): Same.
9418 2018-06-12 Martin Sebor <msebor@redhat.com>
9421 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9423 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9425 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9426 BUILTIN_VEC_XST entries for pointer to double and long long.
9428 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9431 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9433 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9436 2018-06-12 Martin Liska <mliska@suse.cz>
9438 * doc/options.texi: Document IntegerRange.
9440 2018-06-12 Martin Liska <mliska@suse.cz>
9442 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9443 * opt-functions.awk: Handle Deprecated flag.
9444 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9446 (read_cmdline_option): Report warning for a deprecated option.
9447 * opts.h (struct cl_option): Add new field cl_deprecated.
9448 (CL_ERR_DEPRECATED): New.
9450 2018-06-12 Martin Liska <mliska@suse.cz>
9452 * doc/options.texi: Document Deprecated option flag.
9454 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9456 * config/arc/arc-arch.h (arc_extras): New enum.
9457 (arc_cpu_t):Add field extra.
9458 (arc_cpu_types): Consider the extras.
9459 * config/arc/arc-cpus.def: Add extras info.
9460 * config/arc/arc-opts.h (processor_type): Consider extra field.
9461 * config/arc/arc.c (arc_override_options): Handle extra field.
9463 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9465 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9466 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9468 * config/arc/arc.c (arc_init): Update pic warning.
9469 (irq_range): Update irq range parsing warnings.
9470 (arc_override_options): Update various warning messages.
9471 (arc_handle_aux_attribute): Likewise.
9473 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9475 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9477 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9479 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9480 compiler messages directives.
9482 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9484 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9485 * config/mips/mips-tables.opt: Regenerate.
9486 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9488 * doc/invoke.texi: Document -march=i6500.
9490 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9492 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9493 (i6400_gpmul): Add cpu_unit.
9494 (i6400_gpdiv): Likewise.
9495 (i6400_msa_add_d): Update reservations.
9496 (i6400_msa_int_add) Likewise.
9497 (i6400_msa_short_logic3) Likewise.
9498 (i6400_msa_short_logic2) Likewise.
9499 (i6400_msa_short_logic) Likewise.
9500 (i6400_msa_move) Likewise.
9501 (i6400_msa_cmp) Likewise.
9502 (i6400_msa_short_float2) Likewise.
9503 (i6400_msa_div_d) Likewise.
9504 (i6400_msa_long_logic1) Likewise.
9505 (i6400_msa_long_logic2) Likewise.
9506 (i6400_msa_mult) Likewise.
9507 (i6400_msa_long_float2) Likewise.
9508 (i6400_msa_long_float4) Likewise.
9509 (i6400_msa_long_float5) Likewise.
9510 (i6400_msa_long_float8) Likewise.
9511 (i6400_fpu_fadd): Include frint type.
9512 (i6400_fpu_store): New define_insn_reservation.
9513 (i6400_fpu_load): Likewise.
9514 (i6400_fpu_move): Likewise.
9515 (i6400_fpu_fcmp): Likewise.
9516 (i6400_fpu_fmadd): Likewise.
9517 (i6400_int_mult): Include imul3nc type and update reservation.
9518 (i6400_int_div): Include idiv3 type and update reservation.
9519 (i6400_int_load): Update to check type not move_type.
9520 (i6400_int_store): Likewise.
9521 (i6400_int_prefetch): Set zero latency.
9523 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9525 * gcc.c: Document new %@{...} sequence.
9526 (LINK_COMMAND_SPEC): Use it for the -L switches.
9527 (cpp_unique_options): Use it for the -I switches.
9528 (at_file_argbuf): New global variable.
9529 (in_at_file): Likewise.
9530 (alloc_args): Create at_file_argbuf.
9531 (clear_args): Truncate at_file_argbuf.
9532 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9533 (open_at_file): New function.
9534 (close_at_file): Likewise.
9535 (create_at_file): Delete.
9536 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9539 (validate_switches_from_spec): Deal with %@{...} sequence.
9540 (validate_switches): Likewise.
9541 (driver::finalize): Call clear_args.
9543 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9545 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9547 2018-06-11 Martin Sebor <msebor@redhat.com>
9549 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9550 (Wno-class-memaccess): Add @opindex.
9551 (Wno-templates, Wno-multiple-inheritance): Same.
9552 (Wno-virtual-inheritance, Wno-namespaces): Same.
9553 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9554 (Wno-format-overflow, Wno-format-truncation): Same.
9555 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9556 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9557 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9558 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9559 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9560 (Wno-misspelled-isr): Same.
9562 2018-06-11 Martin Sebor <msebor@redhat.com>
9564 * PR tree-optimization/86083
9565 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9567 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9569 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9571 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9574 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9575 on the correct operand.
9576 (*movdi_internal64): Ditto.
9578 2018-06-11 Martin Liska <mliska@suse.cz>
9580 PR tree-optimization/86089
9581 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9583 2018-06-11 Julia Koval <julia.koval@intel.com>
9585 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9586 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9587 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9589 2018-06-11 Olivier Hainque <hainque@adacore.com>
9591 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9592 for Ada with strict dwarf2.
9594 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9597 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9600 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9602 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9604 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9606 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9609 2018-06-08 Martin Liska <mliska@suse.cz>
9611 * tree-cfg.h (debug_function): Fix argument type to match
9614 2018-06-08 Martin Liska <mliska@suse.cz>
9616 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9617 Remove usage of MPX-related (and removed) fields.
9618 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9620 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9622 * cfg.c (debug): Use TDF_NONE rather than 0.
9623 * cfghooks.c (debug): Likewise.
9624 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9625 (struct dump_option_value_info): Convert to...
9626 (struct kv_pair): ...this template type.
9627 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9629 (optinfo_verbosity_options): Likewise.
9630 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9632 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9633 than int for "optgroup_flags" param.
9634 (dump_generic_expr_loc): Use dump_flags_t rather than int for
9636 (dump_dec): Likewise.
9637 (dump_finish): Use TDF_NONE rather than 0.
9638 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9639 rather than int for "optgroup_flags" param. Use TDF_NONE rather
9640 than 0. Update for change to option_ptr.
9641 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9642 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
9643 0. Update for changes to optinfo_verbosity_options and
9645 (opt_info_switch_p): Convert optgroup_flags from int to
9647 (dump_basic_block): Use dump_flags_t rather than int
9648 for "dump_kind" param.
9649 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9650 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9651 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9652 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9653 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9654 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9655 TDF_NONE): Convert from macros to...
9656 (enum dump_flag): ...this new enum.
9657 (dump_flags_t): Update to use enum.
9658 (operator|, operator&, operator~, operator|=, operator&=):
9659 Implement for dump_flags_t.
9660 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9661 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9662 Convert from macros to...
9663 (enum optgroup_flag): ...this new enum.
9664 (optgroup_flags_t): New typedef.
9665 (operator|, operator|=): Implement for optgroup_flags_t.
9666 (struct dump_file_info): Convert field "alt_flags" to
9667 dump_flags_t. Convert field "optgroup_flags" to
9669 (dump_basic_block): Use dump_flags_t rather than int for param.
9670 (dump_generic_expr_loc): Likewise.
9671 (dump_dec): Likewise.
9672 (dump_register): Convert param "optgroup_flags" to
9674 (opt_info_enable_passes): Likewise.
9675 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
9677 * gimple-pretty-print.c (debug): Likewise.
9678 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
9679 (merged_store_group::apply_stores): Likewise.
9680 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
9681 * gimple.c (verify_gimple_pp): Likewise.
9682 * graphite-poly.c (print_pbb_body): Likewise.
9683 * passes.c (pass_manager::register_one_dump_file): Convert
9684 local "optgroup_flags" to optgroup_flags_t.
9685 * print-tree.c (print_node): Use TDF_NONE rather than 0.
9687 (debug_body): Likewise.
9688 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
9689 to optgroup_flags_t.
9690 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
9692 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
9693 (convert_mult_to_fma): Likewise.
9694 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
9695 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
9696 * tree-vect-data-refs.c (dump_lower_bound): Convert param
9697 "dump_kind" to dump_flags_t.
9699 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9701 * config/rs6000/rs6000.c (min, max): Delete.
9703 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9705 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
9706 -mabi=spe and -mabi=no-spe.
9708 2018-06-08 Martin Liska <mliska@suse.cz>
9710 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
9711 where we expect an existing summary.
9713 2018-06-08 Martin Liska <mliska@suse.cz>
9715 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
9716 * ipa-inline.h (estimate_edge_growth): Likewise.
9718 2018-06-08 Martin Liska <mliska@suse.cz>
9720 * cgraph.c (function_version_hasher::hash): Use
9721 cgraph_node::get_uid ().
9722 (function_version_hasher::equal):
9723 * cgraph.h (cgraph_node::get_uid): New method.
9724 * ipa-inline.c (update_caller_keys): Use
9725 cgraph_node::get_uid ().
9726 (update_callee_keys): Likewise.
9727 * ipa-utils.c (searchc): Likewise.
9728 (ipa_reduced_postorder): Likewise.
9729 * lto-cgraph.c (input_node): Likewise.
9730 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
9731 * symbol-summary.h (symtab_insertion): Likewise.
9732 (symtab_removal): Likewise.
9733 (symtab_duplication): Likewise.
9734 * tree-pretty-print.c (dump_function_header): Likewise.
9735 * tree-sra.c (convert_callers_for_node): Likewise.
9737 2018-06-08 Martin Liska <mliska@suse.cz>
9739 * cgraph.c (symbol_table::create_edge): Always assign a new
9741 (symbol_table::free_edge): Do not recycle numbers.
9742 * cgraph.h (cgraph_edge::get): New method.
9743 * symbol-summary.h (symtab_removal): Use it.
9744 (symtab_duplication): Likewise.
9745 (call_summary::hashable_uid): Remove.
9747 2018-06-08 Martin Liska <mliska@suse.cz>
9749 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
9750 (initialize_growth_caches): Remove.
9751 (free_growth_caches): Likewise.
9752 (do_estimate_edge_time): Use edge_growth_cache.
9753 (do_estimate_edge_size): Likewise.
9754 (do_estimate_edge_hints): Likewise.
9755 * ipa-inline.c (reset_edge_caches): Likewise.
9756 (recursive_inlining): Likewise.
9757 (inline_small_functions): Likewise.
9758 * ipa-inline.h (initialize_growth_caches): Remove.
9759 (estimate_edge_size): Likewise.
9760 (estimate_edge_time): Likewise.
9761 (estimate_edge_hints): Likewise.
9762 (reset_edge_growth_cache): Likewise.
9763 * symbol-summary.h (call_summary::remove): New method.
9765 2018-06-08 Martin Liska <mliska@suse.cz>
9767 * ipa-cp.c (class edge_clone_summary): New summary.
9768 (grow_edge_clone_vectors): Remove.
9769 (ipcp_edge_duplication_hook): Remove.
9770 (class edge_clone_summary_t): New call_summary class.
9771 (ipcp_edge_removal_hook): Remove.
9772 (edge_clone_summary_t::duplicate): New function.
9773 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
9774 (create_specialized_node): Likewise.
9775 (ipcp_driver): Initialize edge_clone_summaries and do not
9778 2018-06-08 Martin Liska <mliska@suse.cz>
9780 * symbol-summary.h (get): New function.
9781 (call_summary::m_initialize_when_cloning): New class member.
9783 2018-06-08 Martin Liska <mliska@suse.cz>
9785 * cgraph.c (cgraph_node::remove): Do not recycle uid.
9786 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9787 (symbol_table::allocate_cgraph_symbol): Do not set uid.
9788 * passes.c (uid_hash_t): Record removed_nodes by their uids.
9789 (remove_cgraph_node_from_order): Use the removed_nodes set.
9790 (do_per_function_toporder): Likwise.
9791 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9792 instead of summary_uid.
9793 (symtab_removal): Likewise.
9794 (symtab_duplication): Likewise.
9796 2018-06-08 Martin Liska <mliska@suse.cz>
9798 * ipa-cp.c (ipcp_store_bits_results): Use
9799 ipcp_transformation_sum.
9800 (ipcp_store_vr_results): Likewise.
9801 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9803 (ipcp_transformation_initialize): ... this.
9804 (ipa_set_node_agg_value_chain):
9805 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9806 (write_ipcp_transformation_info): Likewise.
9807 (read_ipcp_transformation_info): Likewise.
9808 (ipcp_update_bits): Likewise.
9809 (ipcp_update_vr): Likewise.
9810 (ipcp_transform_function): Likewise.
9811 * ipa-prop.h: Rename ipcp_transformation_summary to
9812 ipcp_transformation.
9813 (class ipcp_transformation_t): New function summary.
9814 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9815 (ipa_get_agg_replacements_for_node): Likewise.
9817 2018-06-08 Martin Liska <mliska@suse.cz>
9819 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9821 (class funct_state_summary_t): New function_summary class.
9822 (has_function_state): Remove.
9823 (get_function_state): Likewise.
9824 (set_function_state): Likewise.
9825 (add_new_function): Likewise.
9826 (funct_state_summary_t::insert): New function.
9827 (duplicate_node_data): Remove.
9828 (remove_node_data): Remove.
9829 (funct_state_summary_t::duplicate): New function.
9830 (register_hooks): Create new funct_state_summaries.
9831 (pure_const_generate_summary): Use it.
9832 (pure_const_write_summary): Likewise.
9833 (pure_const_read_summary): Likewise.
9834 (propagate_pure_const): Likewise.
9835 (propagate_nothrow): Likewise.
9836 (dump_malloc_lattice): Likewise.
9837 (propagate_malloc): Likewise.
9838 (execute): Do not register hooks, just remove summary
9840 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9843 2018-06-08 Martin Liska <mliska@suse.cz>
9845 * ipa-reference.c (remove_node_data): Remove.
9846 (duplicate_node_data): Likewise.
9847 (class ipa_ref_var_info_summary_t): New class.
9848 (class ipa_ref_opt_summary_t): Likewise.
9849 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9850 (get_reference_optimization_summary): Use
9851 ipa_ref_opt_sum_summaries.
9852 (set_reference_vars_info): Remove.
9853 (set_reference_optimization_summary): Likewise.
9854 (ipa_init): Create summaries.
9855 (init_function_info): Use function summary.
9856 (ipa_ref_opt_summary_t::duplicate): New function.
9857 (ipa_ref_opt_summary_t::remove): New function.
9858 (get_read_write_all_from_node): Fix GNU coding style.
9859 (propagate): Use function summary.
9860 (write_node_summary_p): Fix GNU coding style.
9861 (stream_out_bitmap): Likewise.
9862 (ipa_reference_read_optimization_summary): Use function summary.
9863 (ipa_reference_c_finalize): Do not release hooks.
9865 2018-06-08 Martin Liska <mliska@suse.cz>
9867 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9868 (analyze_function_body): Extract multiple calls of get_create.
9869 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9870 * ipa-inline.c (recursive_inlining): Use ::get method.
9871 * ipa-inline.h (estimate_edge_growth): Likewise.
9873 2018-06-08 Martin Liska <mliska@suse.cz>
9875 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9877 (hsa_function_summary::hsa_function_summary): Use the new enum
9879 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9880 * hsa-gen.c (hsa_get_host_function): Likewise.
9881 (get_brig_function_name): Likewise.
9882 * ipa-hsa.c (process_hsa_functions): Likewise.
9883 (ipa_hsa_write_summary): Likewise.
9884 * symbol-summary.h (symtab_duplication): Use ::get function/
9885 (get): New function.
9887 2018-06-08 Martin Liska <mliska@suse.cz>
9889 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9891 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9892 (hsa_register_kernel): Likewise.
9893 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9894 * hsa-gen.c (hsa_get_host_function): Likewise.
9895 (get_brig_function_name): Likewise.
9896 (generate_hsa): Likewise.
9897 (pass_gen_hsail::execute): Likewise.
9898 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9899 (devirtualization_time_bonus): Likewise.
9900 (ipcp_propagate_stage): Likewise.
9901 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9902 (edge_set_predicate): Likewise.
9903 (evaluate_conditions_for_known_args): Likewise.
9904 (evaluate_properties_for_edge): Likewise.
9905 (ipa_fn_summary::reset): Likewise.
9906 (ipa_fn_summary_t::duplicate): Likewise.
9907 (dump_ipa_call_summary): Likewise.
9908 (ipa_dump_fn_summary): Likewise.
9909 (analyze_function_body): Likewise.
9910 (compute_fn_summary): Likewise.
9911 (estimate_edge_devirt_benefit): Likewise.
9912 (estimate_edge_size_and_time): Likewise.
9913 (estimate_calls_size_and_time): Likewise.
9914 (estimate_node_size_and_time): Likewise.
9915 (inline_update_callee_summaries): Likewise.
9916 (remap_edge_change_prob): Likewise.
9917 (remap_edge_summaries): Likewise.
9918 (ipa_merge_fn_summary_after_inlining): Likewise.
9919 (ipa_update_overall_fn_summary): Likewise.
9920 (read_ipa_call_summary): Likewise.
9921 (inline_read_section): Likewise.
9922 (write_ipa_call_summary): Likewise.
9923 (ipa_fn_summary_write): Likewise.
9924 (ipa_free_fn_summary): Likewise.
9925 * ipa-hsa.c (process_hsa_functions): Likewise.
9926 (ipa_hsa_write_summary): Likewise.
9927 (ipa_hsa_read_section): Likewise.
9928 * ipa-icf.c (sem_function::merge): Likewise.
9929 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9930 (do_estimate_edge_time): Likewise.
9931 (estimate_size_after_inlining): Likewise.
9932 (estimate_growth): Likewise.
9933 (growth_likely_positive): Likewise.
9934 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9935 (inline_call): Likewise.
9936 * ipa-inline.c (caller_growth_limits): Likewise.
9937 (can_inline_edge_p): Likewise.
9938 (can_inline_edge_by_limits_p): Likewise.
9939 (compute_uninlined_call_time): Likewise.
9940 (compute_inlined_call_time): Likewise.
9941 (want_inline_small_function_p): Likewise.
9942 (edge_badness): Likewise.
9943 (update_caller_keys): Likewise.
9944 (update_callee_keys): Likewise.
9945 (recursive_inlining): Likewise.
9946 (inline_small_functions): Likewise.
9947 (inline_to_all_callers_1): Likewise.
9948 (dump_overall_stats): Likewise.
9949 (early_inline_small_functions): Likewise.
9950 (early_inliner): Likewise.
9951 * ipa-inline.h (estimate_edge_growth): Likewise.
9952 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9953 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9954 * ipa-prop.h (IPA_NODE_REF): Likewise.
9955 (IPA_EDGE_REF): Likewise.
9956 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9957 (propagate_malloc): Likewise.
9958 * ipa-split.c (execute_split_functions): Likewise.
9959 * symbol-summary.h: Rename get to get_create.
9961 (get_create): Likewise.
9962 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9964 2018-06-08 Martin Liska <mliska@suse.cz>
9966 * symbol-summary.h (release): Move definition out of class
9968 (symtab_removal): Likewise.
9969 (symtab_duplication): Likewise.
9971 2018-06-08 Martin Liska <mliska@suse.cz>
9973 * symbol-summary.h (function_summary): Move constructor
9974 implementation out of class declaration.
9975 (release): Likewise.
9976 (symtab_insertion): Likewise.
9977 (symtab_removal): Likewise.
9978 (symtab_duplication): Likewise.
9981 2018-06-08 Martin Liska <mliska@suse.cz>
9983 * Makefile.in: Remove support for MPX (macros, related functions,
9984 fields in cgraph_node, ...).
9985 * builtin-types.def (BT_BND): Likewise.
9986 (BT_FN_BND_CONST_PTR): Likewise.
9987 (BT_FN_CONST_PTR_BND): Likewise.
9988 (BT_FN_VOID_PTR_BND): Likewise.
9989 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9990 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9991 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9992 (expand_builtin_mempcpy_with_bounds): Likewise.
9993 (expand_builtin_memset_with_bounds): Likewise.
9994 (expand_builtin_memset_args): Likewise.
9995 (std_expand_builtin_va_start): Likewise.
9996 (expand_builtin): Likewise.
9997 (expand_builtin_with_bounds): Likewise.
9998 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9999 (DEF_LIB_BUILTIN_CHKP): Likewise.
10000 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10001 (DEF_CHKP_BUILTIN): Likewise.
10002 (BUILT_IN_MEMCPY): Likewise.
10003 (BUILT_IN_MEMMOVE): Likewise.
10004 (BUILT_IN_MEMPCPY): Likewise.
10005 (BUILT_IN_MEMSET): Likewise.
10006 (BUILT_IN_STPCPY): Likewise.
10007 (BUILT_IN_STRCAT): Likewise.
10008 (BUILT_IN_STRCHR): Likewise.
10009 (BUILT_IN_STRCPY): Likewise.
10010 (BUILT_IN_STRLEN): Likewise.
10011 (BUILT_IN_MEMCPY_CHK): Likewise.
10012 (BUILT_IN_MEMMOVE_CHK): Likewise.
10013 (BUILT_IN_MEMPCPY_CHK): Likewise.
10014 (BUILT_IN_MEMSET_CHK): Likewise.
10015 (BUILT_IN_STPCPY_CHK): Likewise.
10016 (BUILT_IN_STRCAT_CHK): Likewise.
10017 (BUILT_IN_STRCPY_CHK): Likewise.
10018 * calls.c (store_bounds): Likewise.
10019 (emit_call_1): Likewise.
10020 (special_function_p): Likewise.
10021 (maybe_warn_nonstring_arg): Likewise.
10022 (initialize_argument_information): Likewise.
10023 (finalize_must_preallocate): Likewise.
10024 (compute_argument_addresses): Likewise.
10025 (expand_call): Likewise.
10026 * cfgexpand.c (expand_call_stmt): Likewise.
10027 (expand_return): Likewise.
10028 (expand_gimple_stmt_1): Likewise.
10029 (pass_expand::execute): Likewise.
10030 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10031 (cgraph_node::remove): Likewise.
10032 (cgraph_node::dump): Likewise.
10033 (cgraph_node::verify_node): Likewise.
10034 * cgraph.h (chkp_function_instrumented_p): Likewise.
10035 (symtab_node::get_alias_target): Likewise.
10036 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10037 (cgraph_local_p): Likewise.
10038 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10039 (cgraph_edge::rebuild_references): Likewise.
10040 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10041 (walk_polymorphic_call_targets): Likewise.
10042 (cgraph_node::expand_thunk): Likewise.
10043 (symbol_table::output_weakrefs): Likewise.
10044 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10045 (ix86_handle_option): Likewise.
10046 * config/i386/constraints.md: Likewise.
10047 * config/i386/i386-builtin-types.def (BND): Likewise.
10051 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10052 (BDESC_FIRST): Likewise.
10054 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10055 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10056 * config/i386/i386.c (enum reg_class): Likewise.
10057 (ix86_target_string): Likewise.
10058 (ix86_option_override_internal): Likewise.
10059 (ix86_conditional_register_usage): Likewise.
10060 (ix86_valid_target_attribute_inner_p): Likewise.
10061 (ix86_set_indirect_branch_type): Likewise.
10062 (ix86_set_current_function): Likewise.
10063 (ix86_function_arg_regno_p): Likewise.
10064 (init_cumulative_args): Likewise.
10065 (ix86_function_arg_advance): Likewise.
10066 (ix86_function_arg): Likewise.
10067 (ix86_pass_by_reference): Likewise.
10068 (ix86_function_value_regno_p): Likewise.
10069 (ix86_function_value_1): Likewise.
10070 (ix86_function_value_bounds): Likewise.
10071 (ix86_return_in_memory): Likewise.
10072 (ix86_setup_incoming_vararg_bounds): Likewise.
10073 (ix86_va_start): Likewise.
10074 (indirect_thunk_need_prefix): Likewise.
10075 (print_reg): Likewise.
10076 (ix86_print_operand): Likewise.
10077 (ix86_expand_call): Likewise.
10078 (ix86_output_function_return): Likewise.
10079 (reg_encoded_number): Likewise.
10080 (BDESC_VERIFYS): Likewise.
10081 (ix86_init_mpx_builtins): Likewise.
10082 (ix86_init_builtins): Likewise.
10083 (ix86_emit_cmove): Likewise.
10084 (ix86_emit_move_max): Likewise.
10085 (ix86_expand_builtin): Likewise.
10086 (ix86_builtin_mpx_function): Likewise.
10087 (ix86_get_arg_address_for_bt): Likewise.
10088 (ix86_load_bounds): Likewise.
10089 (ix86_store_bounds): Likewise.
10090 (ix86_load_returned_bounds): Likewise.
10091 (ix86_store_returned_bounds): Likewise.
10092 (ix86_class_likely_spilled_p): Likewise.
10093 (ix86_hard_regno_mode_ok): Likewise.
10094 (x86_order_regs_for_local_alloc): Likewise.
10095 (ix86_mitigate_rop): Likewise.
10096 (ix86_bnd_prefixed_insn_p): Likewise.
10097 (ix86_mpx_bound_mode): Likewise.
10098 (ix86_make_bounds_constant): Likewise.
10099 (ix86_initialize_bounds): Likewise.
10100 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10101 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10102 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10103 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10104 (TARGET_CHKP_BOUND_MODE): Likewise.
10105 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10106 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10107 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10108 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10109 * config/i386/i386.h (TARGET_MPX): Likewise.
10110 (TARGET_MPX_P): Likewise.
10111 (VALID_BND_REG_MODE): Likewise.
10112 (FIRST_BND_REG): Likewise.
10113 (LAST_BND_REG): Likewise.
10114 (enum reg_class): Likewise.
10115 (BND_REG_P): Likewise.
10116 (BND_REGNO_P): Likewise.
10117 (BNDmode): Likewise.
10118 (ADJUST_INSN_LENGTH): Likewise.
10119 * config/i386/i386.md: Likewise.
10120 * config/i386/i386.opt: Likewise.
10121 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10122 (defined): Likewise.
10123 (LINK_MPX): Likewise.
10124 (MPX_SPEC): Likewise.
10125 (LIBMPX_SPEC): Likewise.
10126 (LIBMPXWRAPPERS_SPEC): Likewise.
10127 (CHKP_SPEC): Likewise.
10128 * config/i386/predicates.md: Likewise.
10129 * dbxout.c (dbxout_type): Likewise.
10130 * doc/extend.texi: Likewise.
10131 * doc/invoke.texi: Likewise.
10132 * doc/md.texi: Likewise.
10133 * doc/tm.texi: Likewise.
10134 * doc/tm.texi.in: Likewise.
10135 * dwarf2out.c (is_base_type): Likewise.
10136 (gen_formal_types_die): Likewise.
10137 (gen_subprogram_die): Likewise.
10138 (gen_type_die_with_usage): Likewise.
10139 (gen_decl_die): Likewise.
10140 (dwarf2out_late_global_decl): Likewise.
10141 * expr.c (expand_assignment): Likewise.
10142 (emit_storent_insn): Likewise.
10143 (store_expr_with_bounds): Likewise.
10144 (store_expr): Likewise.
10145 (expand_expr_real_1): Likewise.
10146 * expr.h (store_expr_with_bounds): Likewise.
10147 * function.c (use_register_for_decl): Likewise.
10148 (struct bounds_parm_data): Likewise.
10149 (assign_parms_augmented_arg_list): Likewise.
10150 (assign_parm_find_entry_rtl): Likewise.
10151 (assign_parm_is_stack_parm): Likewise.
10152 (assign_parm_load_bounds): Likewise.
10153 (assign_bounds): Likewise.
10154 (assign_parms): Likewise.
10155 (expand_function_start): Likewise.
10156 * gcc.c (CHKP_SPEC): Likewise.
10157 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10158 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10159 (wrestrict_dom_walker::check_call): Likewise.
10160 * gimple.c (gimple_build_call_from_tree): Likewise.
10161 * gimple.h (enum gf_mask): Likewise.
10162 (gimple_call_with_bounds_p): Likewise.
10163 (gimple_call_set_with_bounds): Likewise.
10164 * gimplify.c (gimplify_init_constructor): Likewise.
10165 * ipa-cp.c (initialize_node_lattices): Likewise.
10166 (propagate_constants_across_call): Likewise.
10167 (find_more_scalar_values_for_callers_subset): Likewise.
10168 * ipa-hsa.c (process_hsa_functions): Likewise.
10169 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10170 * ipa-icf.c (sem_function::merge): Likewise.
10171 * ipa-inline.c (early_inliner): Likewise.
10172 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10173 (warn_function_cold): Likewise.
10174 (propagate_pure_const): Likewise.
10175 * ipa-ref.h (enum GTY): Likewise.
10176 * ipa-split.c (find_retbnd): Likewise.
10177 (consider_split): Likewise.
10178 (split_function): Likewise.
10179 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10180 * ipa.c (walk_polymorphic_call_targets): Likewise.
10181 (symbol_table::remove_unreachable_nodes): Likewise.
10182 (process_references): Likewise.
10183 (cgraph_build_static_cdtor_1): Likewise.
10184 * lto-cgraph.c (lto_output_node): Likewise.
10185 (output_refs): Likewise.
10186 (compute_ltrans_boundary): Likewise.
10187 (input_overwrite_node): Likewise.
10188 (input_node): Likewise.
10189 (input_cgraph_1): Likewise.
10190 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10191 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10192 (class pass_chkp_instrumentation_passes): Likewise.
10193 (make_pass_chkp_instrumentation_passes): Likewise.
10194 * passes.def: Likewise.
10195 * rtl.h (struct GTY): Likewise.
10196 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10197 * stor-layout.c (layout_type): Likewise.
10198 * symtab.c: Likewise.
10199 * target.def: Likewise.
10200 * targhooks.c (default_chkp_bound_type): Likewise.
10201 (default_chkp_bound_mode): Likewise.
10202 (default_builtin_chkp_function): Likewise.
10203 (default_chkp_function_value_bounds): Likewise.
10204 (default_chkp_make_bounds_constant): Likewise.
10205 (default_chkp_initialize_bounds): Likewise.
10206 * targhooks.h (default_chkp_bound_type): Likewise.
10207 (default_chkp_bound_mode): Likewise.
10208 (default_builtin_chkp_function): Likewise.
10209 (default_chkp_function_value_bounds): Likewise.
10210 (default_chkp_make_bounds_constant): Likewise.
10211 (default_chkp_initialize_bounds): Likewise.
10212 * toplev.c (compile_file): Likewise.
10213 (process_options): Likewise.
10214 * tree-core.h (DEF_BUILTIN): Likewise.
10215 (DEF_BUILTIN_CHKP): Likewise.
10216 * tree-inline.c (declare_return_variable): Likewise.
10217 (remap_gimple_stmt): Likewise.
10218 (copy_bb): Likewise.
10219 (initialize_inlined_parameters): Likewise.
10220 (expand_call_inline): Likewise.
10221 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10222 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10223 (make_pass_ipa_chkp_produce_thunks): Likewise.
10224 (make_pass_chkp): Likewise.
10225 (make_pass_chkp_opt): Likewise.
10226 (make_pass_chkp_instrumentation_passes): Likewise.
10227 * tree-pretty-print.c (dump_generic_node): Likewise.
10228 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10229 * tree-ssa-dce.c (propagate_necessity): Likewise.
10230 (eliminate_unnecessary_stmts): Likewise.
10231 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10232 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10233 * tree-ssa-sccvn.h: Likewise.
10234 * tree-ssa-strlen.c (get_string_length): Likewise.
10235 (valid_builtin_call): Likewise.
10236 (adjust_last_stmt): Likewise.
10237 (handle_builtin_strchr): Likewise.
10238 (handle_builtin_strcpy): Likewise.
10239 (handle_builtin_stxncpy): Likewise.
10240 (handle_builtin_memcpy): Likewise.
10241 (handle_builtin_strcat): Likewise.
10242 (strlen_check_and_optimize_stmt): Likewise.
10243 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10244 * tree-streamer-in.c: Likewise.
10245 * tree-streamer.c (record_common_node): Likewise.
10246 * tree.c (tree_code_size): Likewise.
10247 (wide_int_to_tree_1): Likewise.
10248 (type_contains_placeholder_1): Likewise.
10249 (build_common_tree_nodes): Likewise.
10250 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10251 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10252 (POINTER_BOUNDS_P): Likewise.
10253 (BOUNDED_TYPE_P): Likewise.
10254 (BOUNDED_P): Likewise.
10255 (CALL_WITH_BOUNDS_P): Likewise.
10256 (pointer_bounds_type_node): Likewise.
10257 * value-prof.c (gimple_ic): Likewise.
10258 * var-tracking.c (vt_add_function_parameters): Likewise.
10259 * varasm.c (make_decl_rtl): Likewise.
10260 (assemble_start_function): Likewise.
10261 (output_constant): Likewise.
10262 (maybe_assemble_visibility): Likewise.
10263 * varpool.c (ctor_for_folding): Likewise.
10264 * chkp-builtins.def: Remove.
10265 * ipa-chkp.c: Remove.
10266 * ipa-chkp.h: Remove.
10267 * rtl-chkp.c: Remove.
10268 * rtl-chkp.h: Remove.
10269 * tree-chkp-opt.c: Remove.
10270 * tree-chkp.c: Remove.
10271 * tree-chkp.h: Remove.
10273 2018-06-07 Carl Love <cel@us.ibm.com>
10275 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10276 vextract_fp_from_shortl): Add BE support.
10278 2018-06-07 Paul Koning <ni1d@arrl.net>
10280 * compare-elim.c (try_merge_compare): Don't merge compare if
10281 address contains a side effect.
10282 (try_eliminate_compare): Likewise.
10284 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10286 * config.gcc: Support "tremont".
10287 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10288 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10290 * config/i386/i386.c (m_TREMONT): Define.
10291 (processor_target_table): Add "tremont".
10292 (PTA_TREMONT): Define.
10293 (ix86_lea_outperforms): Add TARGET_TREMONT.
10294 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10295 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10296 and M_INTEL_GOLDMONT_PLUS.
10297 (fold_builtin_cpu): Add "tremont".
10298 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10299 (ix86_option_override_internal): Add "tremont".
10300 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10301 (processor_type): Add PROCESSOR_TREMONT.
10302 * config/i386/x86-tune.def: Add m_TREMONT.
10303 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10305 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10307 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10308 symbol defined for msp430i* devices to be lower case.
10310 2018-06-07 Richard Biener <rguenther@suse.de>
10312 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10313 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10314 Properly wrap signed arithmetic if overflow wraps.
10316 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10318 PR tree-optimization/69615
10319 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10320 of a cast from a same precision integral SSA_NAME in a bb dominated
10321 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10322 cast to utype if rhs2 has already a compatible type.
10324 2018-06-07 Richard Biener <rguenther@suse.de>
10326 PR tree-optimization/85935
10327 * graphite-scop-detection.c (find_params_in_bb): Analyze
10328 condition operands with respect to the correct loop. Assert
10329 the analysis doesn't fail.
10331 2018-06-04 Carl Love <cel@us.ibm.com>
10333 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10334 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10335 as it is slightly cheaper.
10336 (first_match_or_eos_index_<mode>):
10337 Calculate index using natural element order.
10338 (first_match_index_<mode>):
10339 Calculate index using natural element order.
10340 (first_match_or_eos_index_<mode>):
10341 Calculate index using natural order.
10342 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10343 for BE and LE modes.
10344 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10345 P9V_BUILTIN_VCLZLSBB_V16QI.
10346 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10349 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10351 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10352 indentation and line wrap for many prototypes. Add missing
10353 @smallexample directives around block of prototypes for vec_xl and
10356 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10358 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10359 track if we pass or return IEEE 128-bit floating point.
10360 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10361 C++ mangling that is compatible with GCC 8.1.
10362 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10363 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10364 floating point types.
10365 (rs6000_function_arg_advance_1): Likewise.
10366 (rs6000_mangle_type): Optionally generate mangled names that match
10367 what GCC 8.1 generated for IEEE 128-bit floating point types.
10368 (rs6000_globalize_decl_name): If we have an external function that
10369 passes or returns IEEE 128-bit types, generate a weak reference
10370 from the mangled name used in GCC 8.1 to the current mangled
10372 (rs6000_init_builtins): Make __ibm128 use the long double type if
10373 long double is IBM extended double. Make __float128 use the long
10374 double type if long double is IEEE 128-bit.
10377 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10378 macro for __ibm128 built-in functions.
10379 (PACK_IF): Add __ibm128 pack/unpack functions.
10380 (UNPACK_IF): Likewise.
10381 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10382 enable long double built-in functions if long double is IEEE
10383 128-bit floating point.
10384 (rs6000_invalid_builtin): Update long double built-in function
10386 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10387 functions, adjust the built-in function to use the long double
10388 built-in function if __ibm128 and long double are the same type.
10389 * doc/extend.texi (PowerPC builtins): Update documention for
10390 __builtin_{,un}pack_longdouble. Add documentation for
10391 __builtin_{,un}pack_ibm128.
10393 2018-06-06 Jim Wilson <jimw@sifive.com>
10395 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10396 (struct machine_function): New field interrupt_mode.
10397 (riscv_handle_type_attribute): New function. Add forward declaration.
10398 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10399 (riscv_expand_epilogue): Check interrupt_mode field.
10400 (riscv_set_current_function): Check interrupt attribute args and
10401 set interrupt_mode field.
10402 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10403 (riscv_sret, riscv_uret): New.
10404 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10405 new arguments to interrupt attribute.
10407 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10410 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10411 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10413 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10415 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10416 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10417 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10418 several redundant entries.
10420 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10422 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10423 type from "rtx" to "rtx_insn *".
10424 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10425 for local "call_insn", removing cast.
10426 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10428 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10430 PR tree-optimization/86066
10431 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10432 for BIT_INSERT_EXPR stores.
10434 2018-06-06 Richard Biener <rguenther@suse.de>
10436 PR tree-optimization/86062
10437 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10438 component refs ontop
10439 of to be offsetted base.
10441 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10443 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10444 to be static and remove check on interrupt attribute name.
10446 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10448 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10449 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10451 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10454 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10456 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10457 Remove second argument, change how error is called.
10458 (aarch64_layout_arg): Remove second argument from
10459 aarch64_err_no_fpadvsimd call.
10460 (aarch64_init_cumulative_args): Ditto.
10461 (aarch64_gimplify_va_arg_expr): Ditto.
10462 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10464 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10466 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10467 (*simple_return_indirect_internal<mode>): Rename from
10468 simple_return_indirect_internal. Use W mode iterator.
10469 (rstorssp): New expander.
10470 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10471 (clrssbsy): New expander.
10472 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10474 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10476 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10478 (cmse_check_pointed_object): Likewise.
10480 2018-06-05 Martin Liska <mliska@suse.cz>
10482 PR gcov-profile/47618
10483 * doc/invoke.texi: Document how -fprofile-dir format
10486 2018-06-05 Richard Biener <rguenther@suse.de>
10488 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10489 removal pretend DOM info isn't available so we do not update
10490 it and only remove edges, not dominated blocks. Actually free
10491 DOM info in case we removed something. Remove unreachable blocks.
10492 (mfb_keep_latches): Work with either DOM info or marked backedges.
10493 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10494 first. Mark backedges if DOM info isn't available.
10495 (Re-)compute DOM info after cleanup_control_flow_pre.
10497 2018-06-05 Richard Biener <rguenther@suse.de>
10499 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10500 (locus_discrim_hasher::hash): Adjust.
10501 (locus_discrim_hasher::equal): Likewise.
10502 (next_discriminator_for_locus): Work on line directly.
10503 (same_line_p): Pass in expanded locus1 as well.
10504 (assign_discriminators): Avoid redundant location expansions.
10506 2018-06-05 Richard Biener <rguenther@suse.de>
10508 PR tree-optimization/86046
10509 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10510 if required after clearing TREE_ADDRESSABLE.
10512 2018-06-05 Richard Biener <rguenther@suse.de>
10514 PR tree-optimization/86047
10515 * tree-ssa-loop.c (for_each_index): Glob handling of all
10516 decls and constants and really handle all of them.
10518 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10521 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10522 qualifier_void_pointer and qualifier_const_void_pointer.
10523 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10524 (arm_init_builtins): Handle the above.
10525 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10526 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10527 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10530 2018-06-05 Martin Liska <mliska@suse.cz>
10532 * auto-profile.c (read_autofdo_file): Do not use
10533 gcov_ctr_summary struct.
10534 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10535 * coverage.c (struct counts_entry): Likewise.
10536 (read_counts_file): Read just single summary entry.
10537 (get_coverage_counts): Use gcov_summary struct.
10538 * coverage.h (get_coverage_counts): Likewise.
10539 * gcov-dump.c (dump_working_sets): Likewise.
10540 (tag_summary): Dump just single summary.
10541 * gcov-io.c (gcov_write_summary): Write just histogram
10543 (gcov_read_summary): Read just single summary.
10544 (compute_working_sets): Use gcov_summary struct.
10545 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10546 of GCOV_COUNTERS_SUMMABLE.
10547 (GCOV_COUNTERS_SUMMABLE): Remove.
10548 (GCOV_FIRST_VALUE_COUNTER): Replace with
10549 GCOV_COUNTER_V_INTERVAL.
10550 (struct gcov_ctr_summary): Remove.
10551 (struct gcov_summary): Directly use fields of former
10553 (compute_working_sets): Use gcov_summary struct.
10554 * gcov.c (read_count_file): Do not use ctrs fields.
10555 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10557 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10559 * profile.c: Likewise.
10560 * profile.h: Likewise.
10562 2018-06-05 Martin Liska <mliska@suse.cz>
10564 PR gcov-profile/84846
10565 * gcov.c (output_lines): Print working directory only
10566 in intermediate format.
10568 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10570 * config/s390/s390-builtin-types.def: Add void function type.
10571 * config/s390/s390-builtins.def: Use the function type for the
10574 2018-06-04 Jim Wilson <jimw@sifive.com>
10576 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10578 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10579 and maybe_eh_return. Change regno to unsigned int. Use new args to
10580 handle EH_RETURN_DATA_REGNO registers properly.
10581 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10582 (riscv_expand_epilogue): Update comment. Change argument name and
10583 type. Update code to use new name and type. Pass new args to
10584 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10586 * config/riscv/riscv.md (NORMAL_RETURN): New.
10587 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10588 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10589 (eh_return): Call gen_eh_return_internal and emit barrier.
10590 (eh_return_internal): Call riscv_expand_epilogue.
10592 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10594 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10595 bit_insertion field and declare can_be_merged_into method.
10596 (merged_store_group::can_be_merged_into): New method.
10597 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10598 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10599 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10601 2018-06-04 Richard Biener <rguenther@suse.de>
10603 PR tree-optimization/85955
10604 * builtins.c (fold_builtin_sincos): Convert pointers to
10605 destination to appropriate type before dereferencing.
10607 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10609 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10611 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10613 * expr.c (expand_expr_real_1): Force the operand into memory if
10614 its TYPE_MODE is BLKmode and if there is no integer mode for
10615 the number of bits being extracted.
10617 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10621 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10622 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10624 2018-06-04 Richard Biener <rguenther@suse.de>
10626 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10627 (cleanup_tree_cfg_noloop): ... single caller. Do
10628 start_recording_case_labels later.
10630 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10632 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10633 to _IMMINTRIN_H_INCLUDED.
10634 * config/i386/pconfigintrin.h: Ditto.
10635 * config/i386/waitpkgintrin.h: Ditto.
10636 * config/i386/immintrin.h: Add includes for sgxintrin.h,
10637 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10638 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10639 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10640 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10641 waitpkgintrin.h and cldemoteintrin.h.
10643 2018-06-04 Richard Biener <rguenther@suse.de>
10645 PR tree-optimization/86038
10646 * tracer.c (find_best_successor): Check probability for
10647 being initialized, bail out if not.
10649 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
10652 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10653 of bits to ignore when comparing architectures.
10655 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10657 PR tree-optimization/69615
10658 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10659 maximum or minimum of the type, try to merge it also as if
10660 range1 is + [-, x - 1] or + [x + 1, -].
10663 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10665 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
10667 PR tree-optimization/86034
10668 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10669 the unsigned bitfield type in a bit insertion sequence if it does not
10670 have a larger precision than the bitfield size.
10671 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
10673 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
10675 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
10677 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10679 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
10680 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
10681 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10682 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
10684 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10686 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
10687 Disable -fdelete-null-pointer-checks for ELF toolchain.
10689 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10690 Kito Cheng <kito.cheng@gmail.com>
10692 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
10693 (nds32le-*-*, nds32be-*-*): Integrate checking process.
10694 (nds32*-*-*): Add glibc and uclibc conditions.
10695 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
10696 (TARGET_EXCEPT_UNWIND_INFO): Define.
10697 * config/nds32/elf.h: New file.
10698 * config/nds32/linux.h: New file.
10699 * config/nds32/nds32-elf.opt: New file.
10700 * config/nds32/nds32-linux.opt: New file.
10701 * config/nds32/nds32-fp-as-gp.c
10702 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
10703 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
10705 (nds32_asm_file_end): Ditto.
10706 (nds32_print_operand): Ditto.
10707 (nds32_insert_attributes): Ditto.
10708 (nds32_init_libfuncs): New function.
10709 (TARGET_HAVE_TLS): Define.
10710 (TARGET_INIT_LIBFUNCS): Define.
10711 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
10713 (TARGET_ELF): Apply different mcmodel setting.
10714 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
10715 been migrated into elf.h and linux.h files.
10716 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
10717 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
10718 (mcmodel): The content has been migrated into nds32-elf.opt and
10719 nds32-linux.opt files.
10720 * config/nds32/t-elf: New file.
10721 * config/nds32/t-linux: New file.
10723 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10724 Shiva Chen <shiva0217@gmail.com>
10726 * config/nds32/constants.md (unspec_volatile_element): Add
10727 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
10728 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
10730 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
10731 (make_pass_nds32_fp_as_gp): Declare.
10732 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
10734 (nds32_asm_function_end_prologue): Remove unused asm output.
10735 (nds32_asm_function_begin_epilogue): Remove unused asm output.
10736 (nds32_asm_file_start): Output necessary fp_as_gp information.
10737 (nds32_option_override): Adjust register usage.
10738 (nds32_expand_prologue): Consider fp_as_gp situation.
10739 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
10740 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
10743 (simple_return): Ditto.
10744 (omit_fp_begin): Output special directive for fp_as_gp.
10745 (omit_fp_end): Output special directive for fp_as_gp.
10746 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
10747 mforbid-fp-as-gp): New options.
10749 2018-06-01 Mark Wielaard <mark@klomp.org>
10751 * dwarf2out.c (dwarf2out_finish): Remove generation of
10752 DW_AT_loclists_base.
10754 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
10756 * gimple-ssa-store-merging.c: Include gimple-fold.h.
10757 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
10758 (struct merged_store_group): Add bit_insertion field.
10759 (dump_char_array): Use standard hexadecimal format.
10760 (merged_store_group::merged_store_group): Set bit_insertion to false.
10761 (merged_store_group::apply_stores): Use optimal buffer size. Deal
10762 with BIT_INSERT_EXPR stores. Move up code updating the mask and
10763 also print the mask in the dump file.
10764 (pass_store_merging::gate): Minor tweak.
10765 (imm_store_chain_info::coalesce_immediate): Fix wrong association
10766 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
10767 stores with INTEGER_CST stores.
10768 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
10769 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
10770 and use it throughout. Generate bit insertion sequences if need be.
10771 (pass_store_merging::process_store): Remove redundant condition.
10772 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
10774 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
10776 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
10777 the 128-bit floating point types. Fix function comment.
10779 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10781 * config/aarch64/aarch64-simd.md
10782 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10784 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10787 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10789 PR tree-optimization/85989
10790 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10792 (backprop::intersect_uses): Check it when deciding whether this
10793 is a backedge reference.
10794 (backprop::process_block): Add each phi to m_visited_phis
10795 after visiting it, then clear it at the end.
10797 2018-06-01 Richard Biener <rguenther@suse.de>
10799 * tree-vectorizer.h (vect_dr_stmt): New function.
10800 (vect_get_load_cost): Adjust.
10801 (vect_get_store_cost): Likewise.
10802 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10803 Use vect_dr_stmt instead of DR_SMTT.
10804 (vect_record_base_alignments): Likewise.
10805 (vect_calculate_target_alignment): Likewise.
10806 (vect_compute_data_ref_alignment): Likewise and make static.
10807 (vect_update_misalignment_for_peel): Likewise.
10808 (vect_verify_datarefs_alignment): Likewise.
10809 (vector_alignment_reachable_p): Likewise.
10810 (vect_get_data_access_cost): Likewise. Pass down
10811 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10812 (vect_get_peeling_costs_all_drs): Likewise.
10813 (vect_peeling_hash_get_lowest_cost): Likewise.
10814 (vect_enhance_data_refs_alignment): Likewise.
10815 (vect_find_same_alignment_drs): Likewise.
10816 (vect_analyze_data_refs_alignment): Likewise.
10817 (vect_analyze_group_access_1): Likewise.
10818 (vect_analyze_group_access): Likewise.
10819 (vect_analyze_data_ref_access): Likewise.
10820 (vect_analyze_data_ref_accesses): Likewise.
10821 (vect_vfa_segment_size): Likewise.
10822 (vect_small_gap_p): Likewise.
10823 (vectorizable_with_step_bound_p): Likewise.
10824 (vect_prune_runtime_alias_test_list): Likewise.
10825 (vect_analyze_data_refs): Likewise.
10826 (vect_supportable_dr_alignment): Likewise.
10827 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10828 (vect_gen_prolog_loop_niters): Likewise.
10829 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10830 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10832 (vect_recog_mask_conversion_pattern): Likewise.
10833 (vect_try_gather_scatter_pattern): Likewise.
10834 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10835 to vect_get_store_cost.
10836 (vect_get_store_cost): Get stmt_info instead of DR.
10837 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10838 (vect_get_load_cost): Get stmt_info instead of DR.
10840 2018-06-01 Richard Biener <rguenther@suse.de>
10842 PR middle-end/86017
10843 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10844 vars wrapped in MEM_REFs.
10846 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10848 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10849 Fix subreg tests so that we only return a choice between
10850 GENERAL_REGS and FP_REGS if the original classes included both.
10852 2018-06-01 Richard Biener <rguenther@suse.de>
10855 * tree-ssa-structalias.c (get_function_part_constraint):
10856 Handle NULL fi->decl.
10857 (find_func_aliases_for_call): Properly handle indirect
10858 fi from direct call.
10859 (find_func_clobbers): Likewise.
10860 (ipa_pta_execute): Likewise.
10861 (create_variable_info_for): For functions that are ifunc_resolver
10862 resolve to a varinfo that contains the result of the resolver call.
10863 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10866 2018-05-31 Michael Collison <michael.collison@arm.com>
10868 * config/aarch64/aarch64.md:
10869 (*fix_to_zero_extenddfdi2): New pattern.
10870 * gcc.target/aarch64/fix_extend1.c: New testcase.
10872 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10874 PR middle-end/78809
10875 PR middle-end/83026
10876 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10877 and BUILT_IN_STRNCMP_EQ.
10878 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10879 BUILT_IN_STRNCMP_EQ.
10880 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10881 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10882 (gimple_fold_builtin): Likewise.
10883 * tree-ssa-strlen.c (compute_string_length): New function.
10884 (determine_min_obsize): New function.
10885 (handle_builtin_string_cmp): New function to handle calls to
10886 string compare functions.
10887 (strlen_optimize_stmt): Add handling to builtin string compare
10889 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10890 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10891 * tree.c (build_common_builtin_nodes): Add new defines of
10892 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10894 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10897 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10898 functions with naked attribute.
10900 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10902 * config/i386/sse.md (avx_vec_concat<mode>):
10903 Substitute concat_tg_mode mode attribute with xtg_mode.
10904 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10905 (concat_tg_mode): Remove mode attribute.
10907 2018-05-31 Martin Sebor <msebor@redhat.com>
10910 * calls.c (alloc_max_size): Correct a logic error/typo.
10911 Treat excessive arguments as infinite. Warn for invalid arguments.
10912 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10914 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10917 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10918 and movx for Haswell.
10920 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10921 Cesar Philippidis <cesar@codesourcery.com>
10923 PR middle-end/85879
10924 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10925 when emitting error on private/firstprivate reductions.
10926 * omp-low.c (lower_omp_target): Avoid reference-type processing
10927 on pointers for firstprivate clause.
10929 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10931 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10934 * config/aarch64/aarch64-simd.md
10935 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10936 (aarch64_ld1_x3_<mode>): Likewise
10937 (aarch64_st1x2<VALLDIF:mode>): Likewise
10938 (aarch64_st1_x2_<mode>): Likewise
10939 (aarch64_st1x3<VALLDIF:mode>): Likewise
10940 (aarch64_st1_x3_<mode>): Likewise
10941 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10942 (vld1_s8_x3): Likewise.
10943 (vld1_u16_x3): Likewise.
10944 (vld1_s16_x3): Likewise.
10945 (vld1_u32_x3): Likewise.
10946 (vld1_s32_x3): Likewise.
10947 (vld1_u64_x3): Likewise.
10948 (vld1_s64_x3): Likewise.
10949 (vld1_f16_x3): Likewise.
10950 (vld1_f32_x3): Likewise.
10951 (vld1_f64_x3): Likewise.
10952 (vld1_p8_x3): Likewise.
10953 (vld1_p16_x3): Likewise.
10954 (vld1_p64_x3): Likewise.
10955 (vld1q_u8_x3): Likewise.
10956 (vld1q_s8_x3): Likewise.
10957 (vld1q_u16_x3): Likewise.
10958 (vld1q_s16_x3): Likewise.
10959 (vld1q_u32_x3): Likewise.
10960 (vld1q_s32_x3): Likewise.
10961 (vld1q_u64_x3): Likewise.
10962 (vld1q_s64_x3): Likewise.
10963 (vld1q_f16_x3): Likewise.
10964 (vld1q_f32_x3): Likewise.
10965 (vld1q_f64_x3): Likewise.
10966 (vld1q_p8_x3): Likewise.
10967 (vld1q_p16_x3): Likewise.
10968 (vld1q_p64_x3): Likewise.
10969 (vst1_s64_x2): Likewise.
10970 (vst1_u64_x2): Likewise.
10971 (vst1_f64_x2): Likewise.
10972 (vst1_s8_x2): Likewise.
10973 (vst1_p8_x2): Likewise.
10974 (vst1_s16_x2): Likewise.
10975 (vst1_p16_x2): Likewise.
10976 (vst1_s32_x2): Likewise.
10977 (vst1_u8_x2): Likewise.
10978 (vst1_u16_x2): Likewise.
10979 (vst1_u32_x2): Likewise.
10980 (vst1_f16_x2): Likewise.
10981 (vst1_f32_x2): Likewise.
10982 (vst1_p64_x2): Likewise.
10983 (vst1q_s8_x2): Likewise.
10984 (vst1q_p8_x2): Likewise.
10985 (vst1q_s16_x2): Likewise.
10986 (vst1q_p16_x2): Likewise.
10987 (vst1q_s32_x2): Likewise.
10988 (vst1q_s64_x2): Likewise.
10989 (vst1q_u8_x2): Likewise.
10990 (vst1q_u16_x2): Likewise.
10991 (vst1q_u32_x2): Likewise.
10992 (vst1q_u64_x2): Likewise.
10993 (vst1q_f16_x2): Likewise.
10994 (vst1q_f32_x2): Likewise.
10995 (vst1q_f64_x2): Likewise.
10996 (vst1q_p64_x2): Likewise.
10997 (vst1_s64_x3): Likewise.
10998 (vst1_u64_x3): Likewise.
10999 (vst1_f64_x3): Likewise.
11000 (vst1_s8_x3): Likewise.
11001 (vst1_p8_x3): Likewise.
11002 (vst1_s16_x3): Likewise.
11003 (vst1_p16_x3): Likewise.
11004 (vst1_s32_x3): Likewise.
11005 (vst1_u8_x3): Likewise.
11006 (vst1_u16_x3): Likewise.
11007 (vst1_u32_x3): Likewise.
11008 (vst1_f16_x3): Likewise.
11009 (vst1_f32_x3): Likewise.
11010 (vst1_p64_x3): Likewise.
11011 (vst1q_s8_x3): Likewise.
11012 (vst1q_p8_x3): Likewise.
11013 (vst1q_s16_x3): Likewise.
11014 (vst1q_p16_x3): Likewise.
11015 (vst1q_s32_x3): Likewise.
11016 (vst1q_s64_x3): Likewise.
11017 (vst1q_u8_x3): Likewise.
11018 (vst1q_u16_x3): Likewise.
11019 (vst1q_u32_x3): Likewise.
11020 (vst1q_u64_x3): Likewise.
11021 (vst1q_f16_x3): Likewise.
11022 (vst1q_f32_x3): Likewise.
11023 (vst1q_f64_x3): Likewise.
11024 (vst1q_p64_x3): Likewise.
11026 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11028 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11029 user_label_prefix to name.
11031 * tree-core.h: Update comment about the format of NAME string
11032 passed to handler in attribute_spec.
11034 * config/msp430/msp430.md: Remove erroneous subreg expression from
11035 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11036 zero_extend{q,h}isi2.
11038 2018-05-30 Borislav Petkov <bp@suse.de>
11040 * doc/extend.texi: Document some architecture specific
11041 constraints and sort entries.
11043 2018-05-30 Martin Sebor <msebor@redhat.com>
11045 PR middle-end/85369
11046 * builtins.c (expand_builtin_stpcpy_1): New function.
11047 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11048 only if the former succeeds.
11050 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11052 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11055 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11057 * doc/invoke.texi (-flinker-output): Document
11059 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11061 * passes.c (ipa_write_summaries): Only modify statements if body
11063 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11064 incrementally linking.
11065 (ipa_passes): Likewise.
11066 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11067 pass down resolution info.
11068 * common.opt (flag_incremental_link): Update info.
11069 * gcc.c (plugin specs): Turn flinker-output=* to
11070 -plugin-opt=-linker-output-known
11071 * toplev.c (compile_file): Also cut compilation when doing incremental
11073 * flag-types. (enum lto_partition_model): Add
11074 LTO_LINKER_OUTPUT_NOLTOREL.
11075 (invoke.texi): Add -flinker-output docs.
11076 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11077 link same way as WPA; do not stream in dead initializers.
11079 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11080 darf2out_register_external_decl): Support incremental link.
11082 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11084 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11086 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11088 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11089 it down to simple_object_copy_lto_debug_sections.
11090 (run_gcc): Determine incremental LTO link time and configure
11091 lto1 into non-wpa mode, disable renaming of debug sections.
11093 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11095 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11096 descriptions of various incorrectly documented functions.
11098 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11101 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11102 address check not strict.
11104 2018-05-30 Richard Biener <rguenther@suse.de>
11106 PR tree-optimization/85964
11107 * tracer.c (better_p): Drop initialized count check, we only
11108 call the function with initialized counts now.
11109 (find_best_successor): Do find a best edge if one
11110 has uninitialized count.
11111 (find_best_predecessor): Likewise. Do BB frequency check only
11112 if count is initialized.
11114 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11116 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11117 (aarch64_ldrstr_offset_compare): New.
11118 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11119 load/store orderings.
11120 (aarch64_gen_adjusted_ldpstp): Likewise.
11122 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11124 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11125 Check for subset of GENERAL_REGS and FP_REGS.
11126 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11129 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11131 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11132 and wi::to_poly_offset. Add the current offset and then check
11133 whether the sum fits, rather than using an unchecked addition of
11134 a checked term. Check for a shwi rather than a uhwi.
11135 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11136 (store_constructor): Use poly_int_tree_p.
11137 (expand_expr_real_1): Likewise.
11138 * function.c (assign_temp): Likewise.
11139 * fold-const.c (const_binop): Use poly_int_tree_p and
11140 wi::to_poly_offset.
11141 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11143 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11144 to_poly_offset for MEM offsets.
11145 * ipa-icf.c (sem_variable::equals): Likewise.
11146 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11147 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11148 wi::to_poly_offset for BIT_FIELD_REF offsets.
11149 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11150 wi::to_poly_offset.
11151 * var-tracking.c (emit_note_insn_var_location): Use
11152 tree_to_poly_uint64.
11154 2018-05-29 Jim Wilson <jimw@sifive.com>
11156 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11158 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11161 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11162 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11164 (sse4_1_round<mode>2): Use nonimmediate_operand
11165 for operand 1 predicate.
11167 2018-05-29 Martin Sebor <msebor@redhat.com>
11168 Richard Biener <rguenther@suse.de>
11171 * calls.c (get_size_range): Call determine_value_range instead
11172 of get_value_range..
11173 * tree-vrp.h (determine_value_range): Declared new function.
11174 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11176 2018-05-29 Richard Biener <rguenther@suse.de>
11178 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11179 sure to use non-pattern stmts for get_earlier_stmt arguments.
11180 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11181 called on pattern stmts.
11182 (get_later_stmt): Likewise.
11184 2018-05-29 Martin Liska <mliska@suse.cz>
11186 PR gcov-profile/85759
11187 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11190 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11192 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11194 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11198 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11199 VEC_PACK_FLOAT_EXPR): New tree codes.
11200 * tree-pretty-print.c (op_code_prio): Handle
11201 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11202 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11203 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11204 * tree-inline.c (estimate_operator_cost): Likewise.
11205 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11206 * fold-const.c (const_binop): Likewise.
11207 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11208 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11209 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11210 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11211 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11212 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11213 * expr.c (expand_expr_real_2): Likewise.
11214 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11215 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11216 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11218 * optabs.c (expand_widen_pattern_expr): For
11219 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11220 sign from result type rather than operand's type.
11221 (expand_binop_directly): For vec_packu_float_optab and
11222 vec_packs_float_optab allow result type to be different from operand's
11224 * optabs-tree.c (optab_for_tree_code): Handle
11225 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11226 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11227 * tree-vect-generic.c (expand_vector_operations_1): Handle
11228 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11229 VEC_PACK_FLOAT_EXPR.
11230 * tree-vect-stmts.c (supportable_widening_operation): Handle
11232 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11233 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11234 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11235 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11236 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11238 (vec_pack<floatprefix>_float_<mode>): New expander.
11239 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11241 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11242 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11243 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11244 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11245 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11247 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11248 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11249 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11250 VEC_PACK_FLOAT_EXPR): Document.
11252 2018-05-29 Richard Biener <rguenther@suse.de>
11254 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11256 (stmt_vec_info_vec): Make pointer.
11257 (init_stmt_vec_info_vec): Remove.
11258 (free_stmt_vec_info_vec): Likewise.
11259 (set_stmt_vec_info_vec): New function.
11260 (free_stmt_vec_infos): Likewise.
11261 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11262 (set_vinfo_for_stmt): Likewise.
11263 (get_earlier_stmt): Likewise.
11264 (get_later_stmt): Likewise.
11265 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11266 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11267 (vec_info::~vec_info): Free stmt_vec_infos.
11268 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11269 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11270 (pass_slp_vectorize::execute): Likewise.
11271 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11272 (free_stmt_vec_info_vec): Likewise.
11273 (set_stmt_vec_info_vec): New function.
11274 (free_stmt_vec_infos): Likewise.
11275 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11276 the global stmt_vec_info_vec.
11277 * tree-parloops.c (gather_scalar_reductions): Use
11278 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11281 2018-05-29 Richard Biener <rguenther@suse.de>
11283 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11285 2018-05-29 Martin Liska <mliska@suse.cz>
11286 David Malcolm <dmalcolm@redhat.com>
11288 * vec.c (test_reverse): New.
11289 (vec_c_tests): Add new test.
11290 * vec.h (vl_ptr>::reverse): New function.
11292 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11294 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11296 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11299 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11301 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11303 2018-05-28 Richard Biener <rguenther@suse.de>
11305 PR tree-optimization/85933
11306 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11307 look at stmts marked as vectorizable.
11309 2018-05-28 Richard Biener <rguenther@suse.de>
11311 PR tree-optimization/85934
11312 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11313 vector boolean check before scalar optimization.
11315 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11317 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11320 2018-05-28 Mark Wielaard <mark@klomp.org>
11322 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11323 if it is an expression containing a minus sign.
11325 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11327 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11329 2018-05-27 Paul Koning <ni1d@arrl.net>
11331 * config/pdp11/pdp11.md (truncsihi2): Remove.
11333 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11334 Chung-Ju Wu <jasonwucj@gmail.com>
11336 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11338 (unaligned_store_dw): Ditto.
11339 * config/nds32/nds32-memory-manipulation.c
11340 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11341 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11342 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11343 (emit_setmem_word_loop): Rename to ...
11344 (emit_setmem_doubleword_loop): ... this.
11345 (nds32_gen_dup_4_byte_to_word_value): New function.
11346 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11347 (nds32_expand_setmem_loop): Refine implementation.
11348 (nds32_expand_setmem_loop_v3m): Ditto.
11349 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11352 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11354 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11356 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11358 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11359 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11360 machine->attr_no_prologue_p.
11361 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11362 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11363 (nds32_expand_epilogue): Consider attr_naked_p.
11364 (nds32_expand_epilogue_v3pop): Likewise.
11365 (nds32_can_use_return_insn): Likewise.
11366 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11367 attr_no_prologue_p fields.
11368 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11370 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11373 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11375 * config/i386/sse.md
11376 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11378 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11380 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11382 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11384 (*<floatsuffix>floatv2div2sf2): Rename to ...
11385 (*float<floatunssuffix>v2div2sf2): ... this.
11386 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11387 (float<floatunssuffix>v2div2sf2_mask): ... this.
11388 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11389 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11390 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11392 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11394 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11396 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11398 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11400 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11402 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11403 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11404 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11405 gen_ufix_truncv8dfv8si2.
11406 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11407 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11408 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11409 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11410 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11411 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11412 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11413 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11415 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11419 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11421 2018-05-25 Jim Wilson <jimw@sifive.com>
11423 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11424 * config/riscv/riscv.c (struct machine_function): Add
11425 interrupt_handler_p and attribute_checked_p fields.
11426 (riscv_attribute_table): Add interrupt.
11427 (riscv_interrupt_type_p): New.
11428 (riscv_save_reg_p): Save extra regs for interrupt handler.
11429 (riscv_use_save_libcall): Return false for interrupt handler.
11430 (riscv_first_stack_step): Add forward declaration.
11431 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11432 for interrupt handler with large frame. Use it for saved reg list.
11433 (riscv_expand_prologue): Move flag_stack_usage_info support to
11434 eliminate duplication.
11435 (riscv_expand_epilogue): Generate mret for interrupt handler.
11436 (riscv_epilogue_uses): New.
11437 (riscv_can_use_return_insn): Return false for interrupt handler.
11438 (riscv_function_ok_for_sibcall): Likewise.
11439 (riscv_set_current_function): Add interrupt handler support.
11440 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11441 * config/riscv/riscv.md (UNSPECV_MRET): New.
11443 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11445 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11447 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11449 PR tree-optimization/85712
11450 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11451 this candidate has already been replaced in-situ by a copy.
11453 2018-05-25 Jason Merrill <jason@redhat.com>
11455 PR c++/80485 - inline function non-zero address.
11456 * symtab.c (nonzero_address): Check DECL_COMDAT.
11458 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11461 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11462 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11463 extension of SImode operation. Use const123_operand predicate.
11464 (*saddsi_1): Remove.
11465 (*saddl_se_1): Ditto.
11466 (*ssubsi_1): Ditto.
11467 (*ssubl_se_1): Ditto.
11468 * config/alpha/predicates.md (const123_operand): New predicate.
11469 * config/alpha/constraints.md (P): Use IN_RANGE.
11471 2018-05-25 Richard Biener <rguenther@suse.de>
11473 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11475 (ref_maybe_used_by_stmt_p): Likewise.
11476 (stmt_may_clobber_ref_p): Likewise.
11477 (stmt_may_clobber_ref_p_1): Likewise.
11478 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11480 (ref_maybe_used_by_stmt_p): Likewise.
11481 (stmt_may_clobber_ref_p): Likewise.
11482 (stmt_may_clobber_ref_p_1): Likewise.
11483 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11484 the alias oracle to disambiguate DRs with stmts DR analysis
11486 (vect_analyze_data_refs): Do not give up on not analyzable
11487 DRs for BB vectorization. Remove code truncating the dataref
11490 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11493 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11494 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11495 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11497 2018-05-25 Richard Biener <rguenther@suse.de>
11499 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11500 function, combining stmt data ref gathering and fatal analysis
11502 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11503 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11504 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11505 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11506 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11508 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11510 PR tree-optimization/85720
11511 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11512 SCC if all partitions are builtins.
11513 (version_loop_by_alias_check): New parameter. Generate cancelable
11514 runtime alias check if all partitions are builtins.
11515 (distribute_loop): Update call to above function.
11517 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11519 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11520 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11521 (parm_default_def_partition_arg): Ditto.
11522 (set_parm_default_def_partition): Ditto.
11523 (get_parm_default_def_partitions): Ditto and make it static.
11524 (get_undefined_value_partitions): Ditto and make it static.
11525 (remove_ssa_form): Refactor call to init_var_map here.
11526 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11527 computation for loop region.
11528 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11529 (register_default_def): Delete.
11530 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11531 (parm_default_def_partition_arg): Ditto.
11532 (set_parm_default_def_partition): Ditto.
11533 (get_parm_default_def_partitions): Ditto and make it static.
11534 (get_undefined_value_partitions): Ditto and make it static.
11535 (coalesce_with_default, coalesce_with_default): Update comment.
11536 (create_coalesce_list_for_region): New func factored out from
11537 create_outofssa_var_map.
11538 (populate_coalesce_list_for_outofssa): New func factored out from
11539 create_outofssa_var_map and coalesce_ssa_name.
11540 (create_outofssa_var_map): Delete.
11541 (coalesce_ssa_name): Refactor to support live range computation.
11542 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11543 (get_parm_default_def_partitions): Delete.
11544 (get_undefined_value_partitions): Ditto.
11545 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11546 computation for loop region.
11547 (new_tree_live_info, loe_visit_block): Ditto.
11548 (live_worklist, set_var_live_on_entry): Ditto.
11549 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11550 * tree-ssa-live.h (struct _var_map): New fields.
11551 (init_var_map): Change decl.
11552 (region_contains_p): New.
11554 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11556 * tree-ssa-live.h (live_merge_and_clear): Delete.
11558 2018-05-25 Richard Biener <rguenther@suse.de>
11561 * tree-dump.c (dequeue_and_dump): Remove access to removed
11562 operand 2 of a SWITCH_EXPR.
11564 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11566 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11567 multiplication and division.
11568 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11569 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11570 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11571 (cond_udiv_optab, cond_umod_optab): New optabs.
11572 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11573 (IFN_COND_RDIV): New internal functions.
11574 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11575 TRUNC_MOD_EXPR and RDIV_EXPR.
11576 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11577 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11579 (SVE_INT_BINARY): Include mult.
11580 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11581 (optab, sve_int_op): Handle mult.
11582 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11584 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11585 for SVE_INT_BINARY_SD.
11587 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11589 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11590 (optab, sve_int_op): Handle div and udiv.
11591 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11592 for SVE_INT_BINARY_SD.
11593 (*<optab><mode>3): New insn for the same.
11595 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11597 * tree-vect-patterns.c: Include predict.h.
11598 (vect_recog_divmod_pattern): Restrict check for division support
11599 to when optimizing for size.
11601 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11603 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11604 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11605 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11606 (gimple_match_op::set_op): Likewise.
11607 (gimple_resimplify4): Declare.
11608 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11609 (expr::gen_transform): Likewise.
11610 (decision_tree::gen): Generate a simplification routine for 4 operands.
11611 * gimple-match-head.c (gimple_simplify): Add an overload for
11612 4 operands. In the top-level function, handle up to 4 call
11613 arguments and call gimple_resimplify4.
11614 (gimple_resimplify4): New function.
11615 (build_call_internal): Pass a fourth operand.
11616 (maybe_push_to_seq): Likewise.
11617 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11618 Fold VEC_COND_EXPRs of an operation and a default value into
11619 an IFN_COND_* function if possible.
11620 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11622 (SVE_COND_FP_BINARY): Include them.
11623 (optab, sve_fp_op): Handle them.
11624 (SVE_INT_BINARY_REV): New code iterator.
11625 (SVE_COND_FP_BINARY_REV): New int iterator.
11626 (commutative): New int attribute.
11627 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11629 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11631 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11632 (*cond_<optab><mode>): New patterns for reversed operands.
11634 2018-05-25 Richard Biener <rguenther@suse.de>
11636 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11637 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11638 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11639 (STMT_VINFO_GROUPED_ACCESS): Adjust.
11640 * tree-vect-data-refs.c (everywhere): Adjust users.
11641 * tree-vect-loop.c (everywhere): Likewise.
11642 * tree-vect-slp.c (everywhere): Likewise.
11643 * tree-vect-stmts.c (everywhere): Likewise.
11644 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11646 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11648 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11650 (gcc_cv_as_section_exclude): ... this.
11651 Try Solaris as #exclude syntax.
11652 * configure: Regenerate.
11653 * config.in: Regenerate.
11654 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11656 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11657 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11659 * varasm.c (default_elf_asm_named_section): Don't check if
11660 HAVE_GAS_SECTION_EXCLUDE is defined.
11662 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11664 * doc/md.texi: Update the documentation of the cond_* optabs
11665 to mention the new final operand. Fix GET_MODE_NUNITS call.
11666 Describe the scalar case too.
11667 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
11668 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
11670 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
11671 (get_conditional_internal_fn): Update comment.
11672 * tree-vect-loop.c (vectorizable_reduction): Pass the original
11673 accumulator value as a final argument to conditional functions.
11674 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
11675 a define_expand and add an "else" operand. Assert for now that
11676 the else operand is equal to operand 2. Use SVE_INT_BINARY and
11677 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
11678 (*cond_<optab><mode>): New patterns.
11679 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
11680 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11681 (UNSPEC_COND_EOR): Delete.
11682 (optab): Remove associated mappings.
11683 (SVE_INT_BINARY): New code iterator.
11684 (sve_int_op): Remove int attribute and add "minus" to the code
11686 (SVE_COND_INT_OP): Delete.
11687 (SVE_COND_FP_OP): Rename to...
11688 (SVE_COND_FP_BINARY): ...this.
11690 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11692 * optabs.c (can_reuse_operands_p): New function.
11693 (maybe_legitimize_operands): Try to reuse the results for
11696 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11698 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
11699 Add {q} suffix to insn mnemonic.
11701 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11703 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
11704 (msp430_warn_func_return): New.
11706 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
11708 * fold-const.c (tree_nonzero_bits): New function.
11709 * fold-const.h (tree_nonzero_bits): Likewise.
11710 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
11711 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
11713 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11717 * varasm.c (assemble_alias): Check ifunc_resolver only on
11720 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11723 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
11724 when memory input operand is handled.
11726 2018-05-24 Luis Machado <luis.machado@linaro.org>
11728 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
11730 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
11732 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11734 * match.pd: Delay FMA folds until after vectorization.
11736 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
11739 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11740 address check not strict.
11742 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11744 * gimple-match.h (gimple_match_op): New class.
11745 (mprts_hook): Replace parameters with a gimple_match_op *.
11746 (maybe_build_generic_op): Likewise.
11747 (gimple_simplified_result_is_gimple_val): Replace parameters with
11748 a const gimple_match_op *.
11749 (gimple_simplify): Replace code_helper * and tree * parameters with
11750 a gimple_match_op * parameter.
11751 (gimple_resimplify1): Replace code_helper *, tree and tree *
11752 parameters with a gimple_match_op * parameter.
11753 (gimple_resimplify2): Likewise.
11754 (gimple_resimplify3): Likewise.
11755 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
11756 parameters with a gimple_match_op * parameter.
11757 * gimple-match-head.c (gimple_simplify): Change prototypes of
11758 auto-generated functions to take a gimple_match_op * instead of
11759 separate code_helper * and tree * parameters. Make the same
11760 change in the top-level overload and update calls to the
11761 gimple_resimplify routines. Update calls to the auto-generated
11762 functions and to maybe_push_res_to_seq in the publicly-facing
11763 operation-specific gimple_simplify overloads.
11764 (gimple_match_op::MAX_NUM_OPS): Define.
11765 (gimple_resimplify1): Replace rcode and ops with a single res_op
11766 parameter. Update call to gimple_simplify.
11767 (gimple_resimplify2): Likewise.
11768 (gimple_resimplify3): Likewise.
11769 (mprts_hook): Replace parameters with a gimple_match_op *.
11770 (maybe_build_generic_op): Likewise.
11771 (build_call_internal): Replace type, nargs and ops with
11772 a gimple_match_op *.
11773 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
11774 with a single gimple_match_op *. Update calls to mprts_hook,
11775 build_call_internal and gimple_simplified_result_is_gimple_val.
11776 Factor out code that is common to the tree_code and combined_fn cases.
11777 * genmatch.c (expr::gen_transform): Replace tem_code and
11778 tem_ops with a gimple_match_op called tem_op. Update calls
11779 to the gimple_resimplify functions and maybe_push_res_to_seq.
11780 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11781 res_ops. Update call to the gimple_resimplify functions.
11782 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11783 (decision_tree::gen): Make the functions take a gimple_match_op *
11784 called res_op instead of separate res_code and res_ops parameters.
11785 Update call accordingly.
11786 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11787 and ops with a single res_op parameter. Update calls to
11788 maybe_build_generic_op and maybe_push_res_to_seq.
11789 (fold_stmt_1): Update calls to gimple_simplify and
11790 replace_stmt_with_simplification.
11791 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11792 and gimple_simplified_result_is_gimple_val.
11793 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11795 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11796 with a gimple_match_op *.
11797 (vn_nary_build_or_lookup): Likewise. Update call to
11798 vn_nary_build_or_lookup_1.
11799 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11800 gimple_match_op *. Update calls to the gimple_resimplify routines
11801 and to gimple_simplified_result_is_gimple_val.
11802 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11803 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11804 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11805 (visit_nary_op): Likewise.
11806 (visit_reference_op_load): Likewise.
11808 2018-05-23 Luis Machado <luis.machado@linaro.org>
11810 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11811 modifier for printing the step amount.
11813 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11816 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11819 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11821 * doc/sourcebuild.texi (Endianness): New subsubsection.
11823 2018-05-23 Luis Machado <luis.machado@linaro.org>
11825 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11826 <prefetch_dynamic_strides>: New const bool field.
11827 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11828 prefetch_dynamic_strides.
11829 (exynosm1_prefetch_tune): Likewise.
11830 (thunderxt88_prefetch_tune): Likewise.
11831 (thunderx_prefetch_tune): Likewise.
11832 (thunderx2t99_prefetch_tune): Likewise.
11833 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11835 (aarch64_override_options_internal): Update to set
11836 PARAM_PREFETCH_DYNAMIC_STRIDES.
11837 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11838 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11839 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11840 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11841 prefetch-dynamic-strides setting.
11843 2018-05-23 Luis Machado <luis.machado@linaro.org>
11845 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11846 <minimum_stride>: New const int field.
11847 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11848 minimum_stride field defaulting to -1.
11849 (exynosm1_prefetch_tune): Likewise.
11850 (thunderxt88_prefetch_tune): Likewise.
11851 (thunderx_prefetch_tune): Likewise.
11852 (thunderx2t99_prefetch_tune): Likewise.
11853 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11854 <default_opt_level>: Set to 3.
11855 (aarch64_override_options_internal): Update to set
11856 PARAM_PREFETCH_MINIMUM_STRIDE.
11857 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11858 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11859 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11860 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11861 stride is constant and is below the minimum stride threshold.
11863 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11865 * config/arm/arm-cpus.in (mode26): Delete.
11866 (armv4): Delete mode26 reference.
11867 * config/arm/arm.c (arm_configure_build_target): Delete use of
11870 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11872 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11874 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11875 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11877 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11878 (floatunsdidf2): Ditto.
11880 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11882 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11883 (fixuns_trunc<mode>si2_avx512f): Ditto.
11884 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11885 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11886 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11888 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11890 PR rtl-optimization/79985
11891 * df-scan.c (df_insn_refs_collect): Remove special case for
11892 global registers and asm statements.
11894 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11896 * extend.texi (Global Register Variables): Rewrite the bullet list.
11897 Note that the register is available for allocation. Note that access
11898 via inline asm must use constraints. Add note about async-signal
11899 handlers. Remove paragraph about automagic register selection.
11901 2018-05-23 Richard Biener <rguenther@suse.de>
11903 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11904 of fixed offset from memset VN.
11906 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11908 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11909 first_interp field.
11910 (alloc_cand_and_find_basis): Initialize first_interp field.
11911 (slsr_process_mul): Modify first_interp field.
11912 (slsr_process_add): Likewise.
11913 (slsr_process_cast): Modify first_interp field for each new
11915 (slsr_process_copy): Likewise.
11916 (dump_candidate): Dump first_interp field.
11917 (replace_mult_candidate): Process all interpretations, not just
11919 (replace_rhs_if_not_dup): Likewise.
11920 (replace_one_candidate): Likewise.
11922 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11924 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11926 (aarch64_needs_frame_chain): New function.
11927 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11929 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11932 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11933 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11934 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11935 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11936 as true for strict-align.
11937 (aarch64_can_inline_p): Perform checks even when callee has no
11938 attributes to check for strict alignment.
11939 * doc/extend.texi (AArch64 Function Attributes): Document
11941 * doc/invoke.texi: (AArch64 Options): Likewise.
11943 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11945 PR tree-optimization/85853
11946 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11947 the handling of the root of the node to...
11948 (vect_slp_analyze_node_operations_1): ...this new function,
11949 and run the whole thing with the child nodes' def types
11950 set according to their SLP node's def type.
11952 2018-05-23 Richard Biener <rguenther@suse.de>
11954 PR middle-end/85874
11955 * tree-data-ref.c (create_runtime_alias_checks): Defer
11956 and ignore overflow warnings.
11958 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11960 PR tree-optimization/85822
11961 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11964 2018-05-23 Richard Biener <rguenther@suse.de>
11966 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11967 memset constants via native_interpret_expr.
11969 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11972 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11974 (cgraph_node::create_alias): Likewise.
11975 (cgraph_node::get_availability): Check ifunc_resolver instead
11976 of looking up ifunc attribute.
11977 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11978 * varasm.c (do_assemble_alias): Likewise.
11979 (assemble_alias): Likewise.
11980 (default_binds_local_p_3): Likewise.
11981 * cgraph.h (cgraph_node): Add ifunc_resolver.
11982 (cgraph_node::only_called_directly_or_aliased_p): Return false
11983 for IFUNC resolver.
11984 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11986 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11987 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11988 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11989 instead of looking up ifunc attribute.
11991 2018-05-22 Luis Machado <luis.machado@linaro.org>
11993 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11995 2018-05-22 Martin Sebor <msebor@redhat.com>
11997 PR middle-end/85359
11998 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11999 only when expasion succeeds.
12000 (expand_builtin_strcmp): Same.
12001 (expand_builtin_strncmp): Same.
12003 2018-05-22 Martin Sebor <msebor@redhat.com>
12005 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12007 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12008 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12010 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12011 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12013 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12014 Add check for legitimate_address.
12015 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12016 (aarch64_swap_ldrstr_operands): New.
12017 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12020 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12021 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12023 * config/aarch64/aarch64.md: New patterns to generate stp
12025 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12026 single words and double words.
12027 (load_pair_sw, load_pair_dw): Likewise.
12028 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12030 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12032 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12033 for different mode ldpstp and add peephole for merged zero stores.
12034 Likewise for loads.
12035 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12037 (aarch64_gen_store_pair): Rename calls to match new patterns.
12038 (aarch64_gen_load_pair): Rename calls to match new patterns.
12039 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12040 (load_pair<DREG:mode><DREG2:mode>): ... This.
12041 (store_pair<mode>): Rename to...
12042 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12043 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12044 New mode iterators.
12045 (V_INT_EQUIV): Handle SImode.
12046 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12049 2018-05-22 Martin Sebor <msebor@redhat.com>
12052 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12053 or ajust the presumed bound on an operation to avoid unnecessary
12056 2018-05-22 Martin Sebor <msebor@redhat.com>
12058 PR tree-optimization/85826
12059 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12060 assuming that a DECL necesarily has a constant size.
12062 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12064 PR middle-end/85862
12065 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12067 2018-05-22 Richard Biener <rguenther@suse.de>
12069 PR tree-optimization/85834
12070 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12071 non-constant and non-zero memset arguments.
12073 2018-05-22 Martin Liska <mliska@suse.cz>
12076 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12078 2018-05-22 Richard Biener <rguenther@suse.de>
12080 PR tree-optimization/85863
12081 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12082 comparisons when vectype is specified.
12083 (vectorizable_condition): Do not specify vectype for
12084 vect_is_simple_cond when SLP vectorizing.
12086 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12089 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12090 define __ibm128 as long double.
12091 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12092 as a distinct type when IEEE 128-bit support is enabled.
12093 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12094 128-bit types to use the correct functions.
12095 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12096 convert between 128-bit floating point types that have different
12097 modes but the same representation, instead of using gen_lowpart to
12099 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12101 (IFKF_reg): New attributes to give the register constraints for
12103 (extend<mode>tf2_internal): New insns to mark an explicit
12104 conversion between 128-bit floating point types that have a
12105 different mode but share the same representation.
12107 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12109 PR tree-optimization/85814
12110 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12111 a null return from get_strinfo when unsharing the next
12112 strinfo in the chain.
12114 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12117 * varasm.c (weak_finish): Clean up weak_decls.
12119 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12121 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12122 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12123 UNSPEC_UADALP values.
12124 * config/aarch64/iterators.md (ABAL): New int iterator.
12127 (sur): Add mappings for the above.
12128 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12130 (aarch64_<sur>abal<mode>_4): Likewise.
12131 (aarch64_<sur>adalp<mode>_3): Likewise.
12132 (<sur>sadv16qi): New define_expand.
12134 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12136 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12137 (*movdf_internal): Ditto.
12138 (*rcpsf2_sse): Ditto.
12139 (*rsqrtsf2_sse): Ditto.
12140 (*sqrt<mode>2_sse): Ditto.
12142 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12144 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12146 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12147 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12148 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12149 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12151 * config/aarch64/arm_neon.h: Likewise.
12152 * config/aarch64/iterators.md (VQ_I): New.
12154 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12156 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12158 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12161 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12163 * config/nds32/constraints.md (S): New constraint.
12164 * config/nds32/nds32.md (call_internal): Use constraint S.
12165 (call_value_internal): Likewise.
12166 (sibcall_internal): Likewise.
12167 (sibcall_value_internal): Likewise.
12169 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12170 Chung-Ju Wu <jasonwucj@gmail.com>
12172 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12173 into consideration.
12175 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12176 Chung-Ju Wu <jasonwucj@gmail.com>
12178 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12179 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12180 (nds32_rtx_costs_impl): Simplify.
12181 (nds32_address_cost_impl): Simplify.
12182 (nds32_init_rtx_costs): New function.
12183 (nds32_rtx_costs_speed_prefer): Likewise.
12184 (nds32_rtx_costs_size_prefer): Likewise.
12185 (nds32_address_cost_speed_prefer): Likewise.
12186 (nds32_address_cost_speed_fwprop): Likewise.
12187 (nds32_address_cost_size_prefer): Likewise.
12188 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12189 * config/nds32/nds32.c (nds32_option_override): Use
12190 nds32_init_rtx_costs function.
12192 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12194 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12195 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12196 (TARGET_PIPELINE_N8): Likewise.
12197 (TARGET_PIPELINE_N10): Likewise.
12198 (TARGET_PIPELINE_N13): Likewise.
12199 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12201 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12203 * config/nds32/nds32-fpu.md: Update copyright year.
12205 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12207 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12209 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12211 * config/nds32/nds32.c
12212 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12213 * config/nds32/nds32.opt (minline-asm-r15): New option.
12215 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12217 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12219 * config/nds32/nds32.md (abssi2): New pattern.
12221 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12223 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12224 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12225 Merge insn pattern from sse_cvtsi2ss<round_name> and
12226 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12227 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12228 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12229 using SWI48 mode iterator.
12230 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12231 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12232 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12233 pattern from sse_cvttss2si<round_saeonly_name>
12234 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12235 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12236 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12237 using SWI48 mode iterator.
12238 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12239 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12240 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12241 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12242 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12243 using SWI48 mode iterator.
12244 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12245 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12246 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12247 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12248 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12249 SWI48 mode iterator.
12250 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12251 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12252 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12253 pattern from sse_cvttsd2si<round_saeonly_name>
12254 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12256 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12258 * config/nds32/nds32-md-auxiliary.c
12259 (nds32_valid_smw_lwm_base_p): Refine.
12260 (nds32_output_smw_single_word): Refine.
12261 (nds32_output_smw_double_word): New.
12262 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12264 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12266 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12267 (nds32_output_stack_pop): Refine.
12268 (nds32_expand_unaligned_load): Refine.
12269 (nds32_expand_unaligned_store): Refine.
12271 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12272 Chung-Ju Wu <jasonwucj@gmail.com>
12274 * config/nds32/constants.md: Add TP_REGNUM constant.
12275 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12276 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12278 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12279 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12280 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12281 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12282 and PIC code generation.
12283 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12285 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12287 * config/nds32/nds32.md: Support TLS and PIC.
12288 * config/nds32/nds32.c: Support TLS and PIC.
12289 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12290 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12293 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12295 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12296 mode with E_ prefix.
12298 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12299 Chung-Ju Wu <jasonwucj@gmail.com>
12301 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12302 * config/nds32/nds32-md-auxiliary.c
12303 (symbolic_reference_mentioned_p): New.
12304 (nds32_legitimize_ict_address): New.
12305 (nds32_expand_ict_move): New.
12306 (nds32_indirect_call_referenced_p): New.
12307 (nds32_symbol_binds_local_p): Delete.
12308 (nds32_long_call_p): Modify.
12309 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12310 * config/nds32/nds32-protos.h
12311 (symbolic_reference_mentioned_p): Declare.
12312 (nds32_legitimize_ict_address): Declare.
12313 (nds32_expand_ict_move): Declare.
12314 (nds32_indirect_call_referenced_p): Declare.
12315 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12316 (nds32_relax_group): Use nds32_ict_const_p as condition.
12317 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12318 (nds32_asm_file_start): Output ict_model directive in asm code.
12319 (nds32_legitimate_address_p): Consider indirect call.
12320 (nds32_print_operand): Consider indirect call.
12321 (nds32_print_operand_address): Consider indirect call.
12322 (nds32_insert_attributes): Handle "indirect_call" attribute.
12323 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12324 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12325 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12326 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12327 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12328 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12329 (TARGET_ICT_MODEL_SMALL): Define.
12330 (TARGET_ICT_MODEL_LARGE): Define.
12331 * config/nds32/nds32.md (movsi): Consider ict model.
12332 (call, call_value): Consider ict model.
12333 (sibcall, sibcall_value): Consider ict model.
12334 * config/nds32/nds32.opt (mict-model): New option.
12335 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12338 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12339 Monk Chiang <sh.chiang04@gmail.com>
12340 Jim Wilson <jimw@sifive.com>
12342 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12343 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12344 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12345 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12346 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12347 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12348 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12349 compute save_libcall_adjustment properly.
12350 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12351 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12352 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12353 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12354 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12355 (ABI_SPEC): Handle mabi=ilp32e.
12356 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12357 (RVE): Add RVE mask.
12358 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12359 <-march>: Add rv32e as an example.
12361 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12364 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12365 (intra_create_variable_infos): Handle C++ constructors.
12367 2018-05-18 Martin Liska <mliska@suse.cz>
12369 * passes.def: Remove a redundant pass.
12371 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12374 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12376 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12378 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12380 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12382 (armv2, armv2a, armv3, armv3m): Delete architectures.
12383 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12384 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12385 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12387 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12388 (*mulsidi3adddi): Likewise.
12389 (mulsidi3): Likewise.
12390 (*mulsidi3_nov6): Likewise.
12391 (umulsidi3): Likewise.
12392 (umulsidi3_nov6): Likewise.
12393 (umaddsidi4): Likewise.
12394 (*umulsidi3adddi): Likewise.
12395 (smulsi3_highpart): Likewise.
12396 (*smulsi3_highpart_nov6): Likewise.
12397 (umulsi3_highpart): Likewise.
12398 (*umulsi3_highpart_nov6): Likewise.
12399 * config/arm/arm.h (arm_arch3m): Delete.
12400 * config/arm/arm.c (arm_arch3m): Delete.
12401 (arm_option_override_internal): Update armv3-related comment.
12402 (arm_configure_build_target): Delete use of isa_bit_mode32.
12403 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12404 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12405 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12406 (mulsa3): Likewise.
12407 (mulusa3): Likewise.
12408 * config/arm/arm-protos.h (arm_arch3m): Delete.
12409 * config/arm/arm-tables.opt: Regenerate.
12410 * config/arm/arm-tune.md: Likewise.
12411 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12412 deleted architectures.
12414 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12416 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12417 (armv5t, armv5te): New features.
12418 (ARMv5, ARMv5e): Delete fgroups.
12419 (ARMv5t, ARMv5te): Adjust for above changes.
12420 (ARMv6m): Likewise.
12421 (armv5, armv5e): Delete arches.
12422 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12424 (*call_reg_arm): Likewise.
12425 (*call_value_reg_armv5): Likewise.
12426 (*call_value_reg_arm): Likewise.
12427 (*call_symbol): Likewise.
12428 (*call_value_symbol): Likewise.
12429 (*sibcall_insn): Likewise.
12430 (*sibcall_value_insn): Likewise.
12431 (clzsi2): Likewise.
12432 (prefetch): Likewise.
12433 (define_split and define_peephole2 dependent on arm_arch5):
12435 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12437 (TARGET_ARM_QBIT): Likewise.
12438 (TARGET_DSP_MULTIPLY): Likewise.
12439 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12440 (arm_arch5, arm_arch5e): Delete.
12441 (arm_arch5t, arm_arch5te): Declare.
12442 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12443 (arm_arch5t): Declare.
12444 (arm_option_reconfigure_globals): Update for the above.
12445 (arm_options_perform_arch_sanity_checks): Update comment, replace
12446 use of arm_arch5 with arm_arch5t.
12447 (use_return_insn): Likewise.
12448 (arm_emit_call_insn): Likewise.
12449 (output_return_instruction): Likewise.
12450 (arm_final_prescan_insn): Likewise.
12451 (arm_coproc_builtin_available): Likewise.
12452 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12453 arm_arch5e with arm_arch5t and arm_arch5te.
12454 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12455 (arm_arch5t, arm_arch5te): Declare.
12456 * config/arm/arm-tables.opt: Regenerate.
12457 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12458 * config/arm/t-multilib: Likewise.
12459 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12460 instead of arm_arch5.
12461 (*call_reg_thumb1): Likewise.
12462 (*call_value_reg_thumb1_v5): Likewise.
12463 (*call_value_reg_thumb1): Likewise.
12464 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12466 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12468 2018-05-18 Martin Liska <mliska@suse.cz>
12470 PR gcov-profile/84846
12471 * doc/gcov.texi: Document -t option of gcov tool.
12473 2018-05-18 Martin Liska <mliska@suse.cz>
12475 PR gcov-profile/84846
12476 * gcov.c (print_usage): Add new -t option.
12477 (process_args): Handle the option.
12478 (generate_results): Use stdout as output when requested by
12481 2018-05-18 Martin Liska <mliska@suse.cz>
12483 PR gcov-profile/84846
12484 * coverage.c (coverage_init): Write PWD to .gcno file.
12485 * doc/gcov.texi: Document how working directory is printed.
12486 * gcov-dump.c (dump_gcov_file): Print PWD.
12487 * gcov.c (output_intermediate_file): Likewise.
12488 (read_graph_file): Read PWD string.
12489 (output_lines): Print PWD.
12491 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12493 PR middle-end/85817
12494 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12495 for retval and return false if all args to phi are zero.
12497 2018-05-18 Richard Biener <rguenther@suse.de>
12499 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12501 (evrp_dom_walker::before_dom_children): Call it.
12503 2018-05-18 Richard Biener <rguenther@suse.de>
12505 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12506 results when processing array refs with variable index.
12508 2018-05-18 Toon Moene <toon@moene.org>
12510 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12511 directly after that of -floop-interchange. Indicate that both
12512 options are enabled by default when specifying -O3.
12514 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12516 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12517 iterator. Delete separate integer-mode vec_set<mode> expander.
12518 (aarch64_simd_vec_setv2di): Delete.
12519 (vec_setv2di): Delete.
12520 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12521 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12522 the "w, r" alternative.
12524 2018-05-18 Martin Liska <mliska@suse.cz>
12526 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12527 * tree-pass.h (make_pass_lower_switch_O0): New function.
12528 * tree-switch-conversion.c (node_has_low_bound): Remove.
12529 (node_has_high_bound): Likewise.
12530 (node_is_bounded): Likewise.
12531 (class pass_lower_switch): Make it a template type and create
12533 (pass_lower_switch::execute): Add template argument.
12534 (make_pass_lower_switch): New function.
12535 (make_pass_lower_switch_O0): New function.
12536 (do_jump_if_equal): Remove.
12537 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12538 all the hard work to tree optimization passes.
12540 2018-05-18 Martin Liska <mliska@suse.cz>
12542 * dbgcnt.c (limit_low): Renamed from limit.
12543 (limit_high): New variable.
12544 (dbg_cnt_is_enabled): Check for upper limit.
12545 (dbg_cnt): Adjust dumping.
12546 (dbg_cnt_set_limit_by_index): Add new argument for high
12548 (dbg_cnt_set_limit_by_name): Likewise.
12549 (dbg_cnt_process_single_pair): Parse new format.
12550 (dbg_cnt_process_opt): Use strtok.
12551 (dbg_cnt_list_all_counters): Remove 'value' and add
12553 * doc/invoke.texi: Document changes.
12555 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12557 * doc/sourcebuild.texi (scalar_all_fma): Document.
12558 * tree.def (FMA_EXPR): Delete.
12559 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12560 * internal-fn.c (ternary_direct): New macro.
12561 (expand_ternary_optab_fn): Likewise.
12562 (direct_ternary_optab_supported_p): Likewise.
12563 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12564 * builtins.c (fold_builtin_fma): Delete.
12565 (fold_builtin_3): Don't call it.
12566 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12567 * expr.c (expand_expr_real_2): Likewise.
12568 * fold-const.c (operand_equal_p): Likewise.
12569 (fold_ternary_loc): Likewise.
12570 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12571 * gimple.c (DEFTREECODE): Likewise.
12572 * gimplify.c (gimplify_expr): Likewise.
12573 * optabs-tree.c (optab_for_tree_code): Likewise.
12574 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12575 * tree-eh.c (operation_could_trap_p): Likewise.
12576 (stmt_could_throw_1_p): Likewise.
12577 * tree-inline.c (estimate_operator_cost): Likewise.
12578 * tree-pretty-print.c (dump_generic_node): Likewise.
12579 (op_code_prio): Likewise.
12580 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12581 * tree-ssa-operands.c (get_expr_operands): Likewise.
12582 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12583 * fold-const-call.h (fold_fma): Delete.
12584 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12585 CFN_FNMA and CFN_FNMS.
12586 (fold_fma): Delete.
12587 * genmatch.c (combined_fn): New enum.
12588 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12589 (commutative_op): New function.
12590 (commutate): Use it. Handle more than 2 operands.
12591 (dt_operand::gen_gimple_expr): Use commutative_op.
12592 (parser::parse_expr): Allow :c to be used with non-binary
12593 operators if the commutative operand is known.
12594 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12595 CFN_FMS, CFN_FNMA and CFN_FNMS.
12596 (backprop::process_assign_use): Remove FMA_EXPR handling.
12597 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12598 (gen_hsa_fma): New function.
12599 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12600 IFN_FNMA and IFN_FNMS.
12601 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12602 * gimple-fold.h (follow_all_ssa_edges): Declare.
12603 * gimple-fold.c (follow_all_ssa_edges): New function.
12604 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12605 gimple_build interface and use follow_all_ssa_edges to fold the result.
12606 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12607 instead of checking for optabs directly.
12608 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12609 rather than FMA_EXPRs.
12610 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12611 call to IFN_FMA instead of an FMA_EXPR.
12613 2018-05-17 Jim Wilson <jimw@sifive.com>
12615 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12616 sign extended promoted subreg, and the range does not have the sign bit
12617 set, then do a sign extend.
12619 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12620 test, check for sign extended subreg and/or constant operands, and
12621 do a sign extend in that case.
12623 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12625 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12626 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12628 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12629 Change logics_shift_reg to logics_shift_imm.
12630 (thunderx2t99_fp_loadpair_basic): Delete.
12631 (thunderx2t99_fp_storepair_basic): Delete.
12632 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12633 (thunderx2t99_asimd_polynomial): Delete.
12634 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12635 and neon_fp_mul_d_scalar_q.
12636 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12637 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12638 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12639 (thunderx2t99_asimd_lut): Add missing tbl types.
12640 (thunderx2t99_asimd_ext): Delete.
12641 (thunderx2t99_asimd_load1_1_mult): Delete.
12642 (thunderx2t99_asimd_load1_2_mult): Delete.
12643 (thunderx2t99_asimd_load1_ldp): New.
12644 (thunderx2t99_asimd_load1): New.
12645 (thunderx2t99_asimd_load2): Add missing *load2* types.
12646 (thunderx2t99_asimd_load3): New.
12647 (thunderx2t99_asimd_load4): New.
12648 (thunderx2t99_asimd_store1_1_mult): Delete.
12649 (thunderx2t99_asimd_store1_2_mult): Delete.
12650 (thunderx2t99_asimd_store2_mult): Delete.
12651 (thunderx2t99_asimd_store2_onelane): Delete.
12652 (thunderx2t99_asimd_store_stp): New.
12653 (thunderx2t99_asimd_store1): New.
12654 (thunderx2t99_asimd_store2): New.
12655 (thunderx2t99_asimd_store3): New.
12656 (thunderx2t99_asimd_store4): New.
12658 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
12660 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12661 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
12663 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
12664 Segher Boessenkool <segher@kernel.crashing.org>
12667 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
12670 2018-05-17 Richard Biener <rguenther@suse.de>
12672 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
12673 for pruning loop and prune defs feeding only already visited PHIs.
12675 2018-05-17 Richard Biener <rguenther@suse.de>
12677 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
12679 2018-05-17 Bin Cheng <bin.cheng@arm.com>
12680 Richard Biener <rguenther@suse.de>
12682 PR tree-optimization/85793
12683 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
12684 for VMAT_ELEMENTWISE.
12686 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12688 * internal-fn.h (lookup_internal_fn): Declare
12689 * internal-fn.c (lookup_internal_fn): New function.
12690 * gimple.c (gimple_build_call_from_tree): Handle calls to
12691 internal functions.
12692 * gimple-pretty-print.c (dump_gimple_call): Print "." before
12693 internal function names.
12694 * tree-pretty-print.c (dump_generic_node): Likewise.
12695 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12697 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12699 * gimple-fold.h (gimple_build): Make the function forms take
12700 combined_fn rather than built_in_function.
12701 (gimple_simplify): Likewise.
12702 * gimple-match-head.c (gimple_simplify): Likewise.
12703 * gimple-fold.c (gimple_build): Likewise.
12704 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
12705 rather than gimple_build_call_internal.
12706 (get_initial_defs_for_reduction): Likewise.
12707 (vect_create_epilog_for_reduction): Likewise.
12708 (vectorizable_live_operation): Likewise.
12710 2018-05-17 Martin Liska <mliska@suse.cz>
12712 * gimple-ssa-sprintf.c (format_directive): Do not use
12713 space in between 'G_' and '('.
12715 2018-05-17 Jakub Jelinek <jakub@redhat.com>
12718 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12719 even if the mask is not all ones.
12722 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12724 (ix86_gimple_fold_builtin): Likewise.
12727 * config/i386/i386.c: Include tree-vector-builder.h.
12728 (ix86_vector_shift_count): New function.
12729 (ix86_fold_builtin): Fold shift builtins by scalar count.
12730 (ix86_gimple_fold_builtin): Likewise.
12732 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
12733 _mm512_setzero): New intrinsics.
12735 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
12736 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12738 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
12739 code generation for cases where splatting a value is not useful.
12740 * simplify-rtx.c (simplify_ternary_operation): Simplify
12741 vec_merge across a vec_duplicate and a paradoxical subreg forming
12742 a vector mode to a vec_concat.
12744 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
12746 * config.gcc: Support "goldmont-plus".
12747 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12749 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12750 PROCESSOR_GOLDMONT_PLUS.
12751 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
12752 (processor_target_table): Add "goldmont-plus".
12753 (PTA_GOLDMONT_PLUS): Define.
12754 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
12755 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
12756 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
12757 (fold_builtin_cpu): Add "goldmont-plus".
12758 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
12759 (ix86_option_override_internal): Add "goldmont-plus".
12760 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
12761 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
12762 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
12763 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
12765 2018-05-17 Richard Biener <rguenther@suse.de>
12767 PR tree-optimization/85757
12768 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
12769 remove defs that only feed that PHI from further processing.
12771 2018-05-16 Jim Wilson <jimw@sifive.com>
12773 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
12775 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
12777 2018-05-16 Mark Wielaard <mark@klomp.org>
12779 * dwarf2out.c (count_index_strings): New function.
12780 (output_indirect_strings): Call count_index_strings and generate
12781 header for dwarf_version >= 5.
12783 2018-05-16 Mark Wielaard <mark@klomp.org>
12785 * dwarf2out.c (dwarf_FORM): New function.
12786 (set_indirect_string): Use dwarf_FORM.
12787 (reset_indirect_string): Likewise.
12788 (size_of_die): Likewise.
12789 (value_format): Likewise.
12790 (output_die): Likewise.
12791 (add_skeleton_AT_string): Likewise.
12792 (output_macinfo_op): Likewise.
12793 (index_string): Likewise.
12794 (output_index_string_offset): Likewise.
12795 (output_index_string): Likewise.
12796 (count_index_strings): Likewise.
12798 2018-05-16 Carl Love <cel@us.ibm.com>
12800 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12801 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12803 2018-05-16 Martin Jambor <mjambor@suse.cz>
12805 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12806 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12808 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12810 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12811 (fnma<mode>4): Likewise.
12812 (fms<mode>4): Likewise.
12813 (fnms<mode>4): Likewise.
12814 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12815 (aarch64_fnma<mode>4): Likewise.
12816 (aarch64_fms<mode>4): Likewise.
12817 (aarch64_fnms<mode>4): Likewise.
12818 (aarch64_fnmadd<mode>4): Likewise.
12820 2018-05-16 Jason Merrill <jason@redhat.com>
12822 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12824 2018-05-16 Richard Biener <rguenther@suse.de>
12826 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12827 (dump_stmt_cost): Declare.
12828 (add_stmt_cost): Dump cost we add.
12829 (add_stmt_costs): New function.
12830 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12831 No longer exported.
12832 (vect_analyze_stmt): Adjust prototype.
12833 (vectorizable_condition): Likewise.
12834 (vectorizable_live_operation): Likewise.
12835 (vectorizable_reduction): Likewise.
12836 (vectorizable_induction): Likewise.
12837 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12838 cost vector to pass to vectorizable_ and record afterwards.
12839 (vect_model_reduction_cost): Take cost vector argument and adjust.
12840 (vect_model_induction_cost): Likewise.
12841 (vectorizable_reduction): Likewise.
12842 (vectorizable_induction): Likewise.
12843 (vectorizable_live_operation): Likewise.
12844 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12845 SLP_TREE_NUMBER_OF_VEC_STMTS.
12846 (vect_analyze_slp_cost_1): Remove.
12847 (vect_analyze_slp_cost): Likewise.
12848 (vect_slp_analyze_node_operations): Take visited args and
12849 a target cost vector. Avoid processing already visited stmt sets.
12850 (vect_slp_analyze_operations): Use a local cost vector to gather
12851 costs and register those of non-discarded instances.
12852 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12853 (vect_schedule_slp_instance): Remove copying of
12854 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12856 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12857 adding cost. Record cost entry location.
12858 (vect_prologue_cost_for_slp_op): Function to compute cost of
12859 a constant or invariant generated for SLP vect in the prologue,
12860 split out from vect_analyze_slp_cost_1.
12861 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12862 (vect_model_promotion_demotion_cost): Likewise.
12863 (vect_model_store_cost): Likewise, make static.
12864 (vect_model_load_cost): Likewise.
12865 (vectorizable_bswap): Add cost vector arg and adjust.
12866 (vectorizable_call): Likewise.
12867 (vectorizable_simd_clone_call): Likewise.
12868 (vectorizable_conversion): Likewise.
12869 (vectorizable_assignment): Likewise.
12870 (vectorizable_shift): Likewise.
12871 (vectorizable_operation): Likewise.
12872 (vectorizable_store): Likewise.
12873 (vectorizable_load): Likewise.
12874 (vectorizable_condition): Likewise.
12875 (vectorizable_comparison): Likewise.
12876 (can_vectorize_live_stmts): Likewise.
12877 (vect_analyze_stmt): Likewise.
12878 (vect_transform_stmt): Adjust calls to vectorizable_*.
12879 * tree-vectorizer.c: Include gimple-pretty-print.h.
12880 (dump_stmt_cost): New function.
12882 2018-05-16 Richard Biener <rguenther@suse.de>
12884 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12885 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12886 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12887 (check_name): New callback.
12888 (dse_classify_store): Track cycles via a visited bitmap of PHI
12889 defs and simplify handling of in-loop and across loop dead stores
12890 and properly fail for loop-variant refs. Handle byte-tracking with
12891 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12894 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12896 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12897 (vect_get_mask_type_for_stmt): Likewise.
12898 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12900 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12901 to determine the statement's vector type and the vector type that
12902 should be used for calculating nunits. Deal with cases in which
12903 the type has to be deferred.
12904 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12905 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12906 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12907 (vect_determine_vf_for_stmt): New functions, split out from...
12908 (vect_determine_vectorization_factor): ...here.
12909 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12910 (vect_get_mask_type_for_stmt): New functions, split out from
12911 vect_determine_vectorization_factor.
12913 2018-05-16 Richard Biener <rguenther@suse.de>
12915 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12916 verify the [VEC_]COND_EXPR embedded comparison.
12918 2018-05-15 Martin Sebor <msebor@redhat.com>
12920 PR tree-optimization/85753
12921 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12922 RECORD_TYPE in addition to ARRAY_TYPE.
12924 2018-05-15 Martin Sebor <msebor@redhat.com>
12926 PR middle-end/85643
12927 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12929 2018-05-15 Richard Biener <rguenther@suse.de>
12931 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12932 add by_clobber_p one. Change algorithm to collect all defs
12933 representing uses we need to walk and try reducing them to
12934 a single one before failing.
12935 (dse_dom_walker::dse_optimize_stmt): Adjust.
12937 2018-05-13 Mark Wielaard <mark@klomp.org>
12939 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12940 (size_of_loc_descr): Likewise.
12941 (output_loc_operands): Likewise.
12942 (output_loc_operands_raw): Likewise.
12943 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12944 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12945 (hash_loc_operands): Likewise.
12946 (compare_loc_operands): Likewise.
12948 2018-05-14 Mark Wielaard <mark@klomp.org>
12950 * dwarf2out.c (count_index_addrs): New function.
12951 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12953 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12955 PR tree-optimization/83648
12956 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12957 return value as malloc candidate.
12959 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12962 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12963 param as true in call to suggest_attribute.
12965 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12967 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12968 -mreadonly-in-sdata.
12970 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12972 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12974 (aarch64_crypto_aesd_fused): Likewise.
12976 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12978 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12979 (movsi_aarch64): Likewise.
12980 (load_pairsi): Likewise.
12981 (load_pairdi): Likewise.
12982 (store_pairsi): Likewise.
12983 (store_pairdi): Likewise.
12984 (load_pairsf): Likewise.
12985 (load_pairdf): Likewise.
12986 (store_pairsf): Likewise.
12987 (store_pairdf): Likewise.
12988 (zero_extend): Likewise.
12989 (trunc): Swap alternatives.
12990 (fcvt_target): Add '?' to prefer w over r.
12992 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12995 * config/i386/i386.md: Disallow non-commutative arithmetics in
12996 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12997 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12998 in the peephole2 before it.
13000 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
13002 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13003 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13004 (ix86_handle_option): Handle -mcldemote.
13005 * config.gcc: New header.
13006 * config/i386/cldemoteintrin.h: New file.
13007 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13010 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13011 OPTION_MASK_ISA_CLDEMOTE.
13012 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13013 (ix86_valid_target_attribute_inner_p): Ditto.
13014 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13015 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13016 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13017 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13018 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13020 * config/i386/i386.opt: Add -mcldemote.
13021 * config/i386/x86intrin.h: New header.
13022 * doc/invoke.texi: Add -mcldemote.
13024 2018-05-14 Richard Biener <rguenther@suse.de>
13026 * doc/match-and-simplify.texi: Adjust :s documentation.
13028 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13030 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13031 intended memcpy size.
13032 (REORDER_45): Likewise.
13034 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13036 * sort.cc: New file.
13037 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13038 * vec.c (qsort_chk): Use gcc_qsort.
13039 * Makefile.in (OBJS-libcommon): Add sort.o.
13040 (build/sort.o): New target. Use it...
13041 (BUILD_RTL): ... here, and...
13042 (build/gencfn-macros): ... here, and...
13043 (build/genmatch): ... here.
13045 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13046 Chung-Ju Wu <jasonwucj@gmail.com>
13048 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13049 * config/nds32/nds32-graywolf.md: New file.
13050 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13051 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13053 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13054 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13055 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13056 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13057 * config/nds32/pipelines.md: Include n15 settings.
13059 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13060 Chung-Ju Wu <jasonwucj@gmail.com>
13062 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13063 * config/nds32/nds32-n13.md: New file.
13064 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13065 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13067 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13068 * config/nds32/nds32.md (pipeline_model): Add n13.
13069 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13070 * config/nds32/pipelines.md: Include n13 settings.
13072 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13073 Chung-Ju Wu <jasonwucj@gmail.com>
13075 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13076 * config/nds32/nds32-n10.md: New file.
13077 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13078 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13080 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13081 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13082 * config/nds32/nds32.md (pipeline_model): Add n10.
13083 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13084 * config/nds32/pipelines.md: Include n10 settings.
13086 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13087 Kito Cheng <kito.cheng@gmail.com>
13088 Chung-Ju Wu <jasonwucj@gmail.com>
13090 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13091 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13092 Add enum values for DSP extension instructions.
13093 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13095 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13096 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13097 New code iterators.
13098 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13099 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13100 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13101 * config/nds32/nds32-intrinsic.md: Likewise.
13102 * config/nds32/nds32_intrinsic.h: Likewise.
13103 * config/nds32/nds32-md-auxiliary.c: Likewise.
13104 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13105 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13106 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13107 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13108 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13109 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13110 TYPE_DMAC in switch statement.
13111 * config/nds32/nds32.c: New checking and implementation for DSP
13112 extension instructions.
13113 * config/nds32/nds32.h: Likewise.
13114 * config/nds32/nds32.md: Likewise.
13115 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13116 * config/nds32/predicates.md: Implement new predicates for DSP
13119 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13121 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13122 Reformat alternatives and attributes so it is easier to identify
13123 which constraints/attributes go with which instruction.
13124 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13125 (mov<mode>_softfloat32, FMOVE64): Likewise.
13126 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13127 (mov<mode>_softfloat64, FMOVE64): Likewise.
13129 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13131 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13133 (Basic PowerPC Built-in Functions): The new name of the
13134 subsection previously known as "PowerPC Built-in Functions".
13135 (Basic PowerPC Built-in Functions Available on all Configurations):
13137 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13138 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13139 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13140 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13142 2018-05-11 Martin Jambor <mjambor@suse.cz>
13145 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13148 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13151 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13153 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13155 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13156 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13157 (ix86_handle_option): Handle -mwaitpkg.
13158 * config.gcc: New header.
13159 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13160 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13161 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13163 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13164 OPTION_MASK_ISA_WAITPKG.
13165 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13166 (ix86_option_override_internal): Add PTA_WAITPKG.
13167 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13168 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13169 IX86_BUILTIN_TPAUSE.
13170 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13171 __builtin_ia32_umwait and __builtin_ia32_tpause.
13172 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13173 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13174 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13175 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13176 UNSPECV_TPAUSE): New.
13177 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13178 * config/i386/i386.opt: Add -mwaitpkg.
13179 * config/i386/waitpkgintrin.h: New file.
13180 * config/i386/x86intrin.h: New header.
13181 * doc/invoke.texi: Add -mwaitpkg.
13183 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13186 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13188 (cortex-m0): Use armv6s-m isa.
13189 (cortex-m0plus): Likewise.
13190 (cortex-m1): Likewise.
13191 (cortex-m0.small-multiply): Likewise.
13192 (cortex-m0plus.small-multiply): Likewise.
13193 (cortex-m1.small-multiply): Likewise.
13195 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13196 Jakub Jelinek <jakub@redhat.com>
13198 PR tree-optimization/85692
13199 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13200 source permute as well.
13202 2018-05-11 Martin Liska <mliska@suse.cz>
13205 * doc/extend.texi: Document LLVM style format for no_sanitize
13208 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13210 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13211 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13212 (mode_supports_vsx_dform_quad): Likewise.
13213 (mode_supports_vmx_dform): Move these functions to be next to the
13214 other mode_supports functions.
13215 (mode_supports_dq_form): Likewise.
13216 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13217 mode_supports_dq_form.
13218 (reg_offset_addressing_ok_p): Likewise.
13219 (offsettable_ok_by_alignment): Likewise.
13220 (rs6000_legitimate_offset_address_p): Likewise.
13221 (legitimate_lo_sum_address_p): Likewise.
13222 (rs6000_legitimize_address): Likewise.
13223 (rs6000_legitimize_reload_address): Likewise.
13224 (rs6000_secondary_reload_inner): Likewise.
13225 (rs6000_preferred_reload_class): Likewise.
13226 (rs6000_output_move_128bit): Likewise.
13228 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13230 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13231 Generate SImode target register for null target.
13232 <case IX86_BUILTIN_XGETBV>: Ditto.
13233 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13234 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13236 2018-05-10 Carl Love <cel@us.ibm.com>
13238 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13239 dcbtt and dcbtstt if operands[2] is 0.
13241 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13244 * config/i386/sse.md (usadv64qi): New expander.
13246 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13248 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13249 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13250 -maltivec=be support.
13251 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13252 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13253 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13254 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13255 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13256 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13257 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13258 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13259 altivec_vsumsws): Adjust.
13260 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13261 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13262 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13264 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13265 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13266 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13267 (altivec_lve<VI_char>x): Delete expand.
13268 (*altivec_lve<VI_char>x_internal): Rename to...
13269 (altivec_lve<VI_char>x): ... this.
13270 (altivec_lvxl_<mode>): Delete expand.
13271 (*altivec_lvxl_<mode>_internal): Rename to ...
13272 (altivec_lvxl_<mode>): ... this.
13273 (altivec_stvxl_<mode>): Delete expand.
13274 (*altivec_stvxl_<mode>_internal): Rename to ...
13275 (altivec_stvxl_<mode>): ... this.
13276 (altivec_stve<VI_char>x): Delete expand.
13277 (*altivec_stve<VI_char>x_internal): Rename to ...
13278 (altivec_stve<VI_char>x): ... this.
13279 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13280 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13281 reduc_plus_scal_<mode>): Adjust.
13282 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13284 (rs6000_cpu_cpp_builtins): Adjust.
13285 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13286 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13287 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13288 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13289 -maltivec=be support.
13290 (rs6000_split_vec_extract_var): Adjust.
13291 (rs6000_split_v4si_init): Adjust.
13292 (swap_selector_for_mode): Delete.
13293 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13294 altivec_expand_stvex_be): Delete.
13295 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13296 -maltivec=be support.
13297 (rs6000_gimple_fold_builtin): Ditto.
13298 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13300 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13301 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13302 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13303 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13304 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13305 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13306 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13307 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13308 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13309 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13310 anonymous split): Adjust.
13311 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13312 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13314 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13316 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13317 when --with-gxx-include-dir is also specified.
13318 * configure: Regenerate.
13320 2018-05-09 Jim Wilson <jimw@sifive.com>
13323 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13324 * config/riscv/t-withmultilib: New.
13325 * config/riscv/withmultilib.h: New.
13326 * doc/install.texi: Document RISC-V --with-multilib-list support.
13328 2018-05-09 Richard Biener <rguenther@suse.de>
13330 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13332 (vect_bb_vectorization_profitable_p): Adjust. Compute
13333 actual scalar cost using the cost vector and the add_stmt_cost
13336 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13338 PR rtl-optimization/85645
13339 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13340 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13342 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13344 PR rtl-optimization/85645
13345 * shrink-wrap.c (spread_components): Return a boolean saying if
13346 anything was changed.
13347 (try_shrink_wrapping_separate): Iterate spread_components until
13348 nothing changes anymore.
13350 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13352 PR rtl-optimization/85645
13353 * regrename.c (build_def_use): Also kill the chains that include the
13354 destination of a REG_CFA_REGISTER note.
13356 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13358 PR rtl-optimization/85645
13359 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13360 insn that has a REG_CFA_REGISTER note.
13362 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13364 * cfgexpand.c (expand_clobber): New function.
13365 (expand_gimple_stmt_1): Use it.
13366 * tree-vect-stmts.c (vect_clobber_variable): New function,
13368 (vectorizable_simd_clone_call): ...here.
13369 (vectorizable_store): Emit a clobber either side of an
13370 IFN_STORE_LANES sequence.
13371 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13373 2018-05-09 Tom de Vries <tom@codesourcery.com>
13376 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13377 (define_insn "trap_if_false"): Add exit after trap.
13379 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13381 PR rtl-optimization/85638
13382 * bb-reorder.c: Include common/common-target.h.
13383 (create_forwarder_block): New function extracted from...
13384 (fix_up_crossing_landing_pad): ...here. Rename into...
13385 (dw2_fix_up_crossing_landing_pad): ...this.
13386 (sjlj_fix_up_crossing_landing_pad): New function.
13387 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13388 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13389 from both partitions and exit the loop after one iteration.
13391 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13394 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13396 (Basic PowerPC Built-in Functions): The new name of the
13397 subsection previously known as "PowerPC Built-in Functions".
13398 (Basic PowerPC Built-in Functions Available on all Configurations):
13400 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13402 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13404 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13406 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13409 2018-05-08 Jim Wilson <jimw@sifive.com>
13411 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13412 (LD_EMUL_SUFFIX): New.
13413 (LINK_SPEC): Use it.
13415 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13417 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13419 (Basic PowerPC Built-in Functions): The new name of the
13420 subsection previously known as "PowerPC Built-in Functions".
13421 (Basic PowerPC Built-in Functions Available on all Configurations):
13423 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13425 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13427 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13429 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13432 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13435 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13436 after cmpelim optimization.
13438 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13440 * config.gcc: Support "goldmont".
13441 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13442 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13443 PROCESSOR_GOLDMONT.
13444 * config/i386/i386.c (m_GOLDMONT): Define.
13445 (processor_target_table): Add "goldmont".
13446 (PTA_GOLDMONT): Define.
13447 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13448 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13449 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13450 (fold_builtin_cpu): Add "goldmont".
13451 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13452 (ix86_option_override_internal): Add "goldmont".
13453 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13454 (processor_type): Add PROCESSOR_GOLDMONT.
13455 * config/i386/i386.md: Add CPU "glm".
13456 * config/i386/glm.md: New file.
13457 * config/i386/x86-tune.def: Add m_GOLDMONT.
13458 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13460 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13463 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13465 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13466 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13467 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13470 * config/i386/i386.c (ix86_fold_builtin): Handle
13471 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13474 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13475 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13477 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13480 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13481 (check_arch): Likewise.
13482 (check_fpu): Return the result rather than printing it.
13483 (end arch): Fix operator precedence.
13484 (end cpu): Likewise.
13485 (END): Print the result from check_fpu.
13487 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13488 Alan Hayward <alan.hayward@arm.com>
13489 David Sherwood <david.sherwood@arm.com>
13491 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13492 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13493 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13494 (*fcmuo<mode>_and): New patterns.
13496 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13498 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13499 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13500 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13501 (cmp_op, sve_imm_con): New code attributes.
13502 (SVE_COND_INT_CMP, imm_con): Delete.
13503 (cmp_op): Remove above unspecs from int attribute.
13504 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13506 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13507 comparison-specific unspecs.
13508 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13509 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13510 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13511 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13512 (*vec_fcm<cmp_op><mode>): Rename to...
13513 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13514 (*vec_fcmuo<mode>): Rename to...
13515 (*fcmuo<mode>): ...this and adjust likewise.
13516 (*pred_fcm<cmp_op><mode>): New pattern.
13517 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13518 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13520 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13522 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13523 (aarch64_emit_sve_predicated_cond): New function.
13524 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13525 (aarch64_emit_unspec_cond_or): Replace with...
13526 (aarch64_emit_sve_or_conds): ...this new function. Use
13527 aarch64_emit_sve_ptrue_op for the individual comparisons and
13528 aarch64_emit_binop to OR them together.
13529 (aarch64_emit_inverted_unspec_cond): Replace with...
13530 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13531 aarch64_emit_sve_ptrue_op for the comparison and
13532 aarch64_emit_unop to invert the result.
13533 (aarch64_expand_sve_vec_cmp_float): Update after the above
13534 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13536 2018-05-07 Nathan Sidwell <nathan@acm.org>
13538 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13539 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13540 (Backwards Compatibility): Likewise.
13542 2018-05-07 Luis Machado <luis.machado@linaro.org>
13546 2018-05-07 Luis Machado <luis.machado@linaro.org>
13548 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13549 <prefetch_dynamic_strides>: New const bool field.
13550 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13551 prefetch_dynamic_strides.
13552 (exynosm1_prefetch_tune): Likewise.
13553 (thunderxt88_prefetch_tune): Likewise.
13554 (thunderx_prefetch_tune): Likewise.
13555 (thunderx2t99_prefetch_tune): Likewise.
13556 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13558 (aarch64_override_options_internal): Update to set
13559 PARAM_PREFETCH_DYNAMIC_STRIDES.
13560 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13561 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13562 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13563 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13564 prefetch-dynamic-strides setting.
13566 2018-05-07 Luis Machado <luis.machado@linaro.org>
13568 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13569 <minimum_stride>: New const int field.
13570 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13571 minimum_stride field.
13572 (exynosm1_prefetch_tune): Likewise.
13573 (thunderxt88_prefetch_tune): Likewise.
13574 (thunderx_prefetch_tune): Likewise.
13575 (thunderx2t99_prefetch_tune): Likewise.
13576 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13577 (aarch64_override_options_internal): Update to set
13578 PARAM_PREFETCH_MINIMUM_STRIDE.
13579 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13580 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13581 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13582 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13583 stride is constant and is below the minimum stride threshold.
13585 2018-05-07 Luis Machado <luis.machado@linaro.org>
13587 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13590 2018-05-07 Luis Machado <luis.machado@linaro.org>
13592 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13593 <prefetch_dynamic_strides>: New const bool field.
13594 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13595 prefetch_dynamic_strides.
13596 (exynosm1_prefetch_tune): Likewise.
13597 (thunderxt88_prefetch_tune): Likewise.
13598 (thunderx_prefetch_tune): Likewise.
13599 (thunderx2t99_prefetch_tune): Likewise.
13600 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13602 (aarch64_override_options_internal): Update to set
13603 PARAM_PREFETCH_DYNAMIC_STRIDES.
13604 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13605 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13606 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13607 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13608 prefetch-dynamic-strides setting.
13610 2018-05-07 Luis Machado <luis.machado@linaro.org>
13612 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13613 <minimum_stride>: New const int field.
13614 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13615 minimum_stride field.
13616 (exynosm1_prefetch_tune): Likewise.
13617 (thunderxt88_prefetch_tune): Likewise.
13618 (thunderx_prefetch_tune): Likewise.
13619 (thunderx2t99_prefetch_tune): Likewise.
13620 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13621 (aarch64_override_options_internal): Update to set
13622 PARAM_PREFETCH_MINIMUM_STRIDE.
13623 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13624 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13625 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13626 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13627 stride is constant and is below the minimum stride threshold.
13629 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13632 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13633 the type is addressable. Don't force op into register if it has
13636 2018-05-05 Roland McGrath <mcgrathr@google.com>
13639 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13640 any section for which we don't know a specific type it should have,
13641 regardless of name. Previously this was done only for the exact
13642 names ".init_array", ".fini_array", and ".preinit_array".
13643 (default_elf_asm_named_section): Add comment about
13644 relationship with default_section_type_flags and SECTION_NOTYPE.
13645 (get_section): Don't consider it a type conflict if one side has
13646 SECTION_NOTYPE and the other doesn't, as long as neither has the
13647 SECTION_BSS et al used in the default_section_type_flags logic.
13649 2018-05-05 Tom de Vries <tom@codesourcery.com>
13652 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13653 (workaround_barsyncs): New function.
13654 (nvptx_reorg): Use workaround_barsyncs.
13655 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13656 (define_expand "nvptx_membar_cta"): New define_expand.
13657 (define_insn "*nvptx_membar_cta"): New insn.
13659 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
13661 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13662 To improve optimization opportunities.
13663 * builtin-types.def: The new needed builtin types for the above.
13665 2018-05-04 Richard Biener <rguenther@suse.de>
13667 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
13668 * gimple-ssa-store-merging.c
13669 (imm_store_chain_info::output_merged_store): Remove redundant create,
13670 release split_store vector contents on failure.
13671 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
13672 scalar stmt vector on cache hit.
13674 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
13676 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
13678 * config.gcc (powerpc-xilinx-eabi*): Remove.
13679 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13681 (fusion_addis_mem_combo_load): Ditto.
13682 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
13684 (rs6000_cpu_cpp_builtins): Ditto.
13685 * config/rs6000/rs6000-linux.c
13686 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
13687 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
13688 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
13690 (rs6000_setup_reg_addr_masks): Ditto.
13691 (rs6000_init_hard_regno_mode_ok): Ditto.
13692 (rs6000_option_override_internal): Ditto.
13693 (legitimate_lo_sum_address_p): Ditto.
13694 (rs6000_legitimize_address): Ditto.
13695 (rs6000_legitimize_reload_address): Ditto.
13696 (rs6000_legitimate_address_p): Ditto.
13697 (abi_v4_pass_in_fpr): Ditto.
13698 (setup_incoming_varargs): Ditto.
13699 (rs6000_gimplify_va_arg): Ditto.
13700 (rs6000_split_multireg_move): Ditto.
13701 (rs6000_savres_strategy): Ditto.
13702 (rs6000_emit_prologue_components): Ditto.
13703 (rs6000_emit_epilogue_components): Ditto.
13704 (rs6000_emit_prologue): Ditto.
13705 (rs6000_emit_epilogue): Ditto.
13706 (rs6000_elf_file_end): Ditto.
13707 (rs6000_function_value): Ditto.
13708 (rs6000_libcall_value): Ditto.
13709 * config/rs6000/rs6000.h: Ditto.
13710 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
13711 (TARGET_MINMAX): ... this. New.
13712 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
13713 * config/rs6000/rs6000.md: Remove Xilinx FP support.
13714 (*movsi_internal1_single): Delete.
13715 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
13716 mfpu=, mxilinx-fpu): Delete.
13717 * config/rs6000/singlefp.h: Delete.
13718 * config/rs6000/sysv4.h: Remove Xilinx FP support.
13719 * config/rs6000/t-rs6000: Ditto.
13720 * config/rs6000/t-xilinx: Delete.
13721 * config/rs6000/titan.md: Adjust for fp_type removal.
13722 * config/rs6000/vsx.md: Remove Xilinx FP support.
13723 (VStype_simple): Delete.
13724 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
13725 * config/rs6000/xfpu.h: Delete.
13726 * config/rs6000/xfpu.md: Delete.
13727 * config/rs6000/xilinx.h: Delete.
13728 * config/rs6000/xilinx.opt: Delete.
13729 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
13730 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
13732 2018-05-04 Tom de Vries <tom@codesourcery.com>
13735 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
13738 2018-05-04 Richard Biener <rguenther@suse.de>
13740 PR middle-end/85627
13741 * tree-complex.c (update_complex_assignment): We are always in SSA form.
13742 (expand_complex_div_wide): Likewise.
13743 (expand_complex_operations_1): Likewise.
13744 (expand_complex_libcall): Preserve EH info of the original stmt.
13745 (tree_lower_complex): Handle removed blocks.
13746 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
13747 on complex multiplication and division libcall builtins.
13749 2018-05-04 Richard Biener <rguenther@suse.de>
13751 PR middle-end/85574
13752 * fold-const.c (negate_expr_p): Restrict negation of operand
13753 zero of a division to when we know that can happen without
13755 (fold_negate_expr_1): Likewise.
13757 2018-05-04 Jakub Jelinek <jakub@redhat.com>
13760 * real.h (real_nextafter): Declare.
13761 * real.c (real_nextafter): New function.
13762 * fold-const-call.c (fold_const_nextafter): New function.
13763 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
13764 CASE_CFN_NEXTTOWARD.
13765 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
13766 even when arg1_mode is different from arg0_mode.
13768 2018-05-03 Nathan Sidwell <nathan@acm.org>
13770 * doc/extend.texi (Deprecated Features): Remove
13771 -ffriend-injection.
13772 (Backwards Compatibility): Likewise.
13773 * doc/invoke.texi (C++ Language Options): Likewise.
13774 (C++ Dialect Options): Likewise.
13776 2018-05-03 Jakub Jelinek <jakub@redhat.com>
13779 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
13780 _mm512_mask_mullox_epi64): New intrinsics.
13782 2018-05-03 Tom de Vries <tom@codesourcery.com>
13785 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13786 dump files): Add offload-tree.
13788 2018-05-03 Richard Biener <rguenther@suse.de>
13790 PR tree-optimization/85615
13791 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13792 to loops not nested in BBs loop father to avoid creating multi-entry
13795 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13797 PR tree-optimization/70291
13798 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13799 arguments. Change return type to tree. Emit libcall as a new
13800 statement rather than replacing existing one when inplace_p is true.
13801 (expand_complex_multiplication_components): New function.
13802 (expand_complex_multiplication): Expand floating-point complex
13803 multiplication using the above.
13804 (expand_complex_division): Rename inner_type parameter to type.
13805 Update expand_complex_libcall call-site.
13806 (expand_complex_operations_1): Update expand_complex_multiplication
13807 and expand_complex_division call-sites.
13809 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13812 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13813 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13814 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13815 the highest significant bit of the shift count mask is clear. In
13816 check whether and[sq]i3 is needed verify that all significant bits
13817 of the shift count other than the highest are set.
13819 2018-05-02 Tom de Vries <tom@codesourcery.com>
13822 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13823 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13824 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13825 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13826 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13827 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13828 __builtin_goacc_parlevel_size.
13830 2018-05-02 Richard Biener <rguenther@suse.de>
13832 PR tree-optimization/85597
13833 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13834 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13837 2018-05-02 Tom de Vries <tom@codesourcery.com>
13840 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13841 dump files): Add ltrans-tree.
13843 2018-05-02 Tom de Vries <tom@codesourcery.com>
13846 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13847 dump files): Add wpa-ipa.
13849 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13851 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13852 powerpc*-*-linux*paired* target.
13853 * config/rs6000/750cl.h: Delete.
13854 * config/rs6000/paired.h: Delete.
13855 * config/rs6000/paired.md: Delete.
13856 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13858 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13859 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13860 comment. Remove paired float support.
13861 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13862 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13864 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13865 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13867 * config/rs6000/rs6000.c: Remove paired float support.
13868 (paired_expand_vector_init, paired_expand_vector_move,
13869 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13870 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13871 paired_expand_builtin, paired_expand_predicate_builtin,
13872 paired_init_builtins): Delete.
13873 * config/rs6000/rs6000.h: Remove paired float support.
13874 * config/rs6000/rs6000.md: Remove paired float support.
13875 (move_from_CR_ov_bit): Delete.
13876 * config/rs6000/rs6000.opt (mpaired): Delete.
13877 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13878 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13880 2018-05-02 Richard Biener <rguenther@suse.de>
13882 PR middle-end/85567
13883 * gimplify.c (gimplify_save_expr): When in SSA form allow
13884 SAVE_EXPRs to compute to SSA vars.
13886 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13889 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13890 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13891 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13892 clobber operands[2], instead use a new pseudo. Formatting fixes.
13894 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13896 PR tree-optimization/85586
13897 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13898 exit early for statements in the same group if the accesses are
13901 2018-05-02 Tom de Vries <tom@codesourcery.com>
13904 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13907 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13909 PR tree-optimization/85143
13910 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13912 2018-05-01 Tom de Vries <tom@codesourcery.com>
13915 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13916 not found" error message.
13918 2018-05-01 Tom de Vries <tom@codesourcery.com>
13921 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13922 * vec.c (test_ordered_remove_if): New function.
13923 (vec_c_tests): Call test_ordered_remove_if.
13924 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13925 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13926 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13927 VEC_ORDERED_REMOVE_IF.
13929 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13931 PR tree-optimization/82665
13932 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13933 pointer subtraction where arguments come from a memchr call.
13935 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13937 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13938 --push-state --as-needed and --pop-state instead of --as-needed and
13939 --no-as-needed if ld supports it.
13940 * configure: Regenerated.
13943 * doc/install.texi2html: Replace _002d with - and _002a with * in
13944 generated html files using sed.
13946 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13949 * gcc-rich-location.c (blank_line_before_p): New function.
13950 (use_new_line): New function.
13951 (gcc_rich_location::add_fixit_insert_formatted): New function.
13952 * gcc-rich-location.h
13953 (gcc_rich_location::add_fixit_insert_formatted): New function.
13955 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13957 * selftest.c (assert_streq): Rename "expected" and "actual" to
13958 "val1" and "val2". Extend NULL-handling to cover both inputs
13959 symmetrically, while still requiring both to be non-NULL for a pass.
13960 * selftest.h (assert_streq): Rename "expected" and "actual" to
13962 (ASSERT_EQ): Likewise.
13963 (ASSERT_EQ_AT): Likewise.
13964 (ASSERT_KNOWN_EQ): Likewise.
13965 (ASSERT_KNOWN_EQ_AT): Likewise.
13966 (ASSERT_NE): Likewise.
13967 (ASSERT_MAYBE_NE): Likewise.
13968 (ASSERT_MAYBE_NE_AT): Likewise.
13969 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13970 the assertion to pass.
13971 (ASSERT_STREQ_AT): Likewise.
13973 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13975 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13976 interaction with -pie.
13978 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13980 * selftest.h: Fix alphabetization of per-source-file selftest
13983 2018-04-30 Jason Merrill <jason@redhat.com>
13985 PR c++/61982 - dead stores to destroyed objects.
13986 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13989 2018-04-30 Jason Merrill <jason@redhat.com>
13991 * tree.c (build_clobber): New.
13992 * tree.h: Declare it.
13993 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13995 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13997 * diagnostic-show-locus.c (layout::layout): Update for
13998 location_get_source_line returning a char_span.
13999 (struct char_span): Move to input.h.
14000 (struct correction): Update for fields in char_span becoming
14002 (struct source_line): Update for location_get_source_line
14003 returning a char_span.
14004 (layout::print_line): Likewise.
14005 * edit-context.c (edited_file::print_content): Likewise.
14006 (edited_file::print_diff_hunk): Likewise.
14007 (edited_file::print_run_of_changed_lines): Likewise.
14008 (edited_file::get_num_lines): Likewise.
14009 (edited_line::edited_line): Likewise.
14010 * final.c (asm_show_source): Likewise.
14011 * input.c (location_get_source_line): Convert return type
14012 from const char * to char_span, losing the final "line_len"
14014 (dump_location_info): Update for the above.
14015 (get_substring_ranges_for_loc): Likewise. Use a char_span
14016 when handling the literal within the line.
14017 (test_reading_source_line): Update for location_get_source_line
14018 returning a char_span.
14019 * input.h (class char_span): Move here from
14020 diagnostic-show-locus.c, converting from a struct to a class.
14021 Make data members private.
14022 (char_span::operator bool): New.
14023 (char_span::length): New.
14024 (char_span::get_buffer): New.
14025 (char_span::operator[]): New.
14026 (char_span::subspan): Make const.
14027 (char_span::xstrdup): New.
14028 (location_get_source_line): Convert return type from const char *
14029 to char_span, losing the final "line_size" param.
14031 2018-04-30 Jan Hubicka <jh@suse.cz>
14033 * lto-wrapper.c (ltrans_priorities): New static var.
14034 (cmp_priority): New.
14035 (run_gcc): Read priorities and if doing parallel build order
14036 the Makefile by them.
14038 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14040 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14042 2018-04-30 Richard Biener <rguenther@suse.de>
14044 * tree-cfg.c (verify_address): Remove base argument, add
14045 flag whether to check TREE_ADDRESSABLE and do that.
14046 (verify_expr): Remove.
14047 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14048 (verify_gimple_assign_single): Likewise.
14049 (verify_gimple_switch): Likewise.
14050 (verify_expr_location_1): Dereference tp once. Add (disabled)
14051 piece from verify_expr.
14052 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14054 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14056 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14058 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14060 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14061 (small_data_pattern): Likewise.
14062 (arc_rewrite_small_data): Likewise.
14063 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14064 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14065 (get_symbol_alignment): New function.
14066 (legitimate_small_data_address_p): Likewise.
14067 (legitimate_scaled_address): Update, call
14068 legitimate_small_data_address_p.
14069 (output_sdata): New static variable.
14070 (arc_print_operand): Update how we handle small data operands.
14071 (arc_print_operand_address): Likewise.
14072 (arc_legitimate_address_p): Update, use
14073 legitimate_small_data_address_p.
14074 (arc_rewrite_small_data_p): Remove.
14075 (arc_rewrite_small_data_1): Likewise.
14076 (arc_rewrite_small_data): Likewise.
14077 (small_data_pattern): Likewise.
14078 (compact_sda_memory_operand): Update to use
14079 legitimate_small_data_address_p and get_symbol_alignment.
14080 (prepare_move_operands): Don't rewite sdata pattern.
14081 (prepare_extend_operands): Remove.
14082 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14084 (zero_extendqisi2): Likewise.
14085 (zero_extendhisi2): Likewise.
14086 (extendqihi2): Likewise.
14087 (extendqisi2): Likewise.
14088 (extendhisi2): Likewise.
14089 (addsi3): Likewise.
14090 (subsi3): Likewise.
14091 (andsi3): Likewise.
14092 * config/arc/constraints.md (Usd): Change it to memory constraint.
14094 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14096 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14097 as source of std instructions.
14098 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14099 allow 6-bit constants as source for store instructions.
14100 (movdi_insn): Update instruction pattern to allow 6-bit constants
14101 as source for store instructions.
14103 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14105 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14107 2018-04-30 Nathan Sidwell <nathan@acm.org>
14108 Sandra Loosemore <sandra@codesourcery.com>
14110 * dumpfile.c (dump_open): Allow '-' for stdout.
14111 * doc/invoke.texi (Developer Options): Document dump filename
14112 determination early. Document stdin/stdout selection.
14114 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14116 Microblaze Target: PIC data text relative
14118 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14119 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14121 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14122 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14123 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14124 New addressing mode for data-text relative position indepenedent code.
14125 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14126 'ADDRESS_SYMBOLIC_TXT_REL'.
14127 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14128 (microblaze_legitimate_pic_operand): Exclude function calls from
14129 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14130 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14132 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14133 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14134 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14135 for 'address + offset'.
14136 (microblaze_expand_prologue): Add new function prologue call for
14138 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14139 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14140 table in case of TARGET_PIC_DATA_TEXT_REL.
14141 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14142 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14143 Add new macros 'UNSPEC_TEXT',
14144 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14145 + exclude function calls from 'UNSPEC_PLT' in case of data text
14147 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14148 new target hook for generating address diff vector tables in case of
14150 * doc/tm.texi : Regenerate.
14151 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14152 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14153 of addr diff vector generation.
14154 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14155 target hook definition.
14156 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14157 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14158 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14159 option for data text relative.
14161 2018-04-30 Richard Biener <rguenther@suse.de>
14163 * tree-chrec.h (evolution_function_is_constant_p): Remove
14165 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14167 2018-04-30 Richard Biener <rguenther@suse.de>
14170 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14172 2018-04-30 Richard Biener <rguenther@suse.de>
14174 PR tree-optimization/28364
14175 PR tree-optimization/85275
14176 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14177 copying first exit test.
14179 2018-04-28 Mark Wielaard <mark@klomp.org>
14181 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14182 dwarf_version >= 5.
14183 (dwarf_AT): Handle DW_AT_addr_base.
14184 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14186 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14189 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14190 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14191 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14192 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14194 2018-04-28 Richard Biener <rguenther@suse.de>
14196 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14197 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14198 to reflect use. Only add interesting stmts.
14200 2018-04-27 Martin Jambor <mjambor@suse.cz>
14203 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14204 the jump function allows for passing through aggregate values.
14206 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14208 * input.h (in_system_header_at): Convert from macro to inline
14210 (from_macro_expansion_at): Likewise.
14211 (from_macro_definition_at): Likewise.
14213 2018-04-27 Jeff Law <law@redhat.com>
14215 * config.gcc: Mark tile* targets as deprecated/obsolete.
14217 2018-04-27 Richard Biener <rguenther@suse.de>
14219 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14222 2018-04-27 Richard Biener <rguenther@suse.de>
14224 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14226 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14228 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14229 with Yd constraint. Set "preferred_for_speed" attribute from
14230 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14231 with Yd constraint.
14232 (*movdi_internal): Ditto.
14233 (movti_interunit splitters): Remove
14234 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14235 (movdi_interunit splitters): Ditto.
14236 * config/i386/constraints.md (Ye): Remove.
14237 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14239 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14242 * config/aarch64/constraints.md (Usg): Limit to 31.
14243 (Usj): Limit to 63.
14245 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14247 PR tree-optimization/85529
14248 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14249 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14250 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14251 zero extension or masking of the MSB bit.
14252 (optimize_range_tests): Add FIRST_BB argument, pass it through
14253 to optimize_range_tests_var_bound.
14254 (maybe_optimize_range_tests, reassociate_bb): Adjust
14255 optimize_range_tests callers.
14257 2018-04-26 Richard Biener <rguenther@suse.de>
14258 Jakub Jelinek <jakub@redhat.com>
14260 * cgraph.h (symbol_table): Just declare debug method here.
14261 * symtab.c (symbol_table::debug): Define.
14263 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14265 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14267 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14269 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14270 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14271 (*movdi_internal): Substitute Yi and Yj constraint with x
14272 and Ym and Yn constraint with y constraint. Update "isa"
14273 attribute and set "preferred_for_speed" attribute from
14274 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14275 (*movsi_internal): Ditto.
14276 (*movdf_internal): Ditto.
14277 (*movsf_internal): Ditto.
14278 (*zero_extendsidi2): Ditto.
14279 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14280 (sse2_loadld): Ditto.
14281 (*vec_extract<ssevecmodelower>_0): Ditto.
14282 (*vec_extractv4si_0_zext_sse4): Ditto.
14283 (vec_concatv2di): Ditto.
14284 (*vec_dup<mode>): Ditto.
14285 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14286 * config/i386/constraints.md (Yi): Remove.
14291 2018-04-26 Nathan Sidwell <nathan@acm.org>
14293 * dumpfile.c (dump_open): New.
14294 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14295 (dump_finish): Detect stdio/stderr by value not name.
14297 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14299 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14301 2018-04-26 Tom de Vries <tom@codesourcery.com>
14304 * config/nvptx/nvptx.c (verify_neutering_jumps)
14305 (verify_neutering_labels): New function
14306 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14308 2018-04-26 Tom de Vries <tom@codesourcery.com>
14311 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14312 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14315 2018-04-26 Richard Biener <rguenther@suse.de>
14316 Tom de Vries <tom@codesourcery.com>
14319 * lto-streamer-out.c (output_function): Fixup loops if required to match
14320 discovery done in the reader.
14322 2018-04-26 Richard Biener <rguenther@suse.de>
14324 PR tree-optimization/85116
14325 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14326 have a loop exit from the single latch predecessor. Remove
14327 case of header with just condition.
14328 (ch_base::copy_headers): Exclude infinite loops from any
14330 (pass_ch::execute): Record exits.
14332 2018-04-26 Richard Biener <rguenther@suse.de>
14334 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14335 prologue cost vector and pass it to vect_get_load_cost.
14336 (vect_get_peeling_costs_all_drs): Likewise.
14337 (vect_peeling_hash_get_lowest_cost): Likewise.
14338 (vect_enhance_data_refs_alignment): Likewise.
14340 2018-04-26 Richard Biener <rguenther@suse.de>
14342 PR middle-end/85450
14343 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14344 checking of integer<->pointer conversions.
14345 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14346 sign-/zero-extending pointer types.
14347 (expand_omp_for_static_chunk): Likewise.
14349 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14350 Jean Lee <xiaoyur347@gmail.com>
14352 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14353 (TARGET_ASAN_SHADOW_OFFSET): Define.
14354 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14355 true for -fsanitize=address.
14357 2018-04-25 Mark Wielaard <mark@klomp.org>
14359 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14362 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14364 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14365 than "alu", remove explicit "memory" and "imm_disp" attributes.
14366 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14368 PR middle-end/85414
14369 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14370 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14371 gen_lowpart_no_emit.
14373 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14376 * config/i386/i386.c (ix86_expand_builtin): Change memory
14377 operand to XI, extend p0 to Pmode.
14378 * config/i386/i386.md: Change unspec volatile and operand
14379 1 mode to XI, change operand 0 mode to P.
14381 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14383 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14384 GET_MODE_MASK before any checking.
14385 (nds32_can_use_bset_p): Likewise.
14386 (nds32_can_use_btgl_p): Likewise.
14388 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14390 * config/nds32/nds32-doubleword.md: New define_split pattern for
14391 illegal register number.
14393 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14395 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14397 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14399 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14401 2018-04-25 Richard Biener <rguenther@suse.de>
14403 * lto-streamer.h (LTO_major_version): Bump to 8.
14405 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14407 * BASE-VER: Set to 9.0.0.
14409 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14411 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14412 in __abskf2 and __powikf2.
14414 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14417 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14418 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14419 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14420 Use the above on operand 2. Reindent.
14421 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14423 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14426 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14428 * config/i386/i386.opt (mcet): Removed.
14429 * doc/install.texi: Remove -mcet documentation.
14430 * doc/invoke.texi: Likewise.
14432 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14435 * doc/install.texi: Remove -mcet from bootstrap-cet.
14437 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14440 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14441 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14445 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14446 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14447 containing a CONST_VECTOR.
14449 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14451 * doc/install.texi: Update newlib dependency for nvptx.
14453 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14456 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14457 instead of INTVAL when shifting x left.
14459 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14461 PR tree-optimization/85478
14462 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14463 vect_grouped_store_supported for single element vectors.
14465 2018-04-24 Richard Biener <rguenther@suse.de>
14468 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14469 load cost increase to the case of non-constant step.
14471 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14474 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14475 destination if any_malformed_asm.
14477 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14479 PR middle-end/85496
14480 * expr.c (store_field): In the bitfield case, if the value comes from
14481 a function call and is returned in registers by means of a PARALLEL,
14482 do not change the mode of the temporary unless BLKmode and VOIDmode.
14484 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14486 PR rtl-optimization/85423
14487 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14488 dependencies to debug insns when the previous insn is non-debug.
14490 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14492 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14493 enums into a single definition.
14494 (fls): Fix predicates and printing.
14497 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14499 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14500 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14501 and short u6 immediate.
14502 (check_if_valid_sleep_operand): Remove.
14503 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14505 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14507 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14508 flag_always_save_lp condition.
14509 * config/nds32/nds32.opt (malways-save-lp): New option.
14511 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14513 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14514 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14515 * config/nds32/nds32.h
14516 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14517 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14519 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14521 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14522 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14524 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14525 Chung-Ju Wu <jasonwucj@gmail.com>
14527 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14529 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14530 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14532 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14534 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14536 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14538 * config/nds32/nds32-protos.h (nds32_data_alignment,
14539 nds32_local_alignment): Declare.
14540 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14541 nds32_local_alignment): New functions.
14542 (TARGET_CONSTANT_ALIGNMENT): Define.
14543 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14545 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14547 * config/nds32/nds32.c
14548 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14549 (TARGET_MODES_TIEABLE_P): Likewise.
14551 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14553 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14554 level Ofast and Og.
14556 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14557 Chung-Ju Wu <jasonwucj@gmail.com>
14559 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14560 for unaligned access.
14561 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14563 * config/nds32/nds32-intrinsic.md: Likewise.
14564 * config/nds32/nds32_intrinsic.h: Likewise.
14565 * config/nds32/nds32.h (nds32_builtins): Likewise.
14566 * config/nds32/nds32.opt (munaligned-access): New option.
14567 * config/nds32/nds32.c (nds32_asm_file_start): Display
14568 flag_unaligned_access status.
14570 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14572 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14573 -mno-relax is present.
14574 * config/riscv/linux.h (LINK_SPEC): Ditto.
14576 2018-04-20 Martin Sebor <msebor@redhat.com>
14579 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14581 (gimple_fold_builtin_stxcpy_chk): Same.
14582 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14584 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14587 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14588 __powikf2 when long double is IEEE 128-bit.
14590 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14592 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14593 step to make sure stack always aligned.
14595 2018-04-20 Carl Love <cel@us.ibm.com>
14598 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14599 size check for arg0.
14601 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14602 Tom de Vries <tom@codesourcery.com>
14605 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14606 Emit insns for calls too.
14607 (nvptx_find_par): Always look for worker-level predecessor insn.
14608 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14610 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14611 (nvptx_process_pars): Propagate frames for calls.
14613 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14616 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14618 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14619 (ix86_handle_option): Don't handle OPT_mibt.
14620 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14622 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14624 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14625 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14626 (ix86_target_macros): Define __CET__ with flag_cf_protection
14627 for -fcf-protection.
14628 * config/i386/i386.c (isa2_opts): Remove -mibt.
14629 * config/i386/i386.h (TARGET_IBT): Removed.
14630 (TARGET_IBT_P): Likewise.
14631 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14632 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14633 * config/i386/i386.opt (mcet): Update help message.
14634 (mshstk): Likewise.
14636 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
14637 -mcet as an alias for -mshstk.
14639 2018-04-20 Richard Biener <rguenther@suse.de>
14641 PR middle-end/85475
14642 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14643 complexity by forcing a single use of the multiply operand.
14645 2018-04-20 Martin Jambor <mjambor@suse.cz>
14648 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14649 recursion dependency to only apply to non-clones.
14651 2018-04-20 Martin Jambor <mjambor@suse.cz>
14654 * ipa-cp.c (create_specialized_node): Check that clones of
14655 self-recursive edges exist during IPA-CP.
14657 2018-04-19 Toon Moene <toon@moene.org>
14659 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14662 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14664 PR tree-optimization/85467
14665 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14666 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
14667 VECTOR_CST element to type.
14669 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14672 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
14673 * config/i386/i386.md (builtin_setjmp_setup): Removed.
14674 (builtin_longjmp): Likewise.
14675 (save_stack_nonlocal): New pattern.
14676 (restore_stack_nonlocal): Likewise.
14678 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14681 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14682 Replace ASM_OUTPUT_LABEL with fprintf.
14684 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14687 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14688 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
14689 * config/i386/i386-c.c (ix86_target_macros_internal): Also
14690 define __IBT__ and __SHSTK__ for -fcf-protection.
14691 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
14693 (ix86_trampoline_init): Likewise.
14694 (x86_output_mi_thunk): Likewise.
14695 (ix86_notrack_prefixed_insn_p): Likewise.
14696 (ix86_option_override_internal): Don't disallow -fcf-protection.
14697 * config/i386/i386.md (rdssp<mode>): Also enable for
14699 (incssp<mode>): Likewise.
14700 (nop_endbr): Likewise.
14701 * config/i386/i386.opt (mcet): Change help message to built-in
14704 (mshstk): Likewise.
14705 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
14706 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
14707 enable CET built-in functions.
14709 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
14711 * common/config/i386/i386-common.c
14712 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
14713 OPTION_MASK_ISA_MOVDIRI_UNSET,
14714 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
14715 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
14716 * config.gcc (movdirintrin.h): New header.
14717 * config/i386/cpuid.h (bit_MOVDIRI,
14718 bit_MOVDIR64B): New bits.
14719 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
14721 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
14722 (VOID, PVOID, PCVOID)): New function types.
14723 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
14724 __builtin_ia32_directstoreu_u64,
14725 __builtin_ia32_movdir64b): New builtins.
14726 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
14727 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
14729 (ix86_valid_target_attribute_inner_p): Ditto.
14730 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
14731 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
14732 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
14733 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
14734 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
14735 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
14736 (movdiri<mode>, movdir64b_<mode>): New.
14737 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
14738 * config/i386/immintrin.h: Include movdirintrin.h.
14739 * config/i386/movdirintrin.h: New file.
14740 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
14742 2018-04-19 Richard Biener <rguenther@suse.de>
14744 PR middle-end/85455
14745 * cfg.c (clear_bb_flags): When loop state says we have
14746 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
14748 2018-04-19 Richard Biener <rguenther@suse.de>
14750 PR tree-optimization/84737
14751 * tree-vect-data-refs.c (vect_copy_ref_info): New function
14752 copying restrict info.
14753 (vect_setup_realignment): Use it.
14754 * tree-vectorizer.h (vect_copy_ref_info): Declare.
14755 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
14756 the first DR to all generated stores.
14757 (vectorizable_load): Likewise for loads.
14759 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14761 PR tree-optimization/85446
14762 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
14763 the integral and pointer types to have the same precision.
14765 * doc/install.texi: Document --disable-cet being the default and
14768 2018-04-18 Martin Liska <mliska@suse.cz>
14770 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14773 2018-04-18 Martin Liska <mliska@suse.cz>
14776 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14779 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14780 arguments if they are comparable.
14782 2018-04-18 Martin Liska <mliska@suse.cz>
14785 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14788 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14791 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
14794 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14795 ENDBR after calling __morestack.
14797 2018-04-18 David Malcolm <dmalcolm@redhat.com>
14800 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14801 by using gcc_base_ver to generate a gcc_driver_version, and use
14802 it when generating GCC_DRIVER_NAME.
14803 * configure: Regenerate.
14805 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14808 * config.gcc: Obsolete powerpc*-*-*spe*.
14810 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14813 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14814 (stabstr_D): Change type of unum from unsigned int to
14815 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14818 2018-04-17 Jim Wilson <jimw@sifive.com>
14821 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14822 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14823 Set arg_pointer_offset after using pretend_args_size.
14825 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14827 PR rtl-optimization/85431
14828 * dse.c (record_store): Ignore zero width stores.
14831 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14832 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14833 __builtin_stack_restore rather than after it.
14834 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14835 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14836 argument instead of virtual_dynamic_stack_rtx.
14838 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14840 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14842 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14843 Add note to error message to explain internal mapping of overloaded
14844 built-in function name to non-overloaded built-in function name.
14845 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14848 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14851 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14852 where the inputs overlap with the output.
14854 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14856 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14857 (=v, v) alternative and explicit "memory" attribute.
14858 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14859 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14861 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14862 "sselog1" type instead of "sselog".
14863 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14864 "sselog". Remove explicit "memory" attribute.
14865 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14866 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14868 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14869 "sselog1" type instead of "sselog". Remove explicit "memory"
14871 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14872 use "sselog1" type instead of "sselog". Remove explicit "memory"
14874 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14875 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14877 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14878 "sselog1" type instead of "sselog". Remove explicit "memory"
14880 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14881 use "sselog1" type instead of "sselog". Remove explicit "memory"
14885 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14887 PR middle-end/85414
14888 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14891 2018-04-17 Martin Jambor <mjambor@suse.cz>
14894 * ipa-cp.c (create_specialized_node): Call
14895 expand_all_artificial_thunks if necessary.
14897 2018-04-17 Martin Liska <mliska@suse.cz>
14900 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14901 in message, remote space in between '_G' and '('.
14903 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14906 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14907 avx512f_vmcmp<mode>3<round_saeonly_name>,
14908 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14909 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14910 avx512f_rndscale<mode><round_saeonly_name>,
14911 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14912 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14913 Use %<iptr>2 instead of %2 for -masm=intel.
14914 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14915 avx512f_vcvttss2usi<round_saeonly_name>,
14916 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14918 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14919 avx512f_vcvttsd2usi<round_saeonly_name>,
14920 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14921 Use %q1 instead of %1 for -masm=intel.
14922 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14923 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14924 of %3 for -masm=intel.
14925 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14927 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14929 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14931 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14932 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14934 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14936 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14937 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14938 %g1 and one with %0 and %1.
14939 (avx512er_vmrcp28<mode><round_saeonly_name>,
14940 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14941 %1 for -masm=intel.
14942 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14943 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14944 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14945 of %0 and %{%4%} for -masm=intel.
14946 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14947 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14948 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14949 order of %0 and %{%5%}%{z%} for -masm=intel.
14951 2018-04-17 Jan Hubicka <jh@suse.cz>
14954 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14956 2018-04-17 Martin Liska <mliska@suse.cz>
14959 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14960 for target_clone error message. Make default implementation
14961 clone to be a local declaration.
14962 (separate_attrs): Add new argument and check for an empty
14964 (expand_target_clones): Handle it.
14965 (ipa_target_clone): Make redirection just for target_clones
14968 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14969 Tom de Vries <tom@codesourcery.com>
14971 PR middle-end/84955
14972 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14973 tiled basic blocks without omp continue statements.
14975 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14978 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14979 vec_extract expression as having side effects to make sure it gets
14982 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14985 * config/i386/i386.c (get_builtin_code_for_version): Check
14988 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14991 * config.gcc: Support "skylake".
14992 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14994 * config/i386/i386.c (m_SKYLAKE): Define.
14995 (processor_target_table): Add "skylake".
14996 (ix86_option_override_internal): Add "skylake".
14997 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14998 PROCESSOR_CANNONLAKE.
14999 (get_builtin_code_for_version): Fix priority for
15000 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15001 PROCESSOR_SKYLAKE-AVX512.
15002 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15003 (processor_type): Add PROCESSOR_SKYLAKE.
15005 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15006 Jason Merrill <jason@redhat.com>
15009 * convert.c (convert_to_integer_1): Use direct recursion for
15010 enumeral types and types with a precision less than the number
15011 of bits in their mode.
15013 2018-04-16 Julia Koval <julia.koval@intel.com>
15016 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15017 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15019 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15022 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15023 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15024 and -mno-direct-move.
15026 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15029 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15030 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15031 Ensure negative shifts result in {0}.
15033 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15035 PR rtl-optimization/79916
15036 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15037 regs (if any) to define how to gnerate SD moves when LRA is in
15040 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15042 PR rtl-optimization/85393
15043 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15044 * except.c (expand_dw2_landing_pad_for_region): Make static.
15045 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15046 a label and unconditional jump to old_bb, rather than
15047 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15050 PR rtl-optimization/85376
15051 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15052 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15053 instead of a specific value.
15055 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15056 Bin Cheng <bin.cheng@arm.com>
15058 PR tree-optimization/82965
15059 PR tree-optimization/83991
15060 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15061 by_profile_only parameter.
15062 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15063 information if the loop was predicted to iterate too many times.
15064 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15066 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15069 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15072 2018-04-13 Martin Liska <mliska@suse.cz>
15073 Jakub Jelinek <jakub@redhat.com>
15075 PR middle-end/81657
15076 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15077 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15078 * builtins.c (expand_builtin_memory_copy_args): Use
15079 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15080 handle dest_addr == pc_rtx.
15082 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15085 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15086 asked to not generate direct moves.
15087 (fix_trunc<mode>si2_stfiwx): Similar.
15088 (fix_trunc<mode>si2_internal): Similar.
15090 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15093 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15094 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15095 lookup if dest in some wider mode is known to be const0_rtx and
15096 if so, record permanent equivalence for it to be ZERO_EXTEND of
15097 the narrower mode destination.
15099 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15101 * lto-streamer-out.c (output_function): Revert 259346.
15102 * omp-expand.c (expand_oacc_for): Likewise.
15104 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15106 PR rtl-optimization/85354
15107 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15108 * sel-sched.c (sel_global_init): ... here.
15110 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15113 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15114 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15115 mode for PE-COFF targets.
15116 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15117 (i386_pe_asm_lto_end): Likewise.
15118 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15119 (TARGET_ASM_LTO_END): Likewise.
15120 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15121 (i386_pe_asm_lto_start): New function.
15122 (i386_pe_asm_lto_end): Likewise.
15124 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15125 Richard Biener <rguenther@suse.de>
15127 PR middle-end/84955
15128 * lto-streamer-out.c (output_function): Fix CFG loop state before
15130 * omp-expand.c (expand_oacc_for): Handle calls to internal
15131 functions like regular functions.
15133 2018-04-12 Richard Biener <rguenther@suse.de>
15136 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15137 for the early LTO debug to properly generate references to it
15138 during DIE emission. Do not re-use that for the skeleton for
15140 (dwarf2out_early_finish): Likewise.
15142 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15145 * config/i386/sse.md
15146 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15147 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15148 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15149 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15150 and output is a reg, avoid creating invalid lowpart subreg, but
15151 instead split into a 512-bit move. Don't split if not AVX512VL,
15152 input is xmm16+ reg and output is a mem.
15153 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15154 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15155 xmm16+ reg and output is a mem.
15157 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15159 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15160 also for flag_dwarf2_cfi_asm.
15162 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15164 PR rtl-optimization/85342
15165 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15166 a bool scalar var inside of the loop instead. Don't try to update
15167 recog_data.operand after failed apply_change_group.
15169 2018-04-12 Tom de Vries <tom@codesourcery.com>
15172 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15173 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15174 array with flexible array member as array without given dimension.
15175 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15176 argument for undefined param to true.
15178 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15181 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15182 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15183 from PowerPC section.
15184 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15185 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15186 help text that is too long.
15187 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15188 help text that is too long.
15189 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15190 help text that is too long.
15192 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15194 * config/alpha/alpha.md (stack_probe_internal): Rename
15195 from "probe_stack". Update all callers.
15197 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15199 PR rtl-optimization/84566
15200 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15201 sched_macro_fuse_insns.
15203 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15206 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15207 (compute_block_dependences): ... from here.
15209 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15211 PR tree-optimization/85331
15212 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15213 from int to HOST_WIDE_INT.
15215 2018-04-11 Martin Jambor <mjambor@suse.cz>
15218 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15219 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15220 not the same as the source val.
15221 (cgraph_edge_brings_value_p): New parameter.
15222 (gather_edges_for_value): Pass destination value to
15223 cgraph_edge_brings_value_p.
15224 (perhaps_add_new_callers): Likewise.
15225 (get_info_about_necessary_edges): Likewise and exclude values brought
15226 only by self-recursive edges.
15227 (create_specialized_node): Redirect only clones of self-calling edges.
15228 (+self_recursive_pass_through_p): New function.
15229 (find_more_scalar_values_for_callers_subset): Use it.
15230 (find_aggregate_values_for_callers_subset): Likewise.
15231 (known_aggs_to_agg_replacement_list): Removed.
15232 (decide_whether_version_node): Re-calculate known constants for all
15233 remaining context clones.
15235 2018-04-11 Richard Biener <rguenther@suse.de>
15238 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15239 from early DWARF output.
15240 (dwarf2out_early_finish): Output line info unconditionally into
15241 early DWARF and add reference to it.
15243 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15246 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15247 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15248 other than V2DFmode using iptr mode attribute.
15249 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15251 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15253 PR rtl-optimization/84659
15254 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15256 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15259 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15261 (output_loc_list): Pass address of a dummy size variable even in the
15262 locview handling loop.
15263 (index_location_lists): Add comment on why skip_loc_list_entry can't
15266 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15269 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15272 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15275 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15276 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15277 and -mstring-compare-inline-limit.
15279 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15282 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15283 for stack clash protection in a register whenever we need it to be in
15286 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15288 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15289 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15291 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15294 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15296 (mlong-double-): Ditto.
15297 * config/rs6000/sysv4.opt (msdata=): Ditto.
15298 (mtls-size=): Ditto.
15300 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15302 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15303 erroneous entries for
15304 "vector int vec_ldl (int, long int *)", and
15305 "vector unsigned int vec_ldl (int, unsigned long int *)".
15306 Add comments and entries for
15307 "vector bool char vec_ldl (int, bool char *)",
15308 "vector bool short vec_ldl (int, bool short *)",
15309 "vector bool int vec_ldl (int, bool int *)",
15310 "vector bool long long vec_ldl (int, bool long long *)",
15311 "vector pixel vec_ldl (int, pixel *)",
15312 "vector long long vec_ldl (int, long long *)",
15313 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15314 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15315 type tree bool_long_long_type_node and correct definition of
15316 bool_V2DI_type_node to make reference to this new type tree.
15317 (rs6000_mangle_type): Replace erroneous reference to
15318 bool_long_type_node with bool_long_long_type_node.
15319 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15320 comments to emphasize sign distinctions for char and int types and
15321 replace RS6000_BTI_bool_long constant with
15322 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15323 use of RS6000_BTI_pixel.
15324 (bool_long_type_node): Remove this macro definition.
15325 (bool_long_long_type_node): New macro definition
15327 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15329 PR rtl-optimization/85300
15330 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15331 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15332 simplify_unary_operation fails.
15334 2018-04-10 Martin Liska <mliska@suse.cz>
15336 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15337 cgraph_edge and ipa_ref.
15339 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15343 * config/i386/sse.md
15344 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15345 computation of the VEC_MERGE selector from mask.
15346 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15347 Fix decoding of the VEC_MERGE selector into mask.
15349 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15351 PR tree-optimization/85286
15352 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15354 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15356 * final.c (final_1): Set insn_last_address as well as
15357 insn_current_address.
15359 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15362 * explow.c (emit_stack_probe): Call validize_mem on memory location
15363 before passing it to gen_probe_stack. Create address operand and
15364 legitimize it for the probe_stack_address case.
15366 2018-04-09 Jan Hubicka <jh@suse.cz>
15369 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15370 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15371 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15373 (free_lang_data): Rebuild type inheritance graph.
15375 2018-04-09 Martin Sebor <msebor@redhat.com>
15377 * invoke.texi (-finline-small-functions): Mention other optimization
15379 (-findirect-inlining, -fpartial-inlining): Same.
15380 (-finline-functions-called-once): Same.
15381 (-freorder-blocks-and-partition): Same.
15383 2018-04-09 Jan Hubicka <jh@suse.cz>
15386 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15387 jumps; choose last target that matches the criteria (i.e.
15388 no partition changes for non-crossing jumps).
15389 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15390 support for redirecting crossing jumps to non-crossing.
15392 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15394 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15395 also for naked functions.
15397 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15399 * config/arc/arc.md (add_shift): New pattern.
15400 (add_shift2): Likewise.
15401 (sub_shift): Likewise.
15402 (sub_shift_cmp0_noout): Likewise.
15403 (compare_si_ashiftsi): Likewise.
15404 (xbfu_cmp0_noout): New combine pattern.
15405 (xbfu_cmp0"): Likewise.
15406 (movsi_set_cc_insn): Place the predicable variant first.
15407 (commutative_binary_cmp0_noout): Remove clobber.
15408 (commutative_binary_cmp0): New pattern.
15409 (noncommutative_binary_cmp0): Likewise.
15410 (noncommutative_binary_cmp0_noout): Likewise.
15411 (noncommutative_binary_comparison_result_used): Removed.
15412 (rsub_cmp0): New pattern.
15413 (rsub_cmp0_noout): Likewise.
15414 (extzvsi): Changed, keep only meaningful variants.
15415 (SQH, SEZ): New iterators.
15416 (SQH_postfix): New mode attribute.
15417 (SEZ_prefix): New code attribute.
15418 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15419 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15420 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15421 of numerical value.
15422 (noncommutative_operator): Check the availability of barrel
15425 2018-04-09 Richard Biener <rguenther@suse.de>
15427 PR tree-optimization/85284
15428 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15429 Only use the niter constraining form of simple_iv when the exit
15430 is always executed.
15432 2018-04-09 Tom de Vries <tom@codesourcery.com>
15435 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15436 (define_expand "*memory_barrier"): New define_expand.
15437 (define_insn "memory_barrier"): New insn.
15439 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15441 PR rtl-optimization/80463
15442 PR rtl-optimization/83972
15443 PR rtl-optimization/83480
15445 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15446 correct producer for the insn.
15447 (tidy_control_flow): Fixup seqnos in case of debug insns.
15449 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15451 PR rtl-optimization/83913
15453 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15454 different sched-times when merging exprs.
15456 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15458 PR rtl-optimization/83962
15460 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15461 tidy_fallthru_edge and tidy_control_flow.
15463 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15465 PR rtl-optimization/83530
15467 * sel-sched.c (force_next_insn): New global variable.
15468 (remove_insn_for_debug): When force_next_insn is true, also leave only
15469 next insn in the ready list.
15470 (sel_sched_region): When the region wasn't scheduled, make another pass
15471 over it with force_next_insn set to 1.
15473 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15475 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15477 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15478 for interrupt control.
15479 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15480 functions for interrupt control.
15481 * config/nds32/nds32-intrinsic.md: Likewise.
15482 * config/nds32/nds32_intrinsic.h: Likewise.
15483 * config/nds32/nds32.h (nds32_builtins): Likewise.
15485 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15487 * config/nds32/nds32.c (nds32_init_machine_status,
15488 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15489 strict_aligned_p field.
15490 (nds32_expand_to_rtl_hook): New function.
15491 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15492 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15494 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15495 Chung-Ju Wu <jasonwucj@gmail.com>
15497 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15498 * config/nds32/nds32-n7.md: New file.
15499 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15500 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15502 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15503 * config/nds32/nds32.md (pipeline_model): Add n7.
15504 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15505 * config/nds32/pipelines.md: Include n7 settings.
15507 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15508 Chung-Ju Wu <jasonwucj@gmail.com>
15510 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15511 * config/nds32/nds32-e8.md: New file.
15512 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15513 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15515 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15516 * config/nds32/nds32.md (pipeline_model): Add e8.
15517 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15518 * config/nds32/pipelines.md: Include e8 settings.
15520 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15521 Chung-Ju Wu <jasonwucj@gmail.com>
15523 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15524 * config/nds32/nds32-n8.md: New file.
15525 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15526 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15528 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15529 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15530 * config/nds32/nds32.md (pipeline_model): Add n8.
15531 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15532 * config/nds32/pipelines.md: Include n8 settings.
15534 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15535 Chung-Ju Wu <jasonwucj@gmail.com>
15537 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15538 * config/nds32/nds32-n9-2r1w.md: New file.
15539 * config/nds32/nds32-n9-3r2w.md: New file.
15540 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15541 nds32_register_ports): New or modify for cpu n9.
15542 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15544 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15545 * config/nds32/nds32-utils.c: New file.
15546 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15547 TARGET_MUL_SLOW): Define.
15548 * config/nds32/nds32.md (pipeline_model): New attribute.
15549 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15550 New options that support cpu n9.
15551 * config/nds32/pipelines.md: Include n9 settings.
15552 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15554 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15556 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15557 information if necessary.
15558 (output_cond_branch_compare_zero): Likewise.
15559 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15560 (nds32_target_alignment): Refine for alignment.
15561 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15562 (FUNCTION_BOUNDARY): Modify.
15563 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15565 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15567 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15569 * config/nds32/constants.md (unspec_volatile_element): Add values for
15570 TLB operation and data prefetch.
15571 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15572 functions for TLB operation and data prefetch.
15573 * config/nds32/nds32-intrinsic.md: Likewise.
15574 * config/nds32/nds32_intrinsic.h: Likewise.
15575 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15576 (nds32_print_operand): Likewise.
15577 * config/nds32/nds32.h (nds32_builtins): Likewise.
15579 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15580 Andrew Pinski <pinsika@gcc.gnu.org>
15582 PR middle-end/82976
15583 * match.pd: Use constant_boolean_node of correct type instead of
15584 boolean_true_node or boolean_false_node for simplifying
15585 pointer comparisons to zero.
15587 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15589 PR tree-optimization/80021
15590 * tree.c (verify_type_variant): Make error call in verify_variant_match
15591 translatable and remove final full stop.
15593 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15595 * config/nds32/constants.md (unspec_volatile_element): Add
15596 UNSPEC_VOLATILE_EH_RETURN.
15597 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15598 nds32_output_stack_pop): Support dwarf exception handling process.
15599 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15600 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15601 exception handling process.
15602 (nds32_compute_stack_frame): Likewise.
15603 (nds32_return_addr_rtx): Likewise.
15604 (nds32_initial_elimination_offset): Likewise.
15605 (nds32_expand_prologue): Likewise.
15606 (nds32_expand_epilogue): Likewise.
15607 (nds32_dynamic_chain_address): New function.
15608 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15609 exception handling.
15610 (DYNAMIC_CHAIN_ADDRESS): Define.
15611 (EH_RETURN_DATA_REGNO): Define.
15612 (EH_RETURN_STACKADJ_RTX): Define.
15613 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15614 patterns for dwarf exception handling.
15616 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15618 * config/nds32/nds32.h: Clean up obsolete macros.
15620 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15622 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15623 Add enum values for particular instructions.
15624 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15625 particular intrinsic functions.
15626 * config/nds32/nds32-intrinsic.md: Likewise.
15627 * config/nds32/nds32_intrinsic.h: Likewise.
15628 * config/nds32/nds32.h (nds32_builtins): Likewise.
15629 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15630 (btst, ave): New patterns for particular instructions.
15632 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15634 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15635 Add enum values for atomic load/store and memory sync.
15636 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15638 * config/nds32/nds32-intrinsic.md: Likewise.
15639 * config/nds32/nds32_intrinsic.h: Likewise.
15640 * config/nds32/nds32.h (nds32_builtins): Likewise.
15642 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15644 PR tree-optimization/85257
15645 * fold-const.c (native_encode_vector): If not all elts could fit
15646 and off is -1, return 0 rather than offset.
15647 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15648 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15649 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
15650 adjust buffer in native_interpret_expr call.
15652 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15654 * config/nds32/constants.md (unspec_volatile_element): Add cache
15655 control enum values.
15656 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15657 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15658 * config/nds32/nds32.c (nds32_cctl_names): New.
15659 (nds32_print_operand): Handle cache control register names.
15660 * config/nds32/nds32.h (nds32_builtins): New enum values.
15661 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15663 * config/nds32/nds32.md (type): Add mmu.
15664 * config/nds32/pipelines.md (simple_insn): Add mmu.
15666 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15668 * config/nds32/nds32.md (type): Remove call.
15669 * config/nds32/pipelines.md (simple_insn): Likewise.
15671 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15673 * config/nds32/constants.md (unspec_volatile_element): Add
15674 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
15675 UNSPEC_VOLATILE_FMFCFG.
15676 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
15677 description for fmfcfg and fmfcsr.
15678 (bdesc_1arg): Add fmtcsr.
15679 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
15680 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
15681 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
15682 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
15683 unspec_fmfcfg): New patterns.
15684 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
15685 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
15686 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
15687 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
15688 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
15689 __nds32__fmfcfg): Define.
15691 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15693 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
15694 intrinsic register names.
15695 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
15696 intrinsic register enum values and macros.
15698 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15700 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
15701 for load/store addressing form.
15702 (nds32_print_operand_address): Likewise.
15704 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
15707 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
15708 based on LABEL_REF. Remove useless assertion.
15709 (pic_address_needs_scratch): Fix formatting.
15710 (sparc_legitimize_pic_address): Minor tweaks.
15711 (sparc_delegitimize_address): Adjust assertion accordingly.
15712 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
15713 into symbolic_operand.
15714 (movsi_high_pic_label_ref): Likewise.
15715 (movsi_lo_sum_pic_label_ref): Likewise.
15716 (movdi_pic_label_ref): Likewise.
15717 (movdi_high_pic_label_ref): Likewise.
15718 (movdi_lo_sum_pic_label_ref): Likewise.
15720 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
15722 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
15723 custom LIB_SPEC setup.
15725 2018-04-06 Ruslan Bukin <br@bsdpad.com>
15726 Kito Cheng <kito.cheng@gmail.com>
15728 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
15729 * config/riscv/freebsd.h: New.
15731 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15733 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
15734 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
15737 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15738 Kito Cheng <kito.cheng@gmail.com>
15740 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
15741 nds32_output_call, nds32_symbol_binds_local_p): New functions.
15742 * config/nds32/nds32-protos.h (nds32_output_call,
15743 nds32_output_return): Declare.
15744 * config/nds32/nds32.md: Refine all the call and return patterns.
15746 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15749 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
15750 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
15752 PR rtl-optimization/84872
15753 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
15754 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
15755 EDGE_CROSSING edge.
15757 2018-04-06 Tamar Christina <tamar.christina@arm.com>
15759 * expr.c (copy_blkmode_to_reg): Revert 254862.
15760 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
15762 2018-04-06 Richard Biener <rguenther@suse.de>
15764 PR middle-end/85244
15765 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
15766 after seeing a component reference with an adjacent field. Treat
15767 refs to arrays at struct end of external decls similar to
15768 refs to unconstrained commons.
15770 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15773 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
15774 look through SAVE_EXPRs with non-side-effects argument. Adjust
15776 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
15779 2018-04-06 Richard Biener <rguenther@suse.de>
15781 PR middle-end/85180
15782 * alias.c (find_base_term): New wrapper around find_base_term
15783 unwinding CSELIB_VAL_PTR changes.
15784 (find_base_term): Do not restore CSELIB_VAL_PTR during the
15787 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15789 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15791 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15792 constant definitions.
15793 ("nop"): lr 0,0 -> nopr r0
15794 ("nop_lr0", "nop_lr1"): New insn definitions.
15796 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15798 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15799 NDS32_V3PUSH_AVAILABLE_P macro.
15801 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
15802 Chung-Ju Wu <jasonwucj@gmail.com>
15804 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15805 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15806 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15807 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15808 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15809 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15810 * config/nds32/constraints.md: New constraints and checking for hard
15811 float configuration.
15812 * config/nds32/iterators.md: New mode iterator and attribute for hard
15813 float configuration.
15814 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15816 * config/nds32/nds32-fpu.md: New file.
15817 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15818 deal with hard float code generation.
15819 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15821 (abi_type, float_reg_number): New enum type.
15822 * config/nds32/nds32-predicates.c: New predicates for hard float.
15823 * config/nds32/nds32-protos.h: Declare functions for hard float.
15824 * config/nds32/nds32.c: Implementation for hard float configuration.
15825 * config/nds32/nds32.h: Definitions for hard float configuration.
15826 * config/nds32/nds32.md: Include hard float machine description and
15827 modify patterns for hard float configuration.
15828 * config/nds32/nds32.opt: New options for hard float configuration.
15829 * config/nds32/predicates.md: New predicates for hard float
15832 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15834 * common/config/nds32/nds32-common.c
15835 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15837 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15839 PR middle-end/85195
15840 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15841 CONSTRUCTOR_ELT (ctor, ...)->value.
15843 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15846 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15848 2018-04-05 Tom de Vries <tom@codesourcery.com>
15851 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15854 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15855 Kito Cheng <kito.cheng@gmail.com>
15857 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15858 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15859 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15860 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15862 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15863 Kito Cheng <kito.cheng@gmail.com>
15865 * config/nds32/constraints.md (Ufe): New memory constraint.
15866 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15867 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15868 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15870 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15871 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15873 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15875 * config/nds32/nds32.md: Use optimize_size in the condition for
15876 alu-shift instructions.
15878 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15880 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15882 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15884 * config/nds32/nds32.md (negsi2): Refine pattern.
15886 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15887 Chung-Ju Wu <jasonwucj@gmail.com>
15889 * config/nds32/iterators.md (shift_rotate): New code iterator.
15890 (shift): New code attribute.
15891 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15892 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15893 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15894 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15895 bit-wise operations.
15896 (andsi3, *andsi3): Ditto.
15897 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15898 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15899 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15900 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15901 nds32_ior_operand, nds32_xor_operand): New predicates.
15903 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15905 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15906 (addsi3, subsi3): ... this.
15908 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15910 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15912 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15914 * config/nds32/nds32.md: Adjust indention.
15916 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15918 * config/nds32/nds32.md (feature): New attribute.
15920 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15922 * config/nds32/nds32.md (subtype): New attribute.
15924 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15927 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15928 expansion to perform a bitwise AND of the argument followed by a
15929 boolean negation of the result.
15931 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15933 PR rtl-optimization/84878
15934 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15935 the basic block. Assert the use reference is not artificial and that
15936 it has an associated insn.
15938 2018-04-04 Michael Matz <matz@suse.de>
15940 * builtins.c (compute_objsize): Pass correct operand
15941 to array_at_struct_end_p.
15943 2018-04-04 Richard Biener <rguenther@suse.de>
15946 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15947 from contexts for DINFO_LEVEL_TERSE and below.
15949 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15951 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15952 resiter_operand condition.
15953 * config/nds32/nds32.md (*move<mode>): Ditto.
15955 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15956 Monk Chiang <sh.chiang04@gmail.com>
15958 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15960 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15962 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15964 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15965 Kito Cheng <kito.cheng@gmail.com>
15967 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15968 nds32_cond_code_str, output_cond_branch,
15969 output_cond_branch_compare_zero, nds32_expand_cbranch,
15970 nds32_expand_cstore, nds32_expand_movcc,
15971 nds32_output_cbranchsi4_equality_zero,
15972 nds32_output_cbranchsi4_equality_reg,
15973 nds32_output_cbranchsi4_equality_reg_or_const_int,
15974 nds32_output_cbranchsi4_greater_less_zero: New functions.
15975 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15976 nds32_expand_cstore, nds32_expand_movcc,
15977 nds32_output_cbranchsi4_equality_zero,
15978 nds32_output_cbranchsi4_equality_reg,
15979 nds32_output_cbranchsi4_equality_reg_or_const_int,
15980 nds32_output_cbranchsi4_greater_less_zero): Declare.
15981 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15982 nds32_rimm11s_operand): New predicates.
15983 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15984 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15987 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15989 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15990 * config/nds32/nds32.md: Ditto.
15991 * config/nds32/pipelines.md: Ditto.
15993 2018-04-04 Richard Biener <rguenther@suse.de>
15995 PR tree-optimization/85168
15996 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15997 propagating abnormals.
15999 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16001 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16003 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16004 Kito Cheng <kito.cheng@gmail.com>
16006 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16007 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16008 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16009 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16010 * config/nds32/nds32.md (sibcall_internal): New.
16011 (sibcall_register): Remove.
16012 (sibcall_immediate): Remove.
16013 (sibcall_value_internal): New.
16014 (sibcall_value_register): Remove.
16015 (sibcall_value_immediate): Remove.
16016 * config/nds32/predicates.md (nds32_general_register_operand): New.
16017 (nds32_call_address_operand): New.
16019 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16021 PR rtl-optimization/85167
16022 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16023 bb_defs if *split_p, instead preinitialize it to NULL.
16025 PR tree-optimization/85156
16026 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16027 evaluating the argument multiple times.
16029 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16031 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16033 (_mm_cvtpd_ps): Likewise.
16034 (_mm_cvttpd_epi32): Likewise.
16035 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16036 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16037 vector, pixel, and bool following altivec.h include.
16039 2018-04-03 Martin Sebor <msebor@redhat.com>
16041 * doc/extend.texi (Common Function Attributes): Clarify.
16042 (const attribute): Likewise.
16043 (pure attribute): Likewise.
16045 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16048 * config/i386/i386.c (ix86_expand_vector_set): Use
16049 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16051 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16053 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16054 instructions when changing rounding bits to preserve precision bits
16055 in the x87 control word.
16057 2018-04-03 Martin Liska <mliska@suse.cz>
16059 PR tree-optimization/82491
16060 * rtl.h (strip_offset_and_add): Replace += suboffset with
16061 poly_uint64 () + suboffset.
16063 2018-03-29 Martin Liska <mliska@suse.cz>
16064 Martin Jambor <mjambor@suse.cz>
16067 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16068 param_type is not an integral or pointer type.
16070 2018-04-03 Richard Biener <rguenther@suse.de>
16072 * sese.h (recompute_all_dominators): Remove.
16074 2018-04-02 Martin Sebor <msebor@redhat.com>
16076 * doc/invoke.texi (-Wrestrict): Fix typos.
16078 2018-04-02 Jim Wilson <jimw@sifive.com>
16080 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16081 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16082 (<optab>di3, <optab>si3_extend): Likewise.
16083 (<optab>si3_mask, <optab>si3_mask_1): New.
16084 (<optab>di3_mask, <optab>di3_mask_1): New.
16085 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16086 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16087 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16089 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16091 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16094 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16096 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16097 (nds32_canonicalize_comparison): New function.
16099 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16100 Kito Cheng <kito.cheng@gmail.com>
16101 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16103 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16104 * config/nds32/constants.md (unspec_volatile_element): Add
16105 UNSPEC_VOLATILE_RELAX_GROUP.
16106 * config/nds32/nds32-relax-opt.c: New file.
16107 * config/nds32/nds32-predicates.c
16108 (nds32_symbol_load_store_p): New function.
16109 * config/nds32/nds32-protos.h
16110 (nds32_symbol_load_store_p): Declare function.
16111 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16112 * config/nds32/nds32.c
16113 (nds32_register_pass): New function to register pass.
16114 (nds32_register_passes): New function to register passes.
16115 * config/nds32/nds32.md (relax_group): New pattern.
16116 * config/nds32/nds32.opt (mrelax-hint): New option.
16117 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16119 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16121 * config/nds32/t-nds32: Modify files dependency.
16123 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16125 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16126 (PROFILE_HOOK): Define its implementation.
16128 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16130 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16131 type and 32-bit size.
16133 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16135 PR middle-end/85090
16136 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16137 (V_128_256): New mode iterator.
16138 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16139 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16140 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16142 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16143 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16145 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16148 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16149 NaN inputs correctly.
16151 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16154 * config/rs6000/vsx.md (??r): New mode attribute.
16155 (*vsx_mov<mode>_64bit): Use it.
16156 (*vsx_mov<mode>_32bit): Likewise.
16158 2018-03-30 Martin Sebor <msebor@redhat.com>
16160 PR tree-optimization/84818
16161 * builtins.c (check_access): Use warning_n.
16163 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16166 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16168 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16171 2018-03-30 Julia Koval <julia.koval@intel.com>
16174 * x86-tune.def (movx, partial_reg_dependency): Enable for
16177 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16179 PR inline-asm/84985
16180 * lra-constraints.c (process_alt_operands): Move setting
16181 this_alternative_matches below.
16183 2018-03-29 Martin Liska <mliska@suse.cz>
16186 * doc/invoke.texi: Document how LTO works with debug info.
16187 Describe auto-load support of binutils. Mention 'x86-64'
16188 as valid option value of -march option.
16190 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16192 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16195 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16196 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16197 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16200 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16203 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16204 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16205 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16206 for RS6000_BTM_POWERPC64.
16207 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16208 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16209 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16212 (DIVDEU): Likewise.
16214 2018-03-28 Carl Love <cel@us.ibm.com>
16217 2017-09-27 Carl Love <cel@us.ibm.com>
16219 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16220 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16221 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16224 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16226 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16227 instead of __vector bool.
16228 (_mm_max_pu8): Likewise.
16229 (_mm_min_pi16): Likewise.
16231 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16234 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16235 (DIVWEUO): Likewise.
16236 (DIVDEO): Likewise.
16237 (DIVDEUO): Likewise.
16238 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16239 DIVWEUO and DIVDEUO.
16240 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16241 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16242 (div_extend): Likewise.
16243 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16245 (__builtin_divweuo): Likewise.
16246 (__builtin_divdeo): Likewise.
16247 (__builtin_divdeuo): Likewise.
16249 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16252 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16253 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16255 PR tree-optimization/82004
16256 * gimple-match-head.c (optimize_pow_to_exp): New function.
16257 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16258 Don't fold to exp if optimize_pow_to_exp is false.
16260 2018-03-28 Martin Liska <mliska@suse.cz>
16263 * calls.c (initialize_argument_information): Fix trailing space.
16264 * common.opt: Fix typo and provide better explanation for
16265 -fsanitize-coverage option.
16266 * config/i386/i386.opt: Fix typo.
16268 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16269 Martin Liska <mliska@suse.cz>
16272 * gimplify.c (asan_poison_variable): Don't do the check for
16273 gimplify_omp_ctxp here.
16274 (gimplify_decl_expr): Do it here.
16275 (gimplify_target_expr): Likewise.
16277 2018-03-28 Martin Liska <mliska@suse.cz>
16280 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16281 chkp_type_bounds_count if MPX is not enabled.
16283 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16285 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16287 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16290 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16291 function to create the function decl for complex long double
16292 multiply and divide for -mabi=ieeelongdouble.
16293 (init_float128_ieee): Call it.
16295 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16298 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16299 -fcf-protection=branch -mibt.
16300 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16302 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16305 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16306 arm_word_relocations.
16308 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16311 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16312 extern array declarations.
16314 2018-03-27 Richard Biener <rguenther@suse.de>
16316 PR middle-end/84067
16317 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16318 explicit single_use checks.
16320 2018-03-27 Richard Biener <rguenther@suse.de>
16322 PR tree-optimization/85082
16323 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16326 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16328 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16329 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16330 Turn on fasynchronous-unwind-tables and funwind-tables.
16332 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16335 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16336 (*bmi_blsr_<mode>_ccz): Ditto.
16338 2018-03-26 Tom de Vries <tom@codesourcery.com>
16340 PR tree-optimization/85063
16341 * omp-general.c (offloading_function_p): New function. Factor out
16343 * omp-offload.c (pass_omp_target_link::gate): ... here.
16344 * omp-general.h (offloading_function_p): Declare.
16345 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16346 with attribute omp declare target for offloading functions.
16348 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16350 PR tree-optimization/84005
16351 * tree-data-ref.h (get_base_for_alignment): Declare.
16352 * tree-data-ref.c (get_base_for_alignment_1): New function.
16353 (get_base_for_alignment): Likewise.
16354 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16355 get_base_for_alignment to find a suitable base object, instead
16356 of always using drb->base_address.
16358 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16360 PR inline-asm/85022
16361 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16362 known size by default.
16364 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16366 PR inline-asm/85030
16367 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16368 and non BLKmode operands.
16370 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16373 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16374 Clean up attributes.
16376 2018-03-23 Richard Biener <rguenther@suse.de>
16379 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16380 we are going to emit early debug for LTO.
16382 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16384 PR inline-asm/85034
16385 * function.c (match_asm_constraints_1): Don't optimize if input
16386 doesn't satisfy general_operand predicate for output's mode.
16388 PR inline-asm/85022
16389 * alias.c (write_dependence_p): Don't require for x_canonicalized
16390 non-VOIDmode if x has VOIDmode.
16393 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16394 just don't try to optimize it rather than assert it never happens.
16396 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16398 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16399 macro expansions for definition of ST_INTERNAL_<mode> and
16400 LD_INTERNAL_<mode> builtins.
16401 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16403 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16405 (altivec_expand_st_builtin): Likewise.
16406 (altivec_expand_builtin): Remove calls to deleted functions.
16407 (rs6000_address_for_altivec): Delete this function.
16408 * config/rs6000/vector.md: Remove expands for
16409 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16411 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16414 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16415 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16416 re-computing once computed.
16417 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16418 (arm_init_machine_status): Initialize
16419 machine->static_chain_stack_bytes.
16421 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16424 * doc/extend.texi: Add four new prototypes for vec_ld.
16425 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16426 definitions for more logical presentation.
16427 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16428 entries for V1TI variants of __builtin_altivec_ld builtin.
16429 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16430 handling of V1TI variant of LVX icode pattern.
16431 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16432 (rs6000_gimple_fold_builtin): Likewise.
16433 (altivec_init_builtins): Add code to define
16434 __builtin_altivec_lvx_v1ti function.
16436 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16438 PR inline-asm/84941
16439 * function.c (match_asm_constraints_1): Don't do the optimization
16440 if input isn't a REG, SUBREG, MEM or constant.
16442 2018-03-22 Tom de Vries <tom@codesourcery.com>
16444 PR tree-optimization/84956
16445 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16446 bb_has_abnormal_pred.
16448 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16451 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16452 DECL_INITIAL (decl) to decl at the end.
16453 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16454 adjust the comment.
16456 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16458 * doc/extend.texi (__builtin_tgmath): Document when complex
16459 integer types are treated as _Complex _Float64.
16461 2018-03-21 Tom de Vries <tom@codesourcery.com>
16463 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16465 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16467 PR tree-optimization/84960
16468 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16469 if it is ENTRY block, move them into single succ of ENTRY in that case.
16471 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16473 PR tree-optimization/84811
16474 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16475 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16476 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16477 (known_subrange_p): Update accordingly. Cast each value involved
16478 in the size comparison, rather than casting the result of the
16481 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16483 PR tree-optimization/84982
16484 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16485 by flipping the least significant bit rather than all bits from
16486 bitpos to bitpos + bitsize - 1.
16488 2018-03-21 Nathan Sidwell <nathan@acm.org>
16490 * doc/extend.texi (Deprecated Features): Remove mention of
16491 long-deleted deprecations.
16493 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16496 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16497 * configure: Regenerate.
16499 2018-03-21 Tom de Vries <tom@codesourcery.com>
16501 PR tree-optimization/83126
16502 * tree-parloops.c (num_phis): New function.
16503 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16505 2018-03-21 Nathan Sidwell <nathan@acm.org>
16507 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16508 mention anon-struct/union members and trailing attributes.
16510 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16512 PR tree-optimization/84969
16513 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16514 builtin memset partitions if they set different rhs values.
16516 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16518 PR rtl-optimization/84989
16519 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16520 VEC_DUPLICATE with scalar result mode.
16522 2018-03-21 Martin Liska <mliska@suse.cz>
16525 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16526 not intended return statement.
16528 2018-03-21 Martin Liska <mliska@suse.cz>
16531 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16532 (chkp_find_bound_slots_1): Limit number of iterations.
16534 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16537 * Minor grammar fixes for x86 options.
16539 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16542 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16543 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16545 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16548 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16549 (altivec_lvx_<mode>_1op): Likewise.
16550 (altivec_stvx_<mode>_2op): Likewise.
16551 (altivec_stvx_<mode>_1op): Likewise.
16552 (altivec_lvx_<VM2:mode>): New define_expand.
16553 (altivec_stvx_<VM2:mode>): Likewise.
16554 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16555 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16556 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16557 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16558 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16559 (rs6000_gen_lvx): Likewise.
16560 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16561 (altivec_expand_stv_builtin): Likewise.
16562 (altivec_expand_builtin): Likewise.
16563 * config/rs6000/vector.md: Likewise.
16565 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16568 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16571 2018-03-20 Richard Biener <rguenther@suse.de>
16574 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16575 sign-conversions as zero, fall back to standard scalar_stmt
16578 2018-03-20 Martin Liska <mliska@suse.cz>
16581 * predict.c (rebuild_frequencies): Handle case when we have
16582 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16584 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16587 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16588 flag_section_anchors.
16589 * varasm.c (use_blocks_for_decl_p): Remove hack for
16590 dw2_force_const_mem.
16593 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16595 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16596 be created, use lowpart_subreg of operands[0] rather than operands[0]
16598 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16599 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16600 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16601 and n constraint instead of aarch64_shift_imm_di and Usd.
16602 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16603 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16605 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16608 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16609 to favor GPR over NEON registers.
16610 (<shift>di3_neon): Likewise.
16612 2018-03-20 Tom de Vries <tom@codesourcery.com>
16615 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16616 (nvptx_process_pars): Emit bar.sync asap and alap.
16618 2018-03-20 Tom de Vries <tom@codesourcery.com>
16621 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16622 seen_label if seen_label is already set.
16624 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16627 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16628 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16629 Use 1U instead of 1. Formatting fixes.
16632 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16633 instead of TREE_TYPE (s1) for the return value.
16635 2018-03-19 Jakub Jelinek <jakub@redhat.com>
16637 PR tree-optimization/84946
16638 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16639 bitsize + bitsize in poly_uint64 rather than poly_int64.
16642 * dwarf2asm.c: Include fold-const.c.
16643 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16644 of decl rather than decl itself.
16646 PR rtl-optimization/84643
16647 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16649 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
16652 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16653 calling assemble_variable.
16655 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16658 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16659 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16661 2018-03-19 Jim Wilson <jimw@sifive.com>
16664 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16665 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16666 (riscv_first_stack_step): Likewise.
16667 (riscv_option_override): Use STACK_BOUNDARY instead of
16668 MIN_STACK_BOUNDARY.
16669 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
16670 MIN_STACK_BOUNDARY.
16671 (BIGGEST_ALIGNMENT): Set to 128.
16672 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
16673 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
16676 2018-03-19 Richard Biener <rguenther@suse.de>
16678 PR tree-optimization/84933
16679 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
16680 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
16682 2018-03-19 Richard Biener <rguenther@suse.de>
16684 PR tree-optimization/84859
16685 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
16686 (cond_if_else_store_replacement): Perform sinking operation on
16687 single-store BBs regardless of MAX_STORES_TO_SINK setting.
16688 Generalize what a BB with a single eligible store is.
16690 2018-03-19 Richard Biener <rguenther@suse.de>
16692 PR tree-optimization/84929
16693 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
16694 chrec_is_positive against non-chrec arg.
16696 2018-03-19 Tamar Christina <tamar.christina@arm.com>
16699 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
16701 2018-03-18 Martin Liska <mliska@suse.cz>
16703 PR rtl-optimization/84635
16704 * regrename.c (build_def_use): Use matches_mode only when
16707 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
16709 PR tree-optimization/84913
16710 * tree-vect-loop.c (vectorizable_reduction): Don't try to
16711 vectorize chains of COND_EXPRs.
16713 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16715 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
16717 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16719 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
16721 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16723 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16725 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16726 Kito Cheng <kito.cheng@gmail.com>
16728 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
16729 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
16730 (nds32_adjust_reg_alloc_order): New function.
16731 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
16733 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
16735 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16736 nds32_print_operand, nds32_print_operand_address): Use
16737 HOST_WIDE_INT_PRINT_DEC instead.
16739 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16741 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
16743 2018-03-17 Jakub Jelinek <jakub@redhat.com>
16746 * config/i386/i386.c (initial_ix86_tune_features,
16747 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
16748 unsigned long long.
16749 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
16750 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
16751 rather than 1u << ix86_tune. Formatting fix.
16752 (ix86_option_override_internal): Change ix86_arch_mask from
16753 unsigned int to unsigned HOST_WIDE_INT, initialize to
16754 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
16755 (ix86_function_specific_restore): Likewise.
16757 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16760 * postreload.c (reload_combine_recognize_pattern): Perform
16761 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
16762 truncate_int_for_mode the result for the destination's mode.
16765 * hsa-gen.c (mem_type_for_type): Fix comment typo.
16766 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16768 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16771 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
16774 * lra-assigns.c (lra_split_hard_reg_for): Don't use
16775 regno_allocno_class_array and sorted_pseudos.
16776 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
16777 insns where regno is used.
16779 2018-03-16 Martin Liska <mliska@suse.cz>
16782 * multiple_target.c (create_dispatcher_calls): Redirect
16783 reference in the symbol table.
16785 2018-03-16 Martin Liska <mliska@suse.cz>
16788 * multiple_target.c (create_dispatcher_calls): Redirect also
16791 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16795 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16796 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16797 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16799 2018-03-16 Julia Koval <julia.koval@intel.com>
16801 * doc/invoke.texi (Skylake Server): Add CLWB.
16802 Cannonlake): Remove CLWB.
16804 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16806 PR tree-optimization/84841
16807 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16809 (FLOAT_ONE_CONST_TYPE): Define.
16810 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16811 (sort_by_operand_rank): Put entries with higher constant_type last
16812 rather than first to match comments.
16814 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16816 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16819 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16822 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16823 check and use address of target always.
16825 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16828 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16829 (indirect_thunk_bnd_needed): Likewise.
16830 (indirect_thunks_used): Likewise.
16831 (indirect_thunks_bnd_used): Likewise.
16832 (indirect_return_needed): New.
16833 (indirect_return_bnd_needed): Likewise.
16834 (output_indirect_thunk_function): Add a bool argument for
16836 (output_indirect_thunk_function): Don't generate alias for
16837 function return thunk.
16838 (ix86_code_end): Call output_indirect_thunk_function to generate
16839 function return thunks.
16840 (ix86_output_function_return): Set indirect_return_bnd_needed
16841 and indirect_return_needed instead of indirect_thunk_bnd_needed
16842 and indirect_thunk_needed.
16844 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16846 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16847 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16848 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16850 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16851 Paul Hua <paul.hua.gm@gmail.com>
16854 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16856 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16858 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16859 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16860 resp. SFmode cases.
16862 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16865 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16866 instead of GET_MODE_SIZE when comparing Units.
16868 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16871 * varasm.c (hash_section): Return an unchangeble hash value
16872 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16873 Return !aarch64_can_use_per_function_literal_pools_p ().
16875 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16878 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16879 rather than address of cmode as last argument to prepare_cmp_insn.
16881 2018-03-15 Julia Koval <julia.koval@intel.com>
16883 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16884 F_AVX512VNNI, F_AVX512BITALG): New.
16886 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16889 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16890 insn for floating-point loads and stores.
16892 2018-03-14 Carl Love <cel@us.ibm.com>
16894 * config/rs6000/rs6000-c.c: Add macro definitions for
16895 ALTIVEC_BUILTIN_VEC_PERMXOR.
16896 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16897 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16898 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16899 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16901 * config/doc/extend.texi: Add prototypes for vec_permxor.
16903 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16906 * diagnostic-show-locus.c (class layout_point): Convert m_line
16907 from int to linenum_type.
16908 (line_span::comparator): Use linenum "compare" function when
16909 comparing line numbers.
16910 (test_line_span): New function.
16911 (layout_range::contains_point): Convert param "row" from int to
16913 (layout_range::intersects_line_p): Likewise.
16914 (layout::will_show_line_p): Likewise.
16915 (layout::print_source_line): Likewise.
16916 (layout::should_print_annotation_line_p): Likewise.
16917 (layout::print_annotation_line): Likewise.
16918 (layout::print_leading_fixits): Likewise.
16919 (layout::annotation_line_showed_range_p): Likewise.
16920 (struct line_corrections): Likewise for field m_row.
16921 (line_corrections::line_corrections): Likewise for param "row".
16922 (layout::print_trailing_fixits): Likewise.
16923 (layout::get_state_at_point): Likewise.
16924 (layout::get_x_bound_for_row): Likewise.
16925 (layout::print_line): Likewise.
16926 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16927 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16928 * input.c (selftest::test_linenum_comparisons): New function.
16929 (selftest::input_c_tests): Call it.
16930 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16931 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16932 * selftest.h (ASSERT_GT): New macro.
16933 (ASSERT_GT_AT): New macro.
16934 (ASSERT_LT): New macro.
16935 (ASSERT_LT_AT): New macro.
16937 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16939 PR rtl-optimization/84780
16940 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16942 2018-03-14 Martin Liska <mliska@suse.cz>
16944 * tree.c (record_node_allocation_statistics): Use
16945 get_stats_node_kind.
16946 (get_stats_node_kind): New function extracted from
16947 record_node_allocation_statistics.
16948 (free_node): Use get_stats_node_kind.
16950 2018-03-14 Richard Biener <rguenther@suse.de>
16952 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16953 that the value-set of ANTIC_IN doesn't grow.
16956 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16958 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16959 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16961 2018-03-14 Julia Koval <julia.koval@intel.com>
16963 * config.gcc (icelake-client, icelake-server): New.
16965 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16966 (initial_ix86_arch_features): Ditto.
16967 (PTA_SKYLAKE): Add SGX.
16968 (PTA_ICELAKE): Remove.
16969 (PTA_ICELAKE_CLIENT): New.
16970 (PTA_ICELAKE_SERVER): New.
16971 (ix86_option_override_internal): Split up icelake on icelake client and
16973 (get_builtin_code_for_version): Ditto.
16974 (fold_builtin_cpu): Ditto.
16975 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16976 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16977 * config/i386/i386.h (processor_type): Ditto.
16978 * doc/invoke.texi: Ditto.
16980 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16983 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16984 INTEGER_CST offset, add it together with bitpos / 8 and
16985 sign extend based on POINTER_SIZE.
16989 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16992 * config/i386/constraints.md (Yc): New register constraint.
16993 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16994 Use Yc constraint for alternative 2 of operand 0. Remove
16995 preferred_for_speed attribute.
16997 2018-03-14 Richard Biener <rguenther@suse.de>
16999 PR tree-optimization/84830
17000 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17001 with the old one to avoid oscillations.
17003 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17006 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17008 (assign_by_spills): Return a flag of reload assignment failure.
17009 Do not process the reload assignment failures. Do not spill other
17010 reload pseudos if they has the same reg class. Update n if
17012 (lra_assign): Add a return arg. Set up from the result of
17013 assign_by_spills call.
17014 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17015 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17016 usage_insns if it is not NULL.
17017 (spill_hard_reg_in_range): New function.
17018 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17019 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17020 function prototypes.
17021 (lra_assign): Change prototype.
17022 * lra.c (lra): Add code to deal with fails by splitting hard reg
17025 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17027 * config/riscv/riscv.opt (mrelax): New option.
17028 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17029 "norelax" when riscv_mrelax is disabled.
17030 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17032 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17035 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17036 reassociation for int modes.
17038 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17040 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17041 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17043 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17044 * config/aarch64/aarch64-sve.md
17045 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17046 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17047 (*extend<mode><Vwide>2): Rename to...
17048 (aarch64_sve_extend<mode><Vwide>2): ...this.
17049 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17050 renaming the old pattern to...
17051 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17053 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17054 define_expand, renaming the old pattern to...
17055 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17056 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17057 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17058 account when deciding which SVE instruction the optab should use.
17059 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17061 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17063 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17064 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17065 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17066 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17067 (tlsdesc_small_<mode>): Turn a define_expand and use
17068 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17069 (tlsdesc_small_advsimd_<mode>): ...this.
17070 (tlsdesc_small_sve_<mode>): New pattern.
17072 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17074 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17075 (UNSPEC_UMUL_HIGHPART): New constants.
17076 (MUL_HIGHPART): New int iteraor.
17077 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17078 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17080 (*<su>mul<mode>3_highpart): New define_insn.
17082 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17085 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17088 2018-03-13 Martin Liska <mliska@suse.cz>
17091 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17093 (sem_item_optimizer::~sem_item_optimizer): Release it.
17094 (sem_item_optimizer::merge_classes): Register variable aliases.
17095 (sem_item_optimizer::fixup_pt_set): New function.
17096 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17097 * ipa-icf.h: Declare new variables and functions.
17099 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17101 PR middle-end/84834
17102 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17103 integer_pow2p@2 and test integer_pow2p in condition.
17104 (A < 0 ? C : 0): Similarly for @1.
17106 PR middle-end/84831
17107 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17108 characters starting at p contain '\0' character, don't look beyond
17112 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17113 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17116 * reg-stack.c (change_stack): Change update_end var from int to
17117 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17118 also call set_block_for_insn on the newly added insns and rescan.
17121 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17122 on the last operand.
17125 * tree.c (stabilize_reference_1): Return save_expr (e) for
17126 STATEMENT_LIST even if it doesn't have side-effects.
17128 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17130 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17132 2018-03-12 Renlin Li <renlin.li@arm.com>
17134 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17135 aarch64_output_scalar_simd_mov_immediate.
17137 2018-03-12 Martin Sebor <msebor@redhat.com>
17139 PR tree-optimization/83456
17140 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17141 for perfectly overlapping calls to memcpy.
17142 (gimple_fold_builtin_memory_chk): Same.
17143 (gimple_fold_builtin_strcpy): Handle no-warning.
17144 (gimple_fold_builtin_stxcpy_chk): Same.
17145 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17147 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17149 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17150 parameter. Use it for SFmode.
17151 (rs6000_function_arg_advance_1): Adjust.
17152 (rs6000_function_arg): Adjust.
17153 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17155 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17157 PR rtl-optimization/84169
17158 PR rtl-optimization/84780
17159 * combine.c (can_combine_p): Check for a 2-insn combination whether
17160 the destination register is used between the two insns, too.
17162 2018-03-12 Richard Biener <rguenther@suse.de>
17164 PR tree-optimization/84803
17165 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17166 for refs DR analysis didn't process.
17168 2018-03-12 Richard Biener <rguenther@suse.de>
17170 PR tree-optimization/84777
17171 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17172 force-vectorize loops ignore whether we are optimizing for size.
17174 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17176 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17177 (TARGET_MD_ASM_ADJUST): Define.
17179 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17180 Kito Cheng <kito.cheng@gmail.com>
17181 Chung-Ju Wu <jasonwucj@gmail.com>
17183 * config/nds32/nds32.c (nds32_compute_stack_frame,
17184 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17185 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17186 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17187 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17188 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17189 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17190 * config/nds32/nds32.md (prologue, epilogue): Use macro
17191 NDS32_V3PUSH_AVAILABLE_P to do checking.
17193 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17196 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17197 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17198 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17199 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17200 addition of most attributes on !orig_type_die or the attribute not
17201 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17203 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17204 Chung-Ju Wu <jasonwucj@gmail.com>
17206 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17207 __NDS32_VH__ macro.
17208 * config/nds32/nds32.opt (mvh): New option.
17210 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17211 Chung-Ju Wu <jasonwucj@gmail.com>
17213 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17215 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17216 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17219 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17220 Chung-Ju Wu <jasonwucj@gmail.com>
17222 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17224 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17225 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17227 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17228 Kito Cheng <kito.cheng@gmail.com>
17229 Chung-Ju Wu <jasonwucj@gmail.com>
17231 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17232 UNSPEC_FFMISM and UNSPEC_FLMISM.
17233 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17234 for ffb, ffmism and flmism.
17235 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17236 (unspec_ffmism): Ditto.
17237 (unspec_flmism): Ditto.
17238 (nds32_expand_builtin_impl): Check if string extension is available.
17239 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17240 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17242 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17245 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17248 * lra-assigns.c (assign_by_spills): Return a flag of reload
17249 assignment failure. Do not process the reload assignment
17250 failures. Do not spill other reload pseudos if they has the same
17252 (lra_assign): Add a return arg. Set up from the result of
17253 assign_by_spills call.
17254 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17255 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17256 usage_insns if it is not NULL.
17257 (spill_hard_reg_in_range): New function.
17258 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17259 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17260 function prototypes.
17261 (lra_assign): Change prototype.
17262 * lra.c (lra): Add code to deal with fails by splitting hard reg
17265 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17268 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17270 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17273 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17274 (dw_val_node): Add val_symbolic_view.
17275 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17276 (symview_upper_bound): New.
17277 (new_line_info_table): Initialize symviews_since_reset.
17278 (dwarf2out_source_line): Count symviews_since_reset and set
17279 symview_upper_bound.
17280 (dw_val_equal_p): Handle symview.
17281 (add_AT_symview): New.
17282 (print_dw_val): Handle symview.
17283 (attr_checksum, attr_checksum_ordered): Likewise.
17284 (same_dw_val_p, size_of_die): Likewise.
17285 (value_format, output_die): Likewise.
17286 (add_high_low_attributes): Use add_AT_symview for entry_view.
17287 (dwarf2out_finish): Reset symview_upper_bound, clear
17290 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17293 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17294 Add strict argument and use it.
17295 (rs6000_split_multireg_move): Update for new strict argument.
17296 (mem_operand_gpr): Disallow all non-offsettable addresses.
17297 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17299 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17302 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17303 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17304 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17307 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17308 decl, use remap_type if we want to use the type.
17310 2018-03-09 Martin Sebor <msebor@redhat.com>
17312 PR tree-optimization/84526
17313 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17315 (builtin_access::generic_overlap): Be prepared to handle non-array
17318 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17320 PR rtl-optimization/84682
17321 * lra-constraints.c (process_address_1): Check is_address flag
17322 for address constraints.
17323 (process_alt_operands): Likewise.
17324 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17325 preprocess_constraints.
17326 * recog.h (preprocess_constraints): Add oploc parameter.
17328 * recog.c (preprocess_constraints): Test address_operand for
17329 CT_ADDRESS constraints.
17331 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17334 * lra-assigns.c (assign_by_spills): Return a flag of reload
17335 assignment failure. Do not process the reload assignment
17336 failures. Do not spill other reload pseudos if they has the same
17338 (lra_assign): Add a return arg. Set up from the result of
17339 assign_by_spills call.
17340 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17341 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17342 usage_insns if it is not NULL.
17343 (spill_hard_reg_in_range): New function.
17344 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17345 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17346 function prototypes.
17347 (lra_assign): Change prototype.
17348 * lra.c (lra): Add code to deal with fails by splitting hard reg
17351 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17354 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17355 Accept complain bool parameter. Only emit errors if it is true.
17356 (arm_parse_cpu_option_name): Likewise.
17357 (arm_target_thumb_only): Adjust callers of the above.
17358 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17359 prototype to take a default true bool parameter.
17360 (arm_parse_arch_option_name): Likewise.
17362 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17363 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17367 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17368 * configure: Regenerate.
17369 * configure.ac ("linker --version-script option"): New.
17370 ("linker soname option"): New.
17372 2018-03-09 Richard Biener <rguenther@suse.de>
17374 PR tree-optimization/84775
17375 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17376 immediate uses of predicate stmts and mark them modified.
17379 PR tree-optimization/84178
17380 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17382 (version_loop_for_if_conversion): Delay update_ssa call.
17383 (tree_if_conversion): Delay update_ssa until after predicate
17386 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17389 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17390 when the function accesses prior frames.
17392 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17395 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17396 gen_llsym, otherwise call maybe_gen_llsym.
17398 PR inline-asm/84742
17399 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17400 has ',' character inside of it.
17402 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17405 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17406 as clobbering CC_REGNUM.
17408 2018-03-08 Richard Biener <rguenther@suse.de>
17410 PR middle-end/84552
17411 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17412 (follow_copies_to_constant): Do not follow SSA names registered
17415 2018-03-08 Richard Biener <rguenther@suse.de>
17417 PR tree-optimization/84178
17418 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17420 (version_loop_for_if_conversion): Delay update_ssa call.
17421 (tree_if_conversion): Delay update_ssa until after predicate
17424 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17426 PR tree-optimization/84178
17427 * tree-if-conv.c (release_bb_predicate): Remove the
17428 the assertion that the stmts have NULL use_ops.
17429 Discard the statements, asserting that they haven't
17430 yet been added to a BB.
17432 2018-03-08 Richard Biener <rguenther@suse.de>
17434 PR tree-optimization/84746
17435 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17436 (phi_translate): Pass in destination ANTIC_OUT set.
17437 (phi_translate_1): Likewise. For a simplified result lookup
17438 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17439 (phi_translate_set): Adjust.
17440 (do_pre_regular_insertion): Likewise.
17441 (do_pre_partial_partial_insertion): Likewise.
17443 2018-03-08 Martin Liska <mliska@suse.cz>
17445 PR gcov-profile/84735
17446 * doc/gcov.texi: Document usage of profile files.
17447 * gcov-io.h: Document changes in the format.
17449 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17453 * dwarf2out.c (struct dw_line_info_table): Update comments for
17455 (FORCE_RESET_NEXT_VIEW): New.
17456 (FORCE_RESETTING_VIEW_P): New.
17457 (RESETTING_VIEW_P): Check for -1 too.
17458 (ZERO_VIEW_P): Likewise.
17459 (new_line_info_table): Force-reset next view.
17460 (dwarf2out_begin_function): Likewise.
17461 (dwarf2out_source_line): Simplify zero_view_p initialization.
17462 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17463 view directly. Omit view when omitting .loc at line 0.
17465 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17467 PR tree-optimization/84740
17468 * tree-switch-conversion.c (process_switch): Call build_constructors
17469 only if info.phi_count is non-zero.
17471 PR tree-optimization/84739
17472 * tree-tailcall.c (find_tail_calls): Check call arguments against
17473 DECL_ARGUMENTS (current_function_decl) rather than
17474 DECL_ARGUMENTS (func) when checking for tail recursion.
17476 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17478 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17479 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17480 Volker Reichelt's entry and add entries for people that perform
17481 GCC fuzzy testing and report numerous bugs.
17483 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17486 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17487 readonly data in sdata, if that is disabled.
17488 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17489 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17490 -mreadonly-in-sdata option.
17492 2018-03-07 Martin Sebor <msebor@redhat.com>
17494 PR tree-optimization/84468
17495 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17496 basic block when looking for nul assignment.
17498 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17501 * except.h (output_function_exception_table): Adjust prototype.
17502 * except.c (output_function_exception_table): Remove FNNAME parameter
17503 and add SECTION parameter. Ouput one part of the table at a time.
17504 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17505 the first part of the exception table and emit unwind directives.
17506 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17507 (i386_pe_seh_cold_init): Likewise.
17508 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17509 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17510 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17511 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17512 * config/i386/winnt.c: Include except.h.
17513 (struct seh_frame_state): Add reg_offset, after_prologue and
17514 in_cold_section fields.
17515 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17516 (i386_pe_seh_cold_init): New function.
17517 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17518 to seh->in_cold_section.
17519 (seh_emit_push): Record the offset of the push.
17520 (seh_emit_save): Record the offet of the save.
17521 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17522 Test seh->after_prologue to disregard the epilogue.
17523 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17524 (i386_pe_end_cold_function): New function.
17526 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17529 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17530 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17533 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17535 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17536 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17538 PR middle-end/84723
17539 * multiple_target.c: Include tree-inline.h and intl.h.
17540 (expand_target_clones): Diagnose and fail if node->definition and
17541 !tree_versionable_function_p (node->decl).
17543 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17545 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17547 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17548 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17549 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17551 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17554 * combine.c (try_combine): Use reg_or_subregno instead of handling
17555 just paradoxical SUBREGs and REGs.
17557 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17559 * config/arc/arc.c (arc_finalize_pic): Remove function.
17560 (arc_must_save_register): We use single base PIC register, remove
17561 checks to save/restore the PIC register.
17562 (arc_expand_prologue): Likewise.
17563 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17565 (arc_verify_short): Likewise.
17566 (arc_attr_type): Likewise.
17567 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17568 (walk_stores): Likewise.
17569 (arc_address_cost): Make it static.
17570 (arc_verify_short): Likewise.
17571 (branch_dest): Likewise.
17572 (arc_attr_type): Likewise.
17573 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17574 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17575 (arc_final_prescan_insn): Remove inserting the nops due to
17576 hardware hazards. It is done in reorg step.
17577 (insn_length_variant_t): Remove.
17578 (insn_length_parameters_t): Likewise.
17579 (arc_insn_length_parameters): Likewise.
17580 (arc_get_insn_variants): Likewise.
17581 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17583 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17585 PR inline-asm/84683
17586 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17589 PR tree-optimization/84687
17590 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17592 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17594 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17596 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17597 Rename to ppc_speculation_barrier.
17598 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17599 __builtin_ppc_speculation_barrier.
17601 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17604 * combine.c (combine_simplify_rtx): Don't try to simplify if
17605 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17608 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17610 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17611 to 32 bytes when compiling for POWER9.
17613 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17616 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17617 regparm >= 3 with no arg reg available also for calls with
17618 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17621 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17623 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17625 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17628 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17630 2018-03-05 Richard Biener <rguenther@suse.de>
17632 PR tree-optimization/84486
17633 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17634 When inserting a __builtin_assume_aligned call set the LHS
17635 SSA name alignment info accordingly.
17637 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
17639 PR tree-optimization/84114
17640 * config/aarch64/aarch64.c (aarch64_reassociation_width)
17641 Avoid reassociation of FLOAT_MODE addition.
17643 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
17645 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17646 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17647 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17648 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17649 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17650 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17651 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17653 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17654 __builtin_ia32_wbinvd): New builtins.
17655 (SPECIAL_ARGS2): New.
17656 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17657 (SPECIAL_ARGS2): New.
17658 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17659 (ix86_valid_target_attribute_inner_p): Ditto.
17660 (ix86_init_mmx_sse_builtins): Add special_args2.
17661 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17662 TARGET_WBNOINVD_P): New.
17663 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17664 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17665 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17666 * config/i386/immintrin.h (_wbinvd): New intrinsic.
17667 * config/i386/pconfigintrin.h: New file.
17668 * config/i386/wbnoinvdintrin.h: Ditto.
17669 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
17671 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
17673 2018-03-05 Richard Biener <rguenther@suse.de>
17675 PR tree-optimization/84670
17676 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17678 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17679 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17680 (compute_antic_aux): Only assert the number of values in ANTIC_IN
17681 doesn't grow if all successors (recursively) were visited at least
17684 2018-03-05 Richard Biener <rguenther@suse.de>
17686 PR tree-optimization/84650
17687 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
17688 if executed in the loop pipeline.
17690 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
17692 * doc/configfiles.texi (Configuration Files): Move info about
17693 conditionalizing $target-protos.h to...
17694 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
17695 differs from $target-protos.h.
17697 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
17698 Chung-Ju Wu <jasonwucj@gmail.com>
17700 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
17701 * config/nds32/nds32-multiple.md (setmemsi): Define.
17702 * config/nds32/nds32-memory-manipulation.c
17703 (nds32_gen_dup_4_byte_to_word_value): New.
17704 (emit_setmem_word_loop): New.
17705 (emit_setmem_byte_loop): New.
17706 (nds32_expand_setmem_loop): New.
17707 (nds32_expand_setmem_loop_v3m): New.
17708 (nds32_expand_setmem_unroll): New.
17709 (nds32_expand_setmem): New.
17711 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17712 Chung-Ju Wu <jasonwucj@gmail.com>
17714 * config/nds32/nds32-memory-manipulation.c
17715 (nds32_emit_load_store): New.
17716 (nds32_emit_post_inc_load_store): New.
17717 (nds32_emit_mem_move): New.
17718 (nds32_emit_mem_move_block): New.
17719 (nds32_expand_movmemsi_loop_unknown_size): New.
17720 (nds32_expand_movmemsi_loop_known_size): New.
17721 (nds32_expand_movmemsi_loop): New.
17722 (nds32_expand_movmemsi_unroll): New.
17723 (nds32_expand_movmemqi): Rename ...
17724 (nds32_expand_movmemsi): ... to this.
17725 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
17726 (movmemsi): ... to this.
17727 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
17728 (nds32_expand_movmemsi): ... to this.
17730 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17731 Monk Chiang <sh.chiang04@gmail.com>
17732 Chung-Ju Wu <jasonwucj@gmail.com>
17734 * config/nds32/nds32-protos.h
17735 (nds32_expand_load_multiple): New arguments.
17736 (nds32_expand_store_multiple): Ditto.
17737 (nds32_valid_multiple_load_store): Rename ...
17738 (nds32_valid_multiple_load_store_p): ... to this.
17739 * config/nds32/nds32-memory-manipulation.c
17740 (nds32_expand_load_multiple): Refine implementation.
17741 (nds32_expand_store_multiple): Ditto.
17742 * config/nds32/nds32-multiple.md
17743 (load_multiple): Update nds32_expand_load_multiple interface.
17744 (store_multiple): Update nds32_expand_store_multiple interface.
17745 * config/nds32/nds32-predicates.c
17746 (nds32_valid_multiple_load_store): Rename ...
17747 (nds32_valid_multiple_load_store_p): ... to this and refine
17749 * config/nds32/predicates.md
17750 (nds32_load_multiple_and_update_address_operation): New predicate.
17751 (nds32_store_multiple_and_update_address_operation): New predicate.
17753 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17754 Chung-Ju Wu <jasonwucj@gmail.com>
17756 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
17757 (combo): New attribute.
17758 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
17760 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
17762 * config/nds32/nds32.opt: Change -mcmodel= default value.
17764 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
17765 Monk Chiang <sh.chiang04@gmail.com>
17766 Chung-Ju Wu <jasonwucj@gmail.com>
17768 * config/nds32/constants.md (unspec_element): New enum.
17769 * config/nds32/constraints.md (Umw): New constraint.
17770 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
17771 * config/nds32/nds32-intrinsic.md: Likewise.
17772 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
17773 (nds32_valid_smw_lwm_base_p): New.
17774 (nds32_output_smw_single_word): New.
17775 (nds32_output_lmw_single_word): New.
17776 (nds32_expand_unaligned_load): New.
17777 (nds32_expand_unaligned_store): New.
17778 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
17779 (nds32_output_smw_single_word): Declare.
17780 (nds32_output_lmw_single_word): Declare.
17781 (nds32_expand_unaligned_load): Declare.
17782 (nds32_expand_unaligned_store): Declare.
17783 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17784 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17785 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17786 NDS32_BUILTIN_UASTORE_DW.
17787 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17790 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
17791 Kito Cheng <kito.cheng@gmail.com>
17792 Chung-Ju Wu <jasonwucj@gmail.com>
17794 * config/nds32/nds32-intrinsic.c
17795 (nds32_expand_builtin_null_ftype_reg): Delete.
17796 (nds32_expand_builtin_reg_ftype_imm): Ditto.
17797 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17798 (nds32_read_argument): New.
17799 (nds32_legitimize_target): Ditto.
17800 (nds32_legitimize_argument): Ditto.
17801 (nds32_check_constant_argument): Ditto.
17802 (nds32_expand_unop_builtin): Ditto.
17803 (nds32_expand_unopimm_builtin): Ditto.
17804 (nds32_expand_binop_builtin): Ditto.
17805 (nds32_builtin_decl_impl): Ditto.
17806 (builtin_description): Ditto.
17807 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17808 (nds32_init_builtins_impl): Ditto.
17809 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17810 (nds32_builtin_decl): New.
17811 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17812 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17814 2018-03-02 Jeff Law <law@redhat.com>
17816 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17817 (redundant_insn, fill_simple_delay_slots): Likewise.
17818 (fill_slots_from_thread): Likewise.
17819 * resource.c (mark_referenced_resources): Likewise.
17820 (mark_set_resources, find_dead_or_set_registers): Likewise.
17822 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17824 * substring-locations.h (format_warning_va): Formatting fix for
17825 ATTRIBUTE_GCC_DIAG.
17826 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17828 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17829 * substring-locations.c: Include intl.h.
17830 (format_warning_va): Turned into small wrapper around
17831 format_warning_n_va, renamed to ...
17832 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17833 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17835 (format_warning_at_substring_n): New function.
17836 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17837 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17838 format_warning_at_substring with just a shorter name instead of
17839 const function pointer.
17840 (fmtwarn_n): New function.
17841 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17842 appropriate, get rid of all the fmtstr temporaries, move conditionals
17843 with G_() wrapped string literals directly into fmtwarn arguments,
17844 cast dir.len to (int), formatting fixes.
17846 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17848 * doc/invoke.texi: Remove "Cilk Plus" references.
17850 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17851 Richard Biener <rguenther@suse.de>
17854 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17855 for error or warning attributes if CALL_FROM_THUNK_P is set.
17858 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17861 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17862 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17865 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17866 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17868 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17869 instead of -1U in last predictors element's probability member.
17871 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17874 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17875 arguments if they are comparable.
17877 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17879 PR tree-optimization/84634
17880 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17881 masks and masked_loop_p with a single loop_masks, making sure it's
17882 null for bb vectorization.
17884 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17886 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17887 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17888 loop->force_vectorize to check whether there is no alias.
17890 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17893 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17895 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17897 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17898 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17899 instead of a loop around prev_real_insn.
17900 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17903 PR inline-asm/84625
17904 * config/i386/i386.c (ix86_print_operand): Use conditional
17905 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17908 2018-03-02 Richard Biener <rguenther@suse.de>
17910 PR tree-optimization/84427
17911 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17912 (bitmap_set_subtract_values): Rewrite to handle multiple
17915 (prune_clobbered_mems): Likewise.
17916 (phi_translate): Take edge instead of pred/phiblock.
17917 (phi_translate_1): Likewise.
17918 (phi_translate_set): Likewise. Insert all translated
17919 exprs for a value into the set, keeping possibly multiple
17920 expressions per value.
17921 (compute_antic_aux): Adjust for phi_translate changes.
17922 When intersecting union the expressions and prune those
17923 not in the final value set, keeping possibly multiple
17924 expressions per value. Do not use value-insertion
17925 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17926 all expressions. Add verification that the value-sets
17927 only shrink during iteration.
17928 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17929 (do_pre_regular_insertion): Likewise.
17930 (do_pre_partial_partial_insertion): Likewise.
17932 2018-03-02 Richard Biener <rguenther@suse.de>
17935 * config/darwin.c (saved_debug_info_level): New static global.
17936 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17937 (darwin_asm_lto_end): Restore debug info generation settings.
17939 2018-03-01 Martin Liska <mliska@suse.cz>
17942 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17943 volatile arguments.
17945 2018-03-01 Richard Biener <rguenther@suse.de>
17948 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17949 type annotation with LTO when debug was disabled at compile-time.
17951 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17953 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17955 (mips_final_postscan_insn): Likewise.
17957 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17959 PR rtl-optimization/84528
17960 * alias.c (init_alias_target): Add commentary.
17961 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17962 a unique base value if the frame pointer is not eliminated
17963 to the stack pointer.
17965 2018-03-01 Tom de Vries <tom@codesourcery.com>
17967 PR rtl-optimization/83327
17968 * lra-int.h (hard_regs_spilled_into): Declare.
17969 * lra.c (hard_regs_spilled_into): Define.
17970 (init_reg_info): Init hard_regs_spilled_into.
17971 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17972 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17973 (process_bb_lives): Handle hard_regs_spilled_into.
17974 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17975 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17977 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17979 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17980 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17981 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17982 * config/rs6000/aix72.h: New file.
17984 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17986 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17987 instead of warning_at with conditional singular and plural messages
17991 * stor-layout.c (update_alignment_for_field): For
17992 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17993 && !DECL_PACKED (field), do the alignment update, just use
17994 only desired_align instead of MAX (type_align, desired_align)
17996 (place_field): Don't do known_align < desired_align handling
17997 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17998 is non-NULL, instead do it after rli->prev_field handling and
17999 only if not within a bitfield word. For DECL_PACKED (field)
18000 use type_align of BITS_PER_UNIT.
18002 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18004 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18005 superfluous parentheses and trailing spaces.
18007 2018-02-28 Richard Biener <rguenther@suse.de>
18009 PR tree-optimization/84584
18010 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18011 SCoPs with fake exit edge.
18013 2018-02-28 Martin Liska <mliska@suse.cz>
18016 * timevar.c (timer::print): Fix format to properly print 100%
18019 2018-02-28 Richard Biener <rguenther@suse.de>
18021 PR middle-end/84607
18022 * genmatch.c (capture_info::walk_match): Do not mark
18023 captured expressions without operands as expr_p given
18024 they act more like predicates and should be subject to
18025 "lost tail" side-effect preserving.
18027 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18029 PR rtl-optimization/81611
18030 * auto-inc-dec.c (attempt_change): Move dead note from
18031 mem_insn if it's the next use of regno
18032 (find_address): Take address use of reg holding
18033 non-incremented value. Add parm to limit search to the named
18035 (merge_in_block): Attempt to use a mem insn that is the next
18036 use of the original regno.
18038 2018-02-27 Martin Sebor <msebor@redhat.com>
18041 * doc/invoke.texi (-Wmissing-attributes): New option.
18042 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18044 2018-02-27 Martin Sebor <msebor@redhat.com>
18046 PR translation/84207
18047 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18048 n argument to unsigned HOST_WIDE_INT.
18049 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18050 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18051 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18052 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18054 2018-02-27 Richard Biener <rguenther@suse.de>
18056 PR tree-optimization/84512
18057 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18058 Do not use the estimate returned from record_stmt_cost for
18059 the scalar iteration cost but sum properly using add_stmt_cost.
18061 2018-02-27 Richard Biener <rguenther@suse.de>
18063 PR tree-optimization/84466
18064 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18065 Adjust last change to less strictly validate use operands.
18067 2018-02-27 Martin Liska <mliska@suse.cz>
18069 PR gcov-profile/84548
18070 * gcov.c (process_file): Allow partial overlap and consider it
18071 also as group functions.
18072 (output_lines): Properly calculate range of lines for a group.
18074 2018-02-27 Martin Liska <mliska@suse.cz>
18076 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18077 'ggc' suffixes. Change first column width.
18078 (timer::print): Fix formatting of the column.
18080 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18082 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18083 preserve inline entry blocks for the sake of debug inline
18084 entry point markers alone.
18085 (remove_unused_locals): Suggest in comments a better place to
18086 force the preservation of inline entry blocks that are
18087 otherwise unused, but do not preserve them.
18089 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18091 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18093 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18096 * config/i386/constraints.md (Bs): Replace
18097 ix86_indirect_branch_register with
18098 TARGET_INDIRECT_BRANCH_REGISTER.
18100 * config/i386/i386.md (indirect_jump): Likewise.
18101 (tablejump): Likewise.
18102 (*sibcall_memory): Likewise.
18103 (*sibcall_value_memory): Likewise.
18104 Peepholes of indirect call and jump via memory: Likewise.
18105 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18106 (*sibcall_value_GOT_32): Likewise.
18107 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18108 (GOT_memory_operand): Likewise.
18109 (call_insn_operand): Likewise.
18110 (sibcall_insn_operand): Likewise.
18111 (GOT32_symbol_operand): Likewise.
18112 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18114 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18116 PR rtl-optimization/83496
18117 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18118 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18119 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18120 redundant insn, if any.
18121 (relax_delay_slots): Likewise.
18122 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18124 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18126 PR tree-optimization/83965
18127 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18128 that grouped statements are part of a reduction chain. Return
18129 true if the statement is not marked as a reduction itself but
18130 is part of a group.
18131 (vect_recog_dot_prod_pattern): Don't check whether the statement
18132 is part of a group here.
18133 (vect_recog_sad_pattern): Likewise.
18134 (vect_recog_widen_sum_pattern): Likewise.
18136 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18139 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18141 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18144 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18146 (ix86_output_indirect_function_return): New prototype.
18147 (ix86_split_simple_return_pop_internal): Likewise.
18148 * config/i386/i386.c (indirect_return_via_cx): New.
18149 (indirect_return_via_cx_bnd): Likewise.
18150 (indirect_thunk_name): Handle return va CX_REG.
18151 (output_indirect_thunk_function): Create alias for
18152 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18153 (ix86_output_indirect_jmp): Remove the bool argument.
18154 (ix86_output_indirect_function_return): New function.
18155 (ix86_split_simple_return_pop_internal): Likewise.
18156 * config/i386/i386.md (*indirect_jump): Don't pass false
18157 to ix86_output_indirect_jmp.
18158 (*tablejump_1): Likewise.
18159 (simple_return_pop_internal): Change it to define_insn_and_split.
18160 Call ix86_split_simple_return_pop_internal to split it for
18161 -mfunction-return=.
18162 (simple_return_indirect_internal): Call
18163 ix86_output_indirect_function_return instead of
18164 ix86_output_indirect_jmp.
18166 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18169 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18170 memset and value initialization afterwards.
18172 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18174 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18176 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18179 * common/config/aarch64/aarch64-common.c
18180 (aarch_option_optimization_table[]): Switch
18181 off fomit-frame-pointer
18183 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18184 Chung-Ju Wu <jasonwucj@gmail.com>
18186 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18188 (store_multiple): Ditto.
18190 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18192 * config.gcc: Add --with-cpu support for nds32 target.
18193 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18194 * config/nds32/nds32.opt: Add -mcpu= option.
18196 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18198 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18199 isel=yes): Warn for these deprecated options.
18201 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18203 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18204 ISA_2_5_MASKS_EMBEDDED.
18206 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18208 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18209 p->max as pointers rather than using iterative_hash_expr.
18211 2018-02-23 Carl Love <cel@us.ibm.com>
18213 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18214 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18216 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18217 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18218 P8V_BUILTIN_VEC_VUNSIGNED2.
18220 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18223 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18224 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18226 * lra-constraints.c (curr_insn_transform): Set up
18227 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18229 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18230 * lra-eliminations.c (spill_pseudos): Ditto.
18231 (process_insn_for_elimination): Ditto.
18232 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18233 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18234 LRA_NON_CLOBBERED_ALT.
18236 2018-02-22 Martin Sebor <msebor@redhat.com>
18238 PR tree-optimization/84480
18239 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18240 to maybe_diag_stxncpy_trunc. Call it.
18241 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18242 from gimple_fold_builtin_strcpy. Print inlining stack.
18243 (handle_builtin_stxncpy): Print inlining stack.
18244 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18246 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18249 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18250 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18251 and -fcheck-pointer-bounds are used together.
18252 (indirect_thunk_prefix): New enum.
18253 (indirect_thunk_need_prefix): New function.
18254 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18255 "_nt" instead of "_bnd" for NOTRACK prefix.
18256 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18257 (output_indirect_thunk_function): Likewise.
18259 (ix86_code_end): Update output_indirect_thunk_function calls.
18260 (ix86_output_indirect_branch_via_reg): Replace
18261 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18262 (ix86_output_indirect_branch_via_push): Likewise.
18263 (ix86_output_function_return): Likewise.
18264 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18265 incompatible with -fcf-protection=branch and
18266 -fcheck-pointer-bounds.
18268 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18271 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18272 Change gcc_assert call to output_operand_lossage.
18274 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18276 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18278 2018-02-22 DJ Delorie <dj@redhat.com>
18279 Sebastian Perta <sebastian.perta@renesas.com>
18280 Oleg Endo <olegendo@gcc.gnu.org>
18282 * config/rx/rx.c (rx_rtx_costs): New function.
18283 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18285 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18287 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18289 2018-02-22 Martin Liska <mliska@suse.cz>
18292 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18293 Add "native" as a possible value.
18295 2018-02-22 Martin Liska <mliska@suse.cz>
18298 * config/i386/i386.c (ix86_option_override_internal):
18299 Add "native" as a possible value for -march and -mtune.
18301 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18304 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18305 to all type variants.
18307 PR tree-optimization/84503
18308 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18309 width as info->bitpos + info->bitsize - start.
18310 (merged_store_group::merge_overlapping): Simplify width computation.
18311 (check_no_overlap): New function.
18312 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18313 start + width and last_order of the group, fail if check_no_overlap
18315 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18316 to group if check_no_overlap fails.
18318 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18320 * config/rs6000/altivec.md: Delete contraint arguments to
18321 define_expand, define_split, and define_peephole2, and in
18322 define_insn_and_split if always unused.
18323 * config/rs6000/darwin.md: Ditto.
18324 * config/rs6000/dfp.md: Ditto.
18325 * config/rs6000/rs6000.md: Ditto.
18326 * config/rs6000/sync.md: Ditto.
18327 * config/rs6000/vector.md: Ditto.
18328 * config/rs6000/vsx.md: Ditto.
18330 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18332 * config/rs6000/altivec.md: Write output control strings as braced
18333 blocks instead of double-quoted strings.
18334 * config/rs6000/darwin.md: Ditto.
18335 * config/rs6000/rs6000.md: Ditto.
18336 * config/rs6000/vector.md: Ditto.
18337 * config/rs6000/vsx.md: Ditto.
18339 2018-02-21 Jason Merrill <jason@redhat.com>
18341 PR c++/84314 - ICE with templates and fastcall attribute.
18342 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18344 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18346 * ipa-cp.c (determine_versionability): Fix comment typos.
18348 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18351 * ipa-cp.c (determine_versionability): Do not version functions caling
18354 2018-02-21 Martin Liska <mliska@suse.cz>
18357 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18358 Add "native" as a possible value.
18359 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18360 the macro when native cpu detection is available.
18362 2018-02-21 Martin Liska <mliska@suse.cz>
18365 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18366 Add "native" as a possible value.
18367 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18368 when native cpu detection is available.
18370 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18371 Martin Sebor <msebor@redhat.com>
18373 PR tree-optimization/84478
18374 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18376 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18377 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18378 support which is conservatively correct, for 2 only stay conservative
18379 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18380 argument to the 2 argument get_range_strlen, adjust 6 arg
18381 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18383 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18384 (gimple_fold_builtin_strlen): Pass true as last argument to
18387 2018-02-20 Martin Sebor <msebor@redhat.com>
18389 PR middle-end/84095
18390 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18391 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18392 (builtin_memref::builtin_memref): Factor out parts into
18393 set_base_and_offset and call it.
18395 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18397 PR middle-end/84406
18398 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18399 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18400 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18401 search at the associated MODE_INT.
18403 2018-02-20 Jeff Law <law@redhat.com>
18405 PR middle-end/82123
18406 PR tree-optimization/81592
18407 PR middle-end/79257
18408 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18409 for range data rather than using global data.
18410 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18411 range data rather than using global data.
18412 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18413 pass it to children as needed.
18414 (struct directive::fmtresult): Similarly.
18415 (struct directive::set_width): Similarly.
18416 (struct directive::set_precision): Similarly.
18417 (format_integer, format_directive, parse_directive): Similarly.
18418 (format_none): Accept unnamed vr_values parameter.
18419 (format_percent, format_floating, format_character): Similarly.
18420 (format_string, format_plain): Similarly.
18421 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18422 the EVRP range analyzer for range data rather than using global data.
18423 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18424 gimple-ssa-evrp-analyze.h
18425 (class sprintf_dom_walker): Add after_dom_children member function.
18426 Add evrp_range_analyzer member.
18427 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18428 range analyzer as needed.
18429 (sprintf_dom_walker::after_dom_children): New member function.
18430 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18432 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18433 (evrp_range_analyzer::pop_to_marker): Likewise.
18435 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18437 PR tree-optimization/84419
18438 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18439 with the required type if its current type is compatible but
18442 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18444 PR middle-end/82004
18445 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18446 after vectorization.
18448 2018-02-20 Martin Liska <mliska@suse.cz>
18451 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18452 possible values if we don't have a hint.
18454 2018-02-20 Martin Liska <mliska@suse.cz>
18458 * final.c (shorten_branches): Build align_tab array with one
18460 * opts.c (finish_options): Add alignment option limit check.
18461 (MAX_CODE_ALIGN): Likewise.
18462 (MAX_CODE_ALIGN_VALUE): Likewise.
18463 * doc/invoke.texi: Document maximum allowed option value for
18464 all -falign-* options.
18466 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18469 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18470 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18471 * var-tracking.c (emit_note_insn_var_location): Remove all references
18472 to NOTE_INSN_CALL_ARG_LOCATION.
18473 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18474 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18475 Use copy_rtx_if_shared.
18476 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18477 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18478 (dwarf2out_var_location): Remove handling of
18479 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18481 * final.c (final_scan_insn): Remove all references to
18482 NOTE_INSN_CALL_ARG_LOCATION.
18483 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18484 before dumping final insns.
18485 * except.c (emit_note_eh_region_end): Remove all references to
18486 NOTE_INSN_CALL_ARG_LOCATION.
18487 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18488 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18489 * config/arc/arc.c (hwloop_optimize): Likewise.
18490 * config/arm/arm.c (create_fix_barrier): Likewise.
18491 * config/s390/s390.c (s390_chunkify_start): Likewise.
18492 * config/sh/sh.c (find_barrier): Likewise.
18493 * config/i386/i386.c (rest_of_insert_endbranch,
18494 ix86_seh_fixup_eh_fallthru): Likewise.
18495 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18496 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18497 * config/frv/frv.c (frv_function_prologue): Likewise.
18498 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18500 (note_outside_basic_block_p): Remove all references to
18501 NOTE_INSN_CALL_ARG_LOCATION.
18502 * gengtype.c (adjust_field_rtx_def): Likewise.
18503 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18505 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18506 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18509 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18512 PR tree-optimization/84452
18513 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18514 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18517 2018-02-19 Martin Liska <mliska@suse.cz>
18520 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18522 2018-02-19 Martin Liska <mliska@suse.cz>
18523 Richard Sandiford <richard.sandiford@linaro.org>
18525 PR tree-optimization/82491
18526 * gimple-fold.c (get_base_constructor): Make earlier bail out
18529 2018-02-19 Carl Love <cel@us.ibm.com>
18531 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18532 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18534 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18535 P8V_BUILTIN_VEC_NEG.
18537 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18539 * config/rl78/rl78.md (movdf): New define expand.
18541 2018-02-19 Martin Liska <mliska@suse.cz>
18544 * doc/invoke.texi: Fix typo.
18545 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18547 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18549 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18550 handle rs6000_single_float and rs6000_double_float specially for
18553 2018-02-16 Jeff Law <law@redhat.com>
18555 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18556 (pop_regs): Use it.
18558 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18561 * ipa-inline.c (inline_small_functions): Fix a typo.
18563 2018-02-16 Nathan Sidwell <nathan@acm.org>
18565 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18567 2018-02-16 Carl Love <cel@us.ibm.com>
18569 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18570 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18571 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18572 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18573 expansion to P8V_BUILTIN_VEC_FLOAT2.
18575 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18577 PR rtl-optimization/70023
18578 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18579 src_regno into account.
18581 2018-02-16 Carl Love <cel@us.ibm.com>
18583 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18584 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18585 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18586 * config/rs6000/rs6000.c: Remove case statements for
18587 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18588 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18589 and P9V_BUILTIN_VEC_VINSERT4B.
18590 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18591 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18592 * config/rs6000/vsx.md:
18593 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18596 2018-02-16 Carl Love <cel@us.ibm.com>
18598 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18600 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18602 * config/rs6000/rs6000-c.c: Add the definitions for
18603 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18604 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18605 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18606 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18607 definition for insert4b and define insn *insert3b_internal.
18608 * doc/extend.texi: Add documentation for vec_extract4b.
18610 2018-02-16 Nathan Sidwell <nathan@acm.org>
18612 * doc/extend.texi (Backwards Compatibility): Mention friend
18613 injection. Note for-scope is deprecated.
18614 * doc/invoke.texi (-ffriend-injection): Deprecate.
18616 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18618 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18619 that moved to I2, also allow destinations that are a paradoxical
18620 subreg (instead of a normal reg).
18622 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18625 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18628 2018-02-16 Richard Biener <rguenther@suse.de>
18630 PR tree-optimization/84037
18631 PR tree-optimization/84016
18633 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18634 Adjust vec_construct for the fact we need additional higher latency
18635 128bit inserts for AVX256 and AVX512 vector builds.
18636 (ix86_add_stmt_cost): Scale vector construction cost for
18639 2018-02-16 Richard Biener <rguenther@suse.de>
18641 PR tree-optimization/84417
18642 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18643 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18644 (non_rewritable_lvalue_p): Likewise, use poly-ints.
18646 2018-02-16 Martin Liska <mliska@suse.cz>
18649 * internal-fn.def (ASAN_CHECK): Set proper flags.
18650 (ASAN_MARK): Likewise.
18652 2018-02-16 Julia Koval <julia.koval@intel.com>
18654 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18655 from PTA_CANNONLAKE.
18657 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18660 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18661 Use ++iter rather than iter++ for std::list iterators.
18662 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
18663 defer deleting them until all nodes in the forest are processed. Do
18664 free even leaf nodes. Change to_process into auto_vec.
18667 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
18668 * vec.h (vec_default_construct): Use memset instead of placement new
18669 if BROKEN_VALUE_INITIALIZATION is defined.
18670 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
18671 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
18674 PR rtl-optimization/83723
18675 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
18676 * lra.c (lra_substitute_pseudo): Likewise. If true, use
18677 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
18679 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
18681 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
18683 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
18685 PR rtl-optimization/81443
18686 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
18687 from inner REGs to paradoxical SUBREGs.
18689 2018-02-16 Richard Biener <rguenther@suse.de>
18691 PR tree-optimization/84399
18692 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18693 For operands we can analyze at their definition make sure we can
18694 analyze them at each use as well.
18696 2018-02-16 Richard Biener <rguenther@suse.de>
18698 PR tree-optimization/84190
18699 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
18700 volatile accesses if the decl isn't volatile.
18702 2018-02-15 Jason Merrill <jason@redhat.com>
18704 PR c++/84314 - ICE with templates and fastcall attribute.
18705 * attribs.c (build_type_attribute_qual_variant): Don't clobber
18706 TYPE_CANONICAL on an existing type.
18708 2018-02-15 Jakub Jelinek <jakub@redhat.com>
18710 PR tree-optimization/84383
18711 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
18712 dstoff nor call operand_equal_p if dstbase is NULL.
18714 PR tree-optimization/84334
18715 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
18716 also a CONSTANT_CLASS_P, punt.
18718 2018-02-14 Jim Wilson <jimw@sifive.com>
18720 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
18721 first SMALL_OPERAND check. New local min_second_step. Move assert
18722 to where locals are set. Add TARGET_RVC support.
18723 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
18725 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
18727 * doc/invoke.texi: Correct -Wformat-overflow code sample.
18729 2018-02-14 Martin Sebor <msebor@redhat.com>
18731 PR tree-optimization/83698
18732 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
18733 arrays constrain the offset range to their bounds.
18734 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
18735 (builtin_access::overlap): Avoid setting the size of overlap if it's
18737 (maybe_diag_overlap): Also consider arrays when deciding what values
18738 of offsets to include in diagnostics.
18740 2018-02-14 Martin Sebor <msebor@redhat.com>
18743 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
18744 that correspond to the kind of a declaration.
18746 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
18749 * config/pa/pa.md: Load address of PIC label using the linkage table
18750 if the label is nonlocal.
18752 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
18754 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
18755 warning message if user requests -maltivec=be.
18756 * doc/invoke.texi: Document deprecation of -maltivec=be.
18758 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
18761 * config/rs6000/rs6000-c.c: Update definitions for
18762 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
18763 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
18765 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18768 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
18769 add _get_ssp intrinsics. Remove argument from
18770 __builtin_ia32_rdssp[d|q].
18771 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
18772 * config/i386/i386-builtin.def: Remove argument from
18773 __builtin_ia32_rdssp[d|q].
18774 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
18775 ix86_expand_special_args_builtin for _rdssp[d|q].
18776 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
18777 Clear register before usage.
18778 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
18779 Add documentation for new _get_ssp and _inc_ssp intrinsics.
18781 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
18783 PR tree-optimization/84357
18784 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18785 operand 1 of an ARRAY_REF too.
18787 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
18790 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18791 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18793 (set_of_reg): New struct.
18794 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18795 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18796 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18798 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18799 Split into bitclr, bitset, bitinvert patterns if appropriate.
18800 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18801 use rx_fuse_in_memory_bitop.
18802 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18803 to named insn, correct maximum insn length.
18805 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18808 * machmode.def: Define a complex mode for PARTIAL_INT.
18809 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18811 * doc/rtl.texi: Document CSPImode.
18812 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18814 (msp430_hard_regno_nregs_with_padding): Likewise.
18816 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18819 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18821 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18823 PR rtl-optimization/84169
18824 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18825 we generated a parallel as new i3 and we split that to new i2 and i3
18826 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18827 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18828 those to i2, not i1. Partially rewrite this scan code.
18830 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18833 * stor-layout.c (place_field): For variable length fields, adjust
18834 offset_align afterwards not just based on the field's alignment,
18835 but also on the size.
18837 PR middle-end/84309
18838 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18839 of exps and logs in the use_exp2 case.
18841 2018-02-13 Jeff Law <law@redhat.com>
18843 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18844 entry for "vector".
18846 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18849 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18853 * common.opt (gas-loc-support, gas-locview-support): New.
18854 (ginline-points, ginternal-reset-location-views): New.
18855 * doc/invoke.texi: Document them. Use @itemx where intended.
18856 (gvariable-location-views): Adjust.
18857 * target.def (reset_location_view): New.
18858 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18859 (TARGET_RESET_LOCATION_VIEW): New.
18860 * doc/tm.texi: Rebuilt.
18861 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18862 (dwarf2out_default_as_locview_support): New.
18863 (output_asm_line_debug_info): Use option variables.
18864 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18865 (output_loc_list): Likewise.
18866 (add_high_low_attributes): Check option variables.
18867 Don't output entry view attribute in strict mode.
18868 (gen_inlined_subroutine_die): Check option variables.
18869 (dwarf2out_inline_entry): Likewise.
18870 (init_sections_and_labels): Likewise.
18871 (dwarf2out_early_finish): Likewise.
18872 (maybe_reset_location_view): New, from...
18873 (dwarf2out_var_location): ... here. Call it.
18874 * debug.h (dwarf2out_default_as_loc_support): Declare.
18875 (dwarf2out_default_as_locview_support): Declare.
18876 * hooks.c (hook_int_rtx_insn_0): New.
18877 * hooks.h (hook_int_rtx_insn_0): Declare.
18878 * toplev.c (process_options): Take -gas-loc-support and
18879 -gas-locview-support from dwarf2out. Enable
18880 -gvariable-location-views by default only with locview
18881 assembler support. Enable -ginternal-reset-location-views by
18882 default only if the target defines the corresponding hook.
18883 Enable -ginline-points by default if location views are
18884 enabled; force it disabled if statement frontiers are
18886 * tree-inline.c (expand_call_inline): Check option variables.
18887 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18889 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18891 PR tree-optimization/84321
18892 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18893 handling. Also check whether the anti-range contains any values
18894 that satisfy the mask; switch to a VR_RANGE if not.
18896 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18899 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18901 2018-02-13 Martin Jambor <mjambor@suse.cz>
18904 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18905 of call statements, also set location of a load to a temporary.
18907 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18909 * config/rl78/rl78.c (add_vector_labels): New function.
18910 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18911 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18912 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18913 which checks that no arguments are passed.
18914 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18915 * doc/extend.texi: Documentation for the new attribute.
18917 2018-02-13 Andreas Schwab <schwab@suse.de>
18919 * config/riscv/linux.h (CPP_SPEC): Define.
18921 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18924 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18925 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18926 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18927 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18928 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18929 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18930 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18931 temporarily for AES and PCLMUL builtins.
18933 PR tree-optimization/84339
18934 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18935 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18938 PR middle-end/84309
18939 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18940 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18941 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18943 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18945 * omp-simd-clone.h: New file.
18946 * omp-simd-clone.c: Include omp-simd-clone.h.
18947 (expand_simd_clones): No longer static.
18948 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18949 cgraph.h and omp-simd-clone.h.
18950 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18951 (vect_recog_widen_shift_pattern): Formatting fix.
18952 (vect_pattern_recog_1): Don't check optab for calls.
18955 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18956 operands[2] into a REG before using gen_lowpart on it.
18958 2018-02-12 Jeff Law <law@redhat.com>
18961 * config/sh/sh.c (find_barrier): Consider a sibling call
18964 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18965 successfully back substituting a reg.
18967 2018-02-12 Richard Biener <rguenther@suse.de>
18969 PR tree-optimization/84037
18970 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18971 parameter, move visited init to caller.
18972 (vect_slp_analyze_operations): Separate cost from validity
18973 check, initialize visited once for all instances.
18974 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18976 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18978 (vect_model_promotion_demotion_cost): Likewise.
18979 (vectorizable_bswap): Guard cost modeling with !slp_node
18980 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18982 (vectorizable_call): Likewise.
18983 (vectorizable_conversion): Likewise.
18984 (vectorizable_assignment): Likewise.
18985 (vectorizable_shift): Likewise.
18986 (vectorizable_operation): Likewise.
18987 (vectorizable_store): Likewise.
18988 (vectorizable_load): Likewise.
18989 (vectorizable_condition): Likewise.
18990 (vectorizable_comparison): Likewise.
18992 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18995 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18996 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18997 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18999 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19001 PR middle-end/83665
19002 * params.def (inline-min-speedup): Increase from 8 to 15.
19003 (max-inline-insns-auto): Decrease from 40 to 30.
19004 * ipa-split.c (consider_split): Add some buffer for function to
19005 be considered inlining candidate.
19006 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19009 2018-02-12 Richard Biener <rguenther@suse.de>
19011 PR tree-optimization/84037
19012 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19013 matched stmts if we cannot swap the non-matched ones.
19015 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19017 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19018 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19019 _mm_maskz_scalef_round_ss): New intrinsics.
19020 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19021 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19022 __builtin_ia32_scalefss_round): Remove.
19023 (__builtin_ia32_scalefsd_mask_round,
19024 __builtin_ia32_scalefss_mask_round): New intrinsics.
19025 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19026 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19027 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19028 "<round_constraint>")): Changed to ...
19029 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19030 "<round_scalar_constraint>")): ... this.
19031 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19032 %0, %1, %2<round_op3>}"): Changed to ...
19033 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19034 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19035 %2<round_scalar_mask_op3>}"): ... this.
19036 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19038 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19040 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19041 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19042 (_mm_maskz_sqrt_round_ss): New intrinsics.
19043 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19044 (__builtin_ia32_sqrtsd_mask_round)
19045 (__builtin_ia32_sqrtss_mask_round): New builtins.
19046 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19047 (__builtin_ia32_sqrtss_round): Remove.
19048 (__builtin_ia32_sqrtsd_mask_round)
19049 (__builtin_ia32_sqrtss_mask_round): New builtins.
19050 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19051 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19052 ((match_operand:VF_128 1 "vector_operand"
19053 "xBm,<round_constraint>")): Changed to ...
19054 ((match_operand:VF_128 1 "vector_operand"
19055 "xBm,<round_scalar_constraint>")): ... this.
19056 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19057 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19058 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19059 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19060 %<iptr>1<round_scalar_mask_op3>}): ... this.
19061 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19062 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19064 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19067 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19068 Cast vec_cmpeq result to correct type.
19069 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19070 Cast vec_cmpgt result to correct type.
19072 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19074 * final.c (final_scan_insn_1): Renamed from...
19075 (final_scan_insn): ... this. New wrapper, to recover
19076 seen from the outermost call in recursive ones.
19077 * config/sparc/sparc.c (output_return): Drop seen from call.
19078 (output_sibcall): Likewise.
19079 * config/visium/visium.c (output_branch): Likewise.
19081 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19083 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19086 2018-02-10 Alan Modra <amodra@gmail.com>
19089 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19090 Specify LR as an input.
19092 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19095 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19096 remove_member_access_dummy_vars): New functions.
19097 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19098 lower_omp_1, execute_lower_omp): Use them.
19100 PR rtl-optimization/84308
19101 * shrink-wrap.c (spread_components): Release todo vector.
19103 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19105 PR rtl-optimization/57193
19106 * ira-color.c (struct allocno_color_data): Add member
19107 conflict_allocno_hard_prefs.
19108 (update_conflict_allocno_hard_prefs): New.
19109 (bucket_allocno_compare_func): Add a preference based on
19110 conflict_allocno_hard_prefs.
19111 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19112 (color_allocnos): Remove a dead code. Initiate
19113 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19115 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19118 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19119 constraint from =wa to wa. Avoid a subreg on the output operand,
19120 instead use a pseudo and subreg it in a move.
19121 (p9_xxbrd_<mode>): Changed to ...
19122 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19123 (p9_xxbrd_v2df): New expander.
19124 (p9_xxbrw_<mode>): Changed to ...
19125 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19126 (p9_xxbrw_v4sf): New expander.
19128 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19130 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19132 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19135 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19136 multiply in 32-bit mode.
19137 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19138 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19141 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19143 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19144 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19145 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19146 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19148 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19151 * dwarf2out.c (is_trivial_indirect_ref): New function.
19152 (dwarf2out_late_global_decl): Do not generate a location
19153 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19154 and that are not defined in the current unit.
19156 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19158 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19159 instead of a libcall for UNORDERED.
19161 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19164 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19165 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19167 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19170 * config/s390/s390.c (s390_set_current_function): Invoke
19171 s390_indirect_branch_settings also if fndecl didn't change.
19173 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19175 * config/rs6000/rs6000.md (blockage): Set length to zero.
19177 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19179 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19181 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19184 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19185 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19189 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19190 PARALLEL incoming that failed vt_get_decl_and_offset check.
19192 PR middle-end/84237
19193 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19194 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19196 (get_variable_section): For decls in named .bss* sections pass true as
19197 second argument to bss_initializer_p.
19199 2018-02-09 Marek Polacek <polacek@redhat.com>
19200 Jakub Jelinek <jakub@redhat.com>
19203 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19204 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19205 Sync some changes from cxx_fold_indirect_ref.
19207 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19209 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19211 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19212 (BLOCK_INLINE_ENTRY_LABEL): New.
19213 (dwarf2out_var_location): Disregard inline entry markers.
19214 (inline_entry_data): New struct.
19215 (inline_entry_data_hasher): New hashtable type.
19216 (inline_entry_data_hasher::hash): New.
19217 (inline_entry_data_hasher::equal): New.
19218 (inline_entry_data_table): New variable.
19219 (add_high_low_attributes): Add DW_AT_entry_pc and
19220 DW_AT_GNU_entry_view attributes if a pending entry is found
19221 in inline_entry_data_table. Add old entry_pc attribute only
19222 if debug nonbinding markers are disabled.
19223 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19224 markers are enabled.
19225 (block_within_block_p, dwarf2out_inline_entry): New.
19226 (dwarf2out_finish): Check that no entries remained in
19227 inline_entry_data_table.
19228 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19229 (final_scan_insn, notice_source_line): Likewise.
19230 (rest_of_clean_state): Skip inline entry markers.
19231 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19233 * gimple.c (gimple_build_debug_inline_entry): New.
19234 * gimple.h (enum gimple_debug_subcode): Add
19235 GIMPLE_DEBUG_INLINE_ENTRY.
19236 (gimple_build_debug_inline_entry): Declare.
19237 (gimple_debug_inline_entry_p): New.
19238 (gimple_debug_nonbind_marker_p): Adjust.
19239 * insn-notes.def (INLINE_ENTRY): New.
19240 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19241 inline entry marker notes.
19242 (print_insn): Likewise.
19243 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19244 (INSN_DEBUG_MARKER_KIND): Likewise.
19245 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19246 * tree-inline.c (expand_call_inline): Build and insert
19247 debug_inline_entry stmt.
19248 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19249 inline entry blocks early, if nonbind markers are enabled.
19250 (dump_scope_block): Dump fragment info.
19251 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19252 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19253 (gimple_build_debug_inline_entry): New.
19254 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19255 Enable/disable inline entry points too.
19256 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19257 (DEBUG_INSN): Describe inline entry markers.
19259 * common.opt (gvariable-location-views): New.
19260 (gvariable-location-views=incompat5): New.
19261 * config.in: Rebuilt.
19262 * configure: Rebuilt.
19263 * configure.ac: Test assembler for view support.
19264 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19265 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19266 * dwarf2out.c (var_loc_view): New typedef.
19267 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19268 (dwarf2out_locviews_in_attribute): New.
19269 (dwarf2out_locviews_in_loclist): New.
19270 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19271 (enum dw_line_info_opcode): Add LI_adv_address.
19272 (struct dw_line_info_table): Add view.
19273 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19274 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19275 (zero_view_p): New variable.
19276 (ZERO_VIEW_P): New macro.
19277 (output_asm_line_debug_info): New.
19278 (struct var_loc_node): Add view.
19279 (add_AT_view_list, AT_loc_list): New.
19280 (add_var_loc_to_decl): Add view param. Test it against last.
19281 (new_loc_list): Add view params. Record them.
19282 (AT_loc_list_ptr): Handle loc and view lists.
19283 (view_list_to_loc_list_val_node): New.
19284 (print_dw_val): Handle dw_val_class_view_list.
19285 (size_of_die): Likewise.
19286 (value_format): Likewise.
19287 (loc_list_has_views): New.
19288 (gen_llsym): Set vl_symbol too.
19289 (maybe_gen_llsym, skip_loc_list_entry): New.
19290 (dwarf2out_maybe_output_loclist_view_pair): New.
19291 (output_loc_list): Output view list or entries too.
19292 (output_view_list_offset): New.
19293 (output_die): Handle dw_val_class_view_list.
19294 (output_dwarf_version): New.
19295 (output_compilation_unit_header): Use it.
19296 (output_skeleton_debug_sections): Likewise.
19297 (output_rnglists, output_line_info): Likewise.
19298 (output_pubnames, output_aranges): Update version comments.
19299 (output_one_line_info_table): Output view numbers in asm comments.
19300 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19301 Call maybe_gen_llsym.
19302 (loc_list_from_tree_1): Adjust.
19303 (add_AT_location_description): Create view list attribute if
19304 needed, check it's absent otherwise.
19305 (convert_cfa_to_fb_loc_list): Adjust.
19306 (maybe_emit_file): Call output_asm_line_debug_info for test.
19307 (dwarf2out_var_location): Reset views as needed. Precompute
19308 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19309 attribute. Set view.
19310 (new_line_info_table): Reset next view.
19311 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19312 (dwarf2out_source_line): Likewise. Output view resets and labels to
19313 the assembler, or select appropriate line info opcodes.
19314 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19315 (optimize_string_length): Catch it. Adjust.
19316 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19317 dw_val_class_view_list, and remove it if no longer needed.
19318 (hash_loc_list): Hash view numbers.
19319 (loc_list_hasher::equal): Compare them.
19320 (optimize_location_lists): Check whether a view list symbol is
19321 needed, and whether the locview attribute is present, and
19322 whether they match. Remove the locview attribute if no longer
19324 (index_location_lists): Call skip_loc_list_entry for test.
19325 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19326 Use output_dwarf_version.
19327 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19328 (struct dw_val_node): Add val_view_list.
19329 * final.c (SEEN_NEXT_VIEW): New.
19330 (set_next_view_needed): New.
19331 (clear_next_view_needed): New.
19332 (maybe_output_next_view): New.
19333 (final_start_function): Rename to...
19334 (final_start_function_1): ... this. Take pointer to FIRST,
19335 add SEEN parameter. Emit param bindings in the initial view.
19336 (final_start_function): Reintroduce SEEN-less interface.
19337 (final): Rename to...
19338 (final_1): ... this. Take SEEN parameter. Output final pending
19339 next view at the end.
19340 (final): Reintroduce seen-less interface.
19341 (final_scan_insn): Output pending next view before switching
19342 sections or ending a block. Mark the next view as needed when
19343 outputting variable locations. Notify debug backend of section
19344 changes, and of location view changes.
19345 (rest_of_handle_final): Adjust.
19346 * toplev.c (process_options): Autodetect value for debug variable
19347 location views option. Warn on incompat5 without -gdwarf-5.
19348 * doc/invoke.texi (gvariable-location-views): New.
19349 (gvariable-location-views=incompat5): New.
19350 (gno-variable-location-views): New.
19352 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19354 PR tree-optimization/84136
19355 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19356 that the result of find_edge is non-NULL.
19358 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19361 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19362 storing integer register in SImode. Fix cost of 256 and 512
19363 byte aligned SSE register store.
19365 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19367 * config/i386/i386.c (ix86_multiplication_cost): Fix
19368 multiplication cost for TARGET_AVX512DQ.
19370 2018-02-08 Marek Polacek <polacek@redhat.com>
19372 PR tree-optimization/84238
19373 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19376 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19378 PR tree-optimization/84265
19379 * tree-vect-stmts.c (vectorizable_store): Don't treat
19380 VMAT_CONTIGUOUS accesses as grouped.
19381 (vectorizable_load): Likewise.
19383 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19385 PR tree-optimization/81635
19386 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19387 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19388 (test_round_for_mask): New functions.
19389 (wide_int_cc_tests): Call test_round_for_mask.
19390 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19391 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19392 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19393 range returned by get_range_info.
19395 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19398 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19399 * symtab.c: Include builtins.h
19400 (symtab_node::output_to_lto_symbol_table_p): Move here
19401 from lto-streamer-out.c:output_symbol_p.
19402 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19403 (output_symbol_p): Move all logic to symtab.c
19404 (produce_symtab): Update.
19406 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19408 * config/s390/s390-opts.h (enum indirect_branch): Define.
19409 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19410 (s390_indirect_branch_via_thunk)
19411 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19412 (enum s390_indirect_branch_type): Define.
19413 * config/s390/s390.c (struct s390_frame_layout, struct
19414 machine_function): Remove.
19415 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19416 (indirect_branch_table_label_no, indirect_branch_table_name):
19418 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19419 (enum s390_indirect_branch_option): Define.
19420 (s390_return_addr_from_memory): New function.
19421 (s390_handle_string_attribute): New function.
19422 (s390_attribute_table): Add new attribute handler.
19423 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19424 (s390_indirect_branch_via_thunk): New function.
19425 (s390_indirect_branch_via_inline_thunk): New function.
19426 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19427 sibling call optimization for non z10 compiles.
19428 (s390_emit_call): Force indirect branch target to be a single
19429 register. Add r1 clobber for non-z10 compiles.
19430 (s390_emit_epilogue): Emit return jump via return_use expander.
19431 (s390_reorg): Handle JUMP_INSNs as execute targets.
19432 (s390_option_override_internal): Perform validity checks for the
19433 new command line options.
19434 (s390_indirect_branch_attrvalue): New function.
19435 (s390_indirect_branch_settings): New function.
19436 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19437 (s390_output_indirect_thunk_function): New function.
19438 (s390_code_end): Implement target hook.
19439 (s390_case_values_threshold): Implement target hook.
19440 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19442 * config/s390/s390.h (struct s390_frame_layout)
19443 (struct machine_function): Move here from s390.c.
19444 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19445 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19446 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19447 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19448 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19449 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19450 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19451 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19452 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19453 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19454 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19455 (mnemonic attribute): Add values which aren't recognized
19457 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19458 pattern for branch conversion. Fix mnemonic attribute.
19459 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19460 indirect branch via thunk if requested.
19461 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19462 ("*indirect_jump"): Disable for branch conversion using out of
19464 ("indirect_jump_via_thunk<mode>_z10")
19465 ("indirect_jump_via_thunk<mode>")
19466 ("indirect_jump_via_inlinethunk<mode>_z10")
19467 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19468 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19469 ("casesi_jump_via_inlinethunk<mode>_z10")
19470 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19471 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19472 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19473 ("*indirect2_jump"): Disable for branch conversion.
19474 ("casesi_jump"): Turn into expander and expand patterns for branch
19476 ("return_use"): New expander.
19477 ("*return"): Emit return via thunk and rename it to ...
19478 ("*return<mode>"): ... this one.
19479 * config/s390/s390.opt: Add new options and and enum for the
19482 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19484 * lra-constraints.c (match_reload): Unconditionally use
19485 gen_lowpart_SUBREG, rather than selecting between that
19486 and equivalent gen_rtx_SUBREG code.
19488 2018-02-08 Richard Biener <rguenther@suse.de>
19490 PR tree-optimization/84233
19491 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19492 changed flag instead of boguously re-using phi_inserted.
19494 2018-02-08 Martin Jambor <mjambor@suse.cz>
19496 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19497 static local variables.
19499 2018-02-08 Richard Biener <rguenther@suse.de>
19501 PR tree-optimization/84278
19502 * tree-vect-stmts.c (vectorizable_store): When looking for
19503 smaller vector types to perform grouped strided loads/stores
19504 make sure the mode is supported by the target.
19505 (vectorizable_load): Likewise.
19507 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19509 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19510 Increase LDP/STP opportunities by adding adjacent callee-saves.
19512 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19514 PR rtl-optimization/84068
19515 PR rtl-optimization/83459
19516 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19518 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19520 PR tree-optimization/84224
19521 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19522 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19523 non-zero arguments.
19525 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19528 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19529 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19530 count, remove one USE.
19532 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19534 * doc/install.texi (Configuration): Document the
19535 --with-long-double-format={ibm,ieee} PowerPC configuration
19539 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19540 Convert from define_expand to be define_insn_and_split. Rework
19541 float/double/_Float128 conversions to QI/HI/SImode to work with
19542 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19543 conversions to QI/HImode types did a store and then a load to
19544 truncate the value. For conversions to VSX registers, don't split
19545 the insn, instead emit the code directly. Use the code iterator
19546 any_fix to combine signed and unsigned conversions.
19547 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19548 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19549 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19550 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19551 (fix_<mode>di2_hw): Likewise.
19552 (fixuns_<mode>di2_hw): Likewise.
19553 (fix_<mode>si2_hw): Likewise.
19554 (fixuns_<mode>si2_hw): Likewise.
19555 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19556 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19557 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19558 fix<uns>_trunc<SFDF:mode>si2_p8.
19559 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19560 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19561 (fix<uns>_<mode>_mem): Likewise.
19562 (fctiw<u>z_<mode>_mem): Likewise.
19563 (fix<uns>_<mode>_mem): Likewise.
19564 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19565 the register allocator from doing a direct move to the GPRs to do
19566 a store, and instead use the ISA 3.0 store byte/half-word from
19567 vector register instruction. For IEEE 128-bit floating point,
19568 also optimize stores of 32-bit ints.
19569 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19571 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19573 * genextract.c (push_pathstr_operand): New function to support
19575 (walk_rtx): Call push_pathstr_operand.
19576 (print_path): Support [a-zA-Z].
19578 2018-02-07 Richard Biener <rguenther@suse.de>
19580 PR tree-optimization/84037
19581 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19582 (cse_and_gimplify_to_preheader): Declare.
19583 (vect_get_place_in_interleaving_chain): Likewise.
19584 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19586 (_loop_vec_info::~_loop_vec_info): Delete it.
19587 (cse_and_gimplify_to_preheader): New function.
19588 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19589 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19590 (vectorizable_load): Likewise. For grouped stores always base
19591 the IV on the first element.
19592 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19593 condition before gimplifying.
19595 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19597 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19598 *DIV_EXPR and *MOD_EXPR.
19600 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19603 * config/i386/i386.c (ix86_option_override_internal): Mask out
19604 the CF_SET bit when checking -fcf-protection.
19606 2018-02-07 Tom de Vries <tom@codesourcery.com>
19609 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19612 2018-02-07 Richard Biener <rguenther@suse.de>
19614 PR tree-optimization/84204
19615 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19618 PR tree-optimization/84205
19619 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19620 special-case isl_ast_op_zdiv_r.
19622 PR tree-optimization/84223
19623 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19624 Only add conditions from within the region.
19625 (gather_bbs::after_dom_children): Adjust.
19627 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19630 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19631 * config/avr/avr.md: Only post-reload split REG-REG moves if
19632 either register is GENERAL_REG_P.
19634 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19636 PR tree-optimization/84235
19637 * tree-ssa-scopedtables.c
19638 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19639 if the subtraction is performed in floating point type where NaNs are
19640 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19641 build 1. Formatting fix.
19643 2018-02-06 Jakub Jelinek <jakub@redhat.com>
19646 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19647 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19648 and skip it regardless of bb boundaries. Use CALL_P macro,
19649 don't test INSN_P (insn) together with CALL_P or JUMP_P check
19650 unnecessarily, formatting fix.
19652 2018-02-06 Michael Collison <michael.collison@arm.com>
19654 * config/arm/thumb2.md:
19655 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19656 (*thumb_mov_notscc): Ditto.
19658 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
19661 * config/rs6000/rs6000.md (su code attribute): Use "u" for
19662 unsigned_fix, not "s".
19664 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19666 * configure.ac (gcc_fn_eh_frame_ro): New function.
19667 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
19668 correct .eh_frame permissions.
19669 * configure: Regenerate.
19671 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
19673 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
19674 irrelevant options.
19676 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19678 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19679 Display warning message for -mno-speculate-indirect-jumps.
19681 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
19683 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
19685 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
19687 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
19689 PR tree-optimization/84225
19690 * tree-eh.c (find_trapping_overflow): Only call
19691 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
19693 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19696 * config/i386/i386.c: Reimplement the check of possible options
19697 -mibt/-mshstk conbination. Change error messages.
19698 * doc/invoke.texi: Fix a typo: remove extra '='.
19700 2018-02-06 Marek Polacek <polacek@redhat.com>
19702 PR tree-optimization/84228
19703 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
19705 2018-02-06 Tamar Christina <tamar.christina@arm.com>
19708 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
19709 emitted arch directives.
19710 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
19711 __ARM_FEATURE_COPROC before changing architectures.
19713 2018-02-06 Richard Biener <rguenther@suse.de>
19715 * config/i386/i386.c (print_reg): Fix typo.
19716 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
19718 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
19720 * configure: Regenerate.
19722 2018-02-05 Martin Sebor <msebor@redhat.com>
19724 PR tree-optimization/83369
19725 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
19728 2018-02-05 Martin Liska <mliska@suse.cz>
19730 * doc/invoke.texi: Cherry-pick upstream r323995.
19732 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
19734 * ira.c (ira_init_register_move_cost): Adjust comment.
19736 2018-02-05 Martin Liska <mliska@suse.cz>
19738 PR gcov-profile/84137
19739 * doc/gcov.texi: Fix typo in documentation.
19741 2018-02-05 Martin Liska <mliska@suse.cz>
19743 PR gcov-profile/83879
19744 * doc/gcov.texi: Document necessity of --dynamic-list-data when
19745 using dlopen functionality.
19747 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
19749 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
19750 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
19751 _mm_maskz_range_ss, _mm_mask_range_round_ss,
19752 _mm_maskz_range_round_ss): New intrinsics.
19753 (__builtin_ia32_rangesd128_round)
19754 (__builtin_ia32_rangess128_round): Remove.
19755 (__builtin_ia32_rangesd128_mask_round,
19756 __builtin_ia32_rangess128_mask_round): New builtins.
19757 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
19758 __builtin_ia32_rangess128_round): Remove.
19759 (__builtin_ia32_rangesd128_mask_round,
19760 __builtin_ia32_rangess128_mask_round): New builtins.
19761 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
19762 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
19763 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
19764 "<round_saeonly_constraint>")): Changed to ...
19765 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
19766 "<round_saeonly_scalar_constraint>")): ... this.
19767 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
19768 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
19769 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
19770 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
19771 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
19773 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
19775 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
19777 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
19778 Remove all values except native, 8540 and 8548.
19780 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
19782 * config/i386/i386.c (ix86_output_function_return): Pass
19783 INVALID_REGNUM, instead of -1, as invalid register number to
19784 indirect_thunk_name and output_indirect_thunk.
19786 2018-02-02 Julia Koval <julia.koval@intel.com>
19788 * config.gcc: Add -march=icelake.
19789 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19790 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19791 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19792 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19793 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19794 (processor_target_table): Add icelake.
19795 (ix86_option_override_internal): Handle new PTAs.
19796 (get_builtin_code_for_version): Handle icelake.
19797 (M_INTEL_COREI7_ICELAKE): New.
19798 (fold_builtin_cpu): Handle icelake.
19799 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19800 * doc/invoke.texi: Add -march=icelake.
19802 2018-02-02 Julia Koval <julia.koval@intel.com>
19804 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19805 to wide_int_bitmask.
19806 * wide-int-bitmask.h: New.
19808 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19811 * config/i386/i386.md: Replace Pmode with word_mode in
19812 builtin_setjmp_setup and builtin_longjmp to support x32.
19814 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19818 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19820 (rs6000_supported_cpu_names): New static variable.
19821 (linux_cpu_translation_table): Likewise.
19822 (elf_platform) <cpu>: Define new static variable and use it.
19823 Translate kernel AT_PLATFORM name to canonical name if needed.
19824 Error if platform name is unknown.
19826 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19829 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19831 2018-02-01 Jeff Law <law@redhat.com>
19834 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19835 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19836 the scratch if RELEASE_VIA_POP is false.
19837 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19838 If we have to save a temporary register, decrement SIZE appropriately.
19839 Pass new arguments to release_scratch_register_on_entry.
19840 (ix86_adjust_stack_and_probe): Likewise.
19841 (ix86_emit_probe_stack_range): Pass new arguments to
19842 release_scratch_register_on_entry.
19844 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19846 PR rtl-optimization/84157
19847 * combine.c (change_zero_ext): Use REG_P predicate in
19848 front of HARD_REGISTER_P predicate.
19850 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19852 * config/avr/avr.c (avr_option_override): Move disabling of
19853 -fdelete-null-pointer-checks to...
19854 * common/config/avr/avr-common.c (avr_option_optimization_table):
19857 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19859 PR tree-optimization/81635
19860 * tree-data-ref.c (split_constant_offset_1): For types that
19861 wrap on overflow, try to use range info to prove that wrapping
19864 2018-02-01 Renlin Li <renlin.li@arm.com>
19867 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19868 TAILCALL_ADDR_REGS.
19869 (aarch64_register_move_cost): Likewise.
19870 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19871 TAILCALL_ADDR_REGS.
19872 (REG_CLASS_NAMES): Likewise.
19873 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19874 TAILCALL_ADDR_REGS. Remove IP registers.
19875 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19877 2018-02-01 Richard Biener <rguenther@suse.de>
19879 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19880 for specifying RPO order and allow NULL for that.
19881 * domwalk.c (dom_walker::dom_walker): Likewise.
19882 (dom_walker::walk): Handle NULL RPO order.
19883 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19885 (rewrite_update_dom_walker): Likewise.
19886 (mark_def_dom_walker): Likewise.
19888 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19890 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19891 (aarch64_maybe_expand_sve_subreg_move): Declare.
19892 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19893 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19895 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19896 that are semantically a reverse operation.
19897 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19898 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19899 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19901 (aarch64_can_change_mode_class): For big-endian, forbid changes
19902 between two SVE modes if they have different element sizes.
19904 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19906 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19907 the TImode handling for big-endian targets.
19909 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19911 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19912 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19914 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19915 Remove BSWAP handing for big-endian targets and use the form of
19916 LD1RQ appropariate for the mode.
19918 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19920 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19921 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19922 duplicated element.
19924 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19927 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19928 check for operands that need to go through aarch64_sve_reload_be.
19930 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19932 PR tree-optimization/81661
19933 PR tree-optimization/84117
19934 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19935 * tree-eh.c: Include gimplify.h.
19936 (find_trapping_overflow, replace_trapping_overflow,
19937 rewrite_to_non_trapping_overflow): New functions.
19938 * tree-vect-loop.c: Include tree-eh.h.
19939 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19940 * tree-data-ref.c: Include tree-eh.h.
19941 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19943 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19945 PR rtl-optimization/84123
19946 * combine.c (change_zero_ext): Check if hard register satisfies
19947 can_change_dest_mode before calling gen_lowpart_SUBREG.
19949 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19952 * ira.c (ira_init_register_move_cost): Remove assert.
19954 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19956 PR rtl-optimization/84071
19957 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19958 * doc/tm.texi: Regenerate.
19960 2018-01-31 Richard Biener <rguenther@suse.de>
19962 PR tree-optimization/84132
19963 * tree-data-ref.c (analyze_miv_subscript): Properly
19964 check whether evolution_function_is_affine_multivariate_p
19965 before calling gcd_of_steps_may_divide_p.
19967 2018-01-31 Julia Koval <julia.koval@intel.com>
19970 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19971 * config/i386/i386.md (rdpid_rex64) New.
19972 (rdpid): Make 32bit only.
19974 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19977 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19978 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19980 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19983 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19985 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19987 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19989 PR rtl-optimization/84071
19990 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19991 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19993 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19995 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19996 (arc_attribute_table): Add 'aux' attribute.
19997 (arc_in_small_data_p): Consider aux like variables.
19998 (arc_is_aux_reg_p): New function.
19999 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20000 (arc_get_aux_arg): New function.
20001 (prepare_move_operands): Handle aux-register access.
20002 (arc_handle_aux_attribute): New function.
20003 * doc/extend.texi (ARC Variable attributes): Add subsection.
20005 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20007 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20008 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20009 (arc_attribute_table): Add 'uncached' attribute.
20010 (arc_print_operand): Print '.di' flag for uncached memory
20012 (arc_in_small_data_p): Do not consider for small data the uncached
20014 (arc_is_uncached_mem_p): New function.
20015 * config/arc/predicates.md (compact_store_memory_operand): Check
20016 for uncached memory accesses.
20017 (nonvol_nonimm_operand): Likewise.
20018 * doc/extend.texi (ARC Type Attribute): New subsection.
20020 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20023 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20024 falign-loops=): Add Optimization flag.
20026 2018-01-30 Jeff Law <law@redhat.com>
20029 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20030 INT_REGISTERS_SAVED. Check it prior to calling
20031 get_scratch_register_on_entry.
20032 (ix86_adjust_stack_and_probe): Similarly.
20033 (ix86_emit_probe_stack_range): Similarly.
20034 (ix86_expand_prologue): Corresponding changes.
20036 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20039 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20040 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20042 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20045 * lra-constraints.c (curr_insn_transform): Process AND in the
20048 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20050 PR rtl-optimization/83986
20051 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20052 dependence against last_pending_memory_flush in addition to
20053 pending_jump_insns.
20055 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20057 PR tree-optimization/81611
20058 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20061 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20064 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20067 2018-01-30 Richard Biener <rguenther@suse.de>
20068 Jakub Jelinek <jakub@redhat.com>
20070 PR tree-optimization/84111
20071 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20072 inner loops added during recursion, as they don't have up-to-date
20075 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20078 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20079 (can_inline_edge_by_limits_p): ... here.
20080 (can_early_inline_edge_p, check_callers,
20081 update_caller_keys, update_callee_keys, recursive_inlining,
20082 add_new_edges_to_heap, speculation_useful_p,
20083 inline_small_functions,
20084 inline_small_functions, flatten_function,
20085 inline_to_all_callers_1): Update.
20087 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20089 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20092 2018-01-30 Richard Biener <rguenther@suse.de>
20094 PR tree-optimization/83008
20095 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20096 invariant and constant vector uses in stmts when they need
20097 more than one stmt.
20099 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20102 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20103 * configure: Regenerate.
20105 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20107 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20109 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20110 Use gen_rtx_REG rather than gen_lowpart.
20112 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20114 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20115 rather than 0 when creating partial subregs.
20117 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20119 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20122 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20125 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20126 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20127 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20128 flags. This restores the settings used before the 2017-07-24.
20129 Turning off pre increment/decrement/modify allows IVOPTS to
20130 optimize DF/SF loops where the index is an int.
20132 2018-01-29 Richard Biener <rguenther@suse.de>
20133 Kelvin Nilsen <kelvin@gcc.gnu.org>
20136 * tree-vect-stmts.c (vectorizable_call): Don't call
20137 targetm.vectorize_builtin_md_vectorized_function if callee is
20140 2018-01-22 Carl Love <cel@us.ibm.com>
20142 * doc/extend.tex: Fix typo in second arg in
20143 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20145 2018-01-29 Richard Biener <rguenther@suse.de>
20147 PR tree-optimization/84086
20148 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20149 (flush_ssaname_freelist): When SSA names were released reset
20150 the SCEV hash table.
20152 2018-01-29 Richard Biener <rguenther@suse.de>
20154 PR tree-optimization/84057
20155 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20156 removed paths when removing edges.
20158 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20160 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20161 -mfunction-return=@var{choice}.
20163 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20165 PR diagnostic/84034
20166 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20167 Handle CR like TAB.
20168 (layout::print_source_line): Likewise.
20169 (test_get_line_width_without_trailing_whitespace): Add test cases.
20171 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20173 PR middle-end/84040
20174 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20177 2018-01-26 Jim Wilson <jimw@sifive.com>
20179 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20181 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20184 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20186 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20187 and CMP + SUB-immediate -> SUBS.
20189 2018-01-26 Martin Sebor <msebor@redhat.com>
20191 PR tree-optimization/83896
20192 * tree-ssa-strlen.c (get_string_len): Rename...
20193 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20194 Avoid assuming length is constant.
20195 (handle_char_store): Use HOST_WIDE_INT for string length.
20197 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20200 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20201 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20203 2018-01-26 Richard Biener <rguenther@suse.de>
20205 PR rtl-optimization/84003
20206 * dse.c (record_store): Only record redundant stores when
20207 the earlier store aliases at least all accesses the later one does.
20209 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20211 PR rtl-optimization/83985
20212 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20213 REG_CFA_RESTORE insns.
20214 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20215 REG_CFA_RESTORE insns here.
20218 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20219 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20221 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20223 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20224 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20225 (arc_init): Likewise.
20226 (arc_override_options): Likewise.
20227 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20229 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20231 * config/arc/arc.h (TARGET_DBNZ): Define.
20232 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20233 properly set the tune attribute.
20234 (dbnz): Use TARGET_DBNZ guard.
20235 * config/arc/arc.opt (mtune): Add core3 option.
20237 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20239 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20240 recognize new pic like addresses.
20241 (arc_delegitimize_address): Clean up.
20243 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20245 * config/arc/arc-arches.def: Option mrf16 valid for all
20247 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20248 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20249 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20250 * config/arc/arc-tables.opt: Regenerate.
20251 * config/arc/arc.c (arc_conditional_register_usage): Handle
20252 reduced register file case.
20253 (arc_file_start): Set must have build attributes.
20254 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20255 mrf16 option value.
20256 * config/arc/arc.opt (mrf16): Add new option.
20257 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20258 * config/arc/genmultilib.awk: Handle new mrf16 option.
20259 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20260 * config/arc/t-multilib: Regenerate.
20261 * doc/invoke.texi (ARC Options): Document mrf16 option.
20263 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20265 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20266 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20267 (arc_attribute_table): Add 'secure_call' attribute.
20268 (arc_print_operand): Print secure call operand.
20269 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20271 (arc_is_secure_call_p): New function. * config/arc/arc.md
20272 (call_i): Add support for sjli instruction.
20273 (call_value_i): Likewise.
20274 * config/arc/constraints.md (Csc): New constraint.
20276 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20277 John Eric Martin <John.Martin@emmicro-us.com>
20279 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20280 * config/arc/arc.c (_arc_jli_section): New struct.
20281 (arc_jli_section): New type.
20282 (rc_jli_sections): New static variable.
20283 (arc_handle_jli_attribute): New function.
20284 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20285 (arc_file_end): New function.
20286 (TARGET_ASM_FILE_END): Define.
20287 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20288 (arc_add_jli_section): New function.
20289 (jli_call_scan): Likewise.
20290 (arc_reorg): Call jli_call_scan.
20291 (arc_output_addsi): Remove 'S' from printing asm operand.
20292 (arc_is_jli_call_p): New function.
20293 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20295 (movhi_insn): Likewise.
20296 (movsi_insn): Likewise.
20297 (movsi_set_cc_insn): Likewise.
20298 (loadqi_update): Likewise.
20299 (load_zeroextendqisi_update): Likewise.
20300 (load_signextendqisi_update): Likewise.
20301 (loadhi_update): Likewise.
20302 (load_zeroextendhisi_update): Likewise.
20303 (load_signextendhisi_update): Likewise.
20304 (loadsi_update): Likewise.
20305 (loadsf_update): Likewise.
20306 (movsicc_insn): Likewise.
20307 (bset_insn): Likewise.
20308 (bxor_insn): Likewise.
20309 (bclr_insn): Likewise.
20310 (bmsk_insn): Likewise.
20311 (bicsi3_insn): Likewise.
20312 (cmpsi_cc_c_insn): Likewise.
20313 (movsi_ne): Likewise.
20314 (movsi_cond_exec): Likewise.
20315 (clrsbsi2): Likewise.
20316 (norm_f): Likewise.
20326 (call_i): Remove 'S' asm letter, add jli instruction.
20327 (call_value_i): Likewise.
20328 * config/arc/arc.op (mjli-always): New option.
20329 * config/arc/constraints.md (Cji): New constraint.
20330 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20332 (subsf3_fpx): Likewise.
20333 (mulsf3_fpx): Likewise.
20334 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20336 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20337 function attrbutes.
20338 * doc/invoke.texi (ARC): Document mjli-always option.
20340 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20342 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20343 avoid addition with 0 and use incw and decw where possible.
20345 2018-01-26 Richard Biener <rguenther@suse.de>
20347 PR tree-optimization/81082
20348 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20349 association if it requires casting to unsigned.
20350 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20351 from fold_plusminus_mult_expr to catch important cases late when
20352 range info is available.
20354 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20356 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20357 * configure.ac (hidden_linkonce): New test.
20358 * configure: Regenerate.
20359 * config.in: Regenerate.
20361 2018-01-26 Julia Koval <julia.koval@intel.com>
20363 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20364 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20365 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20366 _mm_mask_bitshuffle_epi64_mask): Fix type.
20367 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20368 USI_FTYPE_V4DI_V4DI_USI): Remove.
20369 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20370 __builtin_ia32_vpshufbitqmb256_mask,
20371 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20372 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20373 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20375 2018-01-26 Alan Modra <amodra@gmail.com>
20378 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20379 UNSPEC_VBPERMQ. Sort other unspecs.
20381 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20383 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20385 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20387 PR middle-end/83055
20388 * predict.c (drop_profile): Do not push/pop cfun; update also
20390 (handle_missing_profiles): Fix logic looking for zero profiles.
20392 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20394 PR middle-end/83977
20395 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20396 on functions with #pragma omp declare simd or functions with simd
20398 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20399 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20400 Remove trailing \n from warning_at calls.
20402 2018-01-25 Tom de Vries <tom@codesourcery.com>
20405 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20406 for neutered workers.
20408 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20411 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20412 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20414 2018-01-24 Jeff Law <law@redhat.com>
20417 * i386.c (get_probe_interval): Move to earlier point.
20418 (ix86_compute_frame_layout): If -fstack-clash-protection and
20419 the frame is larger than the probe interval, then use pushes
20420 to save registers rather than reg->mem moves.
20421 (ix86_expand_prologue): Remove conditional for int_registers_saved
20424 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20427 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20428 min/max for never referenced object.
20430 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20432 PR middle-end/83977
20433 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20435 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20436 attributes from DECL_ATTRIBUTES (decl) without affecting
20437 DECL_ATTRIBUTES (current_function_decl).
20438 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20439 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20441 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20443 PR tree-optimization/83979
20444 * fold-const.c (fold_comparison): Use constant_boolean_node
20445 instead of boolean_{true,false}_node.
20447 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20449 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20452 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20454 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20455 Simplify the clause that sets the length attribute.
20456 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20457 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20458 clause that sets the length attribute.
20459 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20461 2018-01-24 Tom de Vries <tom@codesourcery.com>
20464 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20465 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20466 Add strict parameter.
20467 (prevent_branch_around_nothing): Insert dummy insn between branch to
20468 label and label with no ptx insn inbetween.
20469 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20471 2018-01-24 Tom de Vries <tom@codesourcery.com>
20474 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20475 for neutered threads in warp.
20476 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20478 2018-01-24 Richard Biener <rguenther@suse.de>
20480 PR tree-optimization/83176
20481 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20484 2018-01-24 Richard Biener <rguenther@suse.de>
20486 PR tree-optimization/82819
20487 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20488 code generating pluses that are no-ops in the target precision.
20490 2018-01-24 Richard Biener <rguenther@suse.de>
20492 PR middle-end/84000
20493 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20495 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20497 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20498 to merge probabilities.
20499 * predict.c (probably_never_executed): Also mark as cold functions
20500 with global 0 profile and guessed local profile.
20501 * profile-count.c (profile_probability::combine_with_count): New
20503 * profile-count.h (profile_probability::operator*,
20504 profile_probability::operator*=, profile_probability::operator/,
20505 profile_probability::operator/=): Reduce precision to adjusted
20506 and set value to guessed on contradictory divisions.
20507 (profile_probability::combine_with_freq): Remove.
20508 (profile_probability::combine_wiht_count): Declare.
20509 (profile_count::force_nonzero):: Set to adjusted.
20510 (profile_count::probability_in):: Set quality to adjusted.
20511 * tree-ssa-tail-merge.c (replace_block_by): Use
20512 combine_with_count.
20514 2018-01-23 Andrew Waterman <andrew@sifive.com>
20515 Jim Wilson <jimw@sifive.com>
20517 * config/riscv/riscv.c (riscv_stack_boundary): New.
20518 (riscv_option_override): Set riscv_stack_boundary. Handle
20519 riscv_preferred_stack_boundary_arg.
20520 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20521 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20522 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20523 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20524 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20525 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20527 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20530 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20531 of struct ix86_frame.
20532 (ix86_expand_epilogue): Likewise. Add a local variable for
20533 the reg_save_offset field in struct ix86_frame.
20535 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20537 PR tree-optimization/82604
20538 * tree-loop-distribution.c (enum partition_kind): New enum item
20539 PKIND_PARTIAL_MEMSET.
20540 (partition_builtin_p): Support above new enum item.
20541 (generate_code_for_partition): Ditto.
20542 (compute_access_range): Differentiate cases that equality can be
20543 proven at all loops, the innermost loops or no loops.
20544 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20545 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20546 (finalize_partitions, distribute_loop): Don't fuse partition of
20547 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20548 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20549 parloop is enabled.
20551 2018-01-23 Martin Liska <mliska@suse.cz>
20553 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20554 order to ignore the predictor.
20555 (PRED_POLYMORPHIC_CALL): Likewise.
20556 (PRED_RECURSIVE_CALL): Likewise.
20558 2018-01-23 Martin Liska <mliska@suse.cz>
20560 * tree-profile.c (tree_profiling): Print function header to
20561 aware reader which function we are working on.
20562 * value-prof.c (gimple_find_values_to_profile): Do not print
20563 not interesting value histograms.
20565 2018-01-23 Martin Liska <mliska@suse.cz>
20567 * profile-count.h (enum profile_quality): Add
20568 profile_uninitialized as the first value. Do not number values
20569 as they are zero based.
20570 (profile_count::verify): Update sanity check.
20571 (profile_probability::verify): Likewise.
20573 2018-01-23 Nathan Sidwell <nathan@acm.org>
20575 * doc/invoke.texi (ffor-scope): Deprecate.
20577 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20579 PR tree-optimization/83510
20580 * domwalk.c (set_all_edges_as_executable): New function.
20581 (dom_walker::dom_walker): Convert bool param
20582 "skip_unreachable_blocks" to enum reachability. Move setup of
20583 edge flags to set_all_edges_as_executable and only do it when
20584 reachability is REACHABLE_BLOCKS.
20585 * domwalk.h (enum dom_walker::reachability): New enum.
20586 (dom_walker::dom_walker): Convert bool param
20587 "skip_unreachable_blocks" to enum reachability.
20588 (set_all_edges_as_executable): New decl.
20589 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20590 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20592 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20593 but converting true to REACHABLE_BLOCKS.
20594 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20596 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20597 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20598 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20600 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20601 if check_all_array_refs will be called.
20603 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20605 * tree.c (selftest::test_location_wrappers): Add more test
20608 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20610 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20611 (selftest::test_bit_in_range): Likewise.
20613 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20616 * doc/sourcebuild.texi (vect_float): Say that the selector
20617 only describes the situation when -funsafe-math-optimizations is on.
20618 (vect_float_strict): Document.
20620 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20622 PR tree-optimization/83965
20623 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20624 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20625 instead of checking only for a reduction.
20626 (vect_recog_widen_sum_pattern): Likewise.
20628 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20630 * predict.c (probably_never_executed): Only use precise profile info.
20631 (compute_function_frequency): Skip after inlining hack since we now
20632 have quality checking.
20634 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20636 * profile-count.h (profile_probability::very_unlikely,
20637 profile_probability::unlikely, profile_probability::even): Set
20638 precision to guessed.
20640 2018-01-23 Richard Biener <rguenther@suse.de>
20642 PR tree-optimization/83963
20643 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20644 Properly terminate dominator walk when crossing the exit edge not
20645 when visiting its source block.
20647 2018-01-23 Jakub Jelinek <jakub@redhat.com>
20650 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20651 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20653 2018-01-22 Jakub Jelinek <jakub@redhat.com>
20655 PR tree-optimization/83957
20656 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20657 semicolon after for body surrounded by braces.
20659 PR tree-optimization/83081
20660 * profile-count.h (profile_probability::split): New method.
20661 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20662 Use profile_probability::split.
20663 (do_compare_rtx_and_jump): Fix adjustment of probabilities
20664 when splitting a single conditional jump into 2.
20666 2018-01-22 David Malcolm <dmalcolm@redhat.com>
20668 PR tree-optimization/69452
20669 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
20672 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20674 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
20675 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
20676 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
20678 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20680 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
20682 * config/rl78/rl78.md (movdi): New define_expand.
20683 * config/rl78/rl78.c (rl78_split_movdi): New function.
20685 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
20688 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
20690 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
20691 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
20692 128-bit to produce an UNSPEC move to get the double word with the
20693 signbit and then a shift directly to do signbit.
20694 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
20695 implementation with a new version that just does either a direct
20696 move or a regular move. Move memory interface to separate insns.
20697 Move insns so they are next to the expander.
20698 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
20699 with signbit move. Split big and little endian case.
20700 (signbit<mode>2_dm_mem_le): Likewise.
20701 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
20702 (signbit<mode>2_dm2): Likewise.
20704 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20706 * config/rl78/rl78.md (anddi3): New define_expand.
20708 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20710 * config/rl78/rl78.md (umindi3): New define_expand.
20712 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20714 * config/rl78/rl78.md (smindi3): New define_expand.
20716 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20718 * config/rl78/rl78.md (smaxdi3): New define_expand.
20720 2018-01-22 Carl Love <cel@us.ibm.com>
20722 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
20723 LVX_V1TI): Add macro expansion.
20724 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
20725 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
20726 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
20727 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
20728 Change check to determine if the instruction is a byte reversing
20729 entry. Fix typo in comment.
20730 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
20731 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
20732 Add def_builtin calls for new builtins.
20733 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
20734 Add define_insn expansion.
20736 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20738 * config/rl78/rl78.md (umaxdi3): New define_expand.
20740 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20742 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
20743 for non-QImode registers.
20745 2018-01-22 Richard Biener <rguenther@suse.de>
20747 PR tree-optimization/83963
20748 * graphite-scop-detection.c (scop_detection::get_sese): Delay
20749 including the loop exit block.
20750 (scop_detection::merge_sese): Likewise.
20751 (scop_detection::add_scop): Do it here instead.
20753 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20755 * doc/sourcebuild.texi (arm_softfloat): Document.
20757 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
20760 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
20761 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
20762 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
20764 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20765 David Edelsohn <dje.gcc@gmail.com>
20768 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20769 Change "crset eq" to "crset 2".
20770 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20771 (*call_indirect_aix<mode>_nospec): Likewise.
20772 (*call_value_indirect_aix<mode>_nospec): Likewise.
20773 (*call_indirect_elfv2<mode>_nospec): Likewise.
20774 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
20775 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
20776 change assembly output from . to $.
20777 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20778 (indirect_jump<mode>_nospec): Change assembly output from . to $.
20779 (*tablejump<mode>_internal1_nospec): Likewise.
20781 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
20784 * config/sh/sh_optimize_sett_clrt.cc:
20785 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20787 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
20789 PR tree-optimization/83940
20790 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20791 offset_dt to vect_constant_def rather than vect_unknown_def_type.
20792 (vect_check_load_store_mask): Add a mask_dt_out parameter and
20793 use it to pass back the definition type.
20794 (vect_check_store_rhs): Likewise rhs_dt_out.
20795 (vect_build_gather_load_calls): Add a mask_dt argument and use
20796 it instead of a call to vect_is_simple_use.
20797 (vectorizable_store): Update calls to vect_check_load_store_mask
20798 and vect_check_store_rhs. Use the dt returned by the latter instead
20799 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
20800 instead of calls to vect_is_simple_use. Pass the scalar rather
20801 than the vector operand to vect_is_simple_use when handling
20802 second and subsequent copies of an rhs value.
20803 (vectorizable_load): Update calls to vect_check_load_store_mask
20804 and vect_build_gather_load_calls. Use the cached mask_dt and
20805 gs_info.offset_dt instead of calls to vect_is_simple_use.
20807 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20809 PR middle-end/83945
20810 * tree-emutls.c: Include gimplify.h.
20811 (lower_emutls_2): New function.
20812 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20813 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20814 it before further processing.
20817 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20818 UINTVAL (trueop1) instead of INTVAL (op1).
20820 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20824 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20825 INCOMING_FRAME_SP_OFFSET if not defined.
20826 (scan_trace): Add ENTRY argument. If true and
20827 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20828 emit a note to adjust the CFA offset.
20829 (create_cfi_notes): Adjust scan_trace callers.
20830 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20831 INCOMING_FRAME_SP_OFFSET in the CIE.
20832 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20833 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20835 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20836 * doc/tm.texi: Regenerated.
20838 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20840 PR rtl-optimization/83147
20841 * lra-constraints.c (remove_inheritance_pseudos): Use
20842 lra_substitute_pseudo_within_insn.
20844 2018-01-19 Tom de Vries <tom@codesourcery.com>
20845 Cesar Philippidis <cesar@codesourcery.com>
20848 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20850 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20853 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20854 spaces for function labels.
20856 2018-01-19 Martin Liska <mliska@suse.cz>
20858 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20859 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20860 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20861 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20862 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20863 (PRED_CONST_RETURN): Change from 69 to 65.
20864 (PRED_NULL_RETURN): Change from 91 to 71.
20865 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20866 (PRED_LOOP_GUARD): Change from 66 to 73.
20868 2018-01-19 Martin Liska <mliska@suse.cz>
20870 * predict.c (predict_insn_def): Add new assert.
20871 (struct branch_predictor): Change type to signed integer.
20872 (test_prediction_value_range): Amend test to cover
20873 PROB_UNINITIALIZED.
20874 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20875 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20876 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20877 (PRED_LOOP_IV_COMPARE): Likewise.
20878 * predict.h (PROB_UNINITIALIZED): Define new constant.
20880 2018-01-19 Martin Liska <mliska@suse.cz>
20882 * predict.c (dump_prediction): Add new format for
20883 analyze_brprob.py script which is enabled with -details
20885 * profile-count.h (precise_p): New function.
20887 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20889 PR tree-optimization/83922
20890 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20891 there are no statements that need masking.
20892 (vect_active_double_reduction_p): New function.
20893 (vect_analyze_loop_operations): Use it when handling phis that
20894 are not in the loop header.
20896 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20898 PR tree-optimization/83914
20899 * tree-vect-loop.c (vectorizable_induction): Don't convert
20900 init_expr or apply the peeling adjustment for inductions
20901 that are nested within the vectorized loop.
20903 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20905 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20908 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20912 * function.h (gimplify_parameters): Add gimple_seq * argument.
20913 * function.c: Include gimple.h and options.h.
20914 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20915 for the added local temporaries if needed.
20916 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20917 if there are any parameter cleanups, wrap whole body into a
20918 try/finally with the cleanups.
20920 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20923 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20924 Use GET_MODE_CLASS for scalar floating point.
20926 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20930 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20931 Fix call of call_cgraph_insertion_hooks.
20933 2018-01-18 Martin Sebor <msebor@redhat.com>
20935 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20937 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20940 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20943 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20946 * common.opt: (-ffile-prefix-map): New option.
20947 * opts.c (common_handle_option): Defer it.
20948 * opts-global.c (handle_common_deferred_options): Handle it.
20949 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20950 * file-prefix-map.h: New file.
20951 (remap_debug_filename, add_debug_prefix_map): ...here.
20952 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20953 * final.c (debug_prefix_map, add_debug_prefix_map
20954 remap_debug_filename): Move to...
20955 * file-prefix-map.c: New file.
20956 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20957 generalize, get rid of alloca(), use strrchr() instead of strchr().
20958 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20959 Implement in terms of add_prefix_map().
20960 (remap_macro_filename, remap_debug_filename): Implement in term of
20962 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20963 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20964 * dbxout.c: Include file-prefix-map.h.
20965 * varasm.c: Likewise.
20966 * vmsdbgout.c: Likewise.
20967 * xcoffout.c: Likewise.
20968 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20969 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20970 * doc/invoke.texi (-ffile-prefix-map): Document.
20971 (-fdebug-prefix-map): Update description.
20973 2018-01-18 Martin Liska <mliska@suse.cz>
20975 * config/i386/i386.c (indirect_thunk_name): Document that also
20977 (output_indirect_thunk): Document why both instructions
20978 (pause and lfence) are generated.
20980 2018-01-18 Richard Biener <rguenther@suse.de>
20982 PR tree-optimization/83887
20983 * graphite-scop-detection.c
20984 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20985 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20986 (scop_detection::merge_sese): Re-implement with a flood-fill
20987 algorithm that properly finds a SESE region if it exists.
20989 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20992 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20993 pointer_diff optimizations use view_convert instead of convert.
20995 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20997 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20998 Generate different code for -mno-speculate-indirect-jumps.
20999 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21000 (*call_indirect_aix<mode>): Disable for
21001 -mno-speculate-indirect-jumps.
21002 (*call_indirect_aix<mode>_nospec): New define_insn.
21003 (*call_value_indirect_aix<mode>): Disable for
21004 -mno-speculate-indirect-jumps.
21005 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21006 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21007 -mno-speculate-indirect-jumps.
21008 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21010 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21012 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21013 long double type, set the flags for noting the default long double
21014 type, even if we don't pass or return a long double type.
21016 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21019 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21022 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21024 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21025 support for merge[hl].
21026 (fold_mergehl_helper): New helper function.
21027 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21028 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21029 (altivec_vmrglw_direct): Add xxmrglw insn.
21031 2018-01-17 Andrew Waterman <andrew@sifive.com>
21033 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21034 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21036 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21039 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21040 call the lto_location_cache before reading the
21041 DECL_SOURCE_LOCATION of the types.
21043 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21044 Richard Sandiford <richard.sandiford@linaro.org>
21046 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21047 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21048 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21049 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21050 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21052 * config/aarch64/constraints.md (aarch64_movti_operand):
21054 * config/aarch64/predicates.md (Uti): Add new constraint.
21056 2018-01-17 Carl Love <cel@us.ibm.com>
21058 * config/rs6000/vsx.md (define_expand xl_len_r,
21059 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21060 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21062 (define_expand, define_insn): Move the shift left from the
21063 define_insn to the define_expand for lxvl and stxvl instructions.
21064 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21065 and XL_LEN_R definitions to PURE.
21067 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21069 * config/i386/i386.c (indirect_thunk_name): Declare regno
21070 as unsigned int. Compare regno with INVALID_REGNUM.
21071 (output_indirect_thunk): Ditto.
21072 (output_indirect_thunk_function): Ditto.
21073 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21074 in the call to output_indirect_thunk_function.
21076 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21078 PR middle-end/83884
21079 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21080 rather than the size of inner_type to determine the stack slot size
21081 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21083 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21086 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21089 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21091 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21092 endian Linux systems to optionally enable multilibs for selecting
21093 the long double type if the user configured an explicit type.
21094 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21095 have no long double multilibs if not defined.
21096 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21097 warn if the user used -mabi={ieee,ibm}longdouble and we built
21098 multilibs for long double.
21099 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21100 appropriate multilib option.
21101 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21103 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21104 for building long double multilibs.
21105 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21107 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21109 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21112 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21114 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21117 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21120 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21123 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21125 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21126 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21128 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21130 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21131 different rtl trees depending on TARGET_64BIT.
21132 (rs6000_gen_lvx): Likewise.
21134 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21136 * config/visium/visium.md (nop): Tweak comment.
21137 (hazard_nop): Likewise.
21139 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21141 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21142 -mspeculate-indirect-jumps.
21143 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21144 for -mno-speculate-indirect-jumps.
21145 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21146 (*call_value_indirect_elfv2<mode>): Disable for
21147 -mno-speculate-indirect-jumps.
21148 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21149 (indirect_jump): Emit different RTL for
21150 -mno-speculate-indirect-jumps.
21151 (*indirect_jump<mode>): Disable for
21152 -mno-speculate-indirect-jumps.
21153 (*indirect_jump<mode>_nospec): New define_insn.
21154 (tablejump): Emit different RTL for
21155 -mno-speculate-indirect-jumps.
21156 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21157 (tablejumpsi_nospec): New define_expand.
21158 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21159 (tablejumpdi_nospec): New define_expand.
21160 (*tablejump<mode>_internal1): Disable for
21161 -mno-speculate-indirect-jumps.
21162 (*tablejump<mode>_internal1_nospec): New define_insn.
21163 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21166 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21168 * caller-save.c (insert_save): Drop unnecessary parameter. All
21171 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21172 Richard Biener <rguenth@suse.de>
21175 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21176 return early, inline manually is_gimple_sizepos. Make sure if we
21177 call gimplify_expr we don't end up with a gimple constant.
21178 * tree.c (variably_modified_type_p): Don't return true for
21179 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21180 * gimplify.h (is_gimple_sizepos): Remove.
21182 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21184 PR tree-optimization/83857
21185 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21186 vectorizable_live_operation for pure SLP statements.
21187 (vectorizable_live_operation): Handle PHIs.
21189 2018-01-16 Richard Biener <rguenther@suse.de>
21191 PR tree-optimization/83867
21192 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21193 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21195 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21198 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21199 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21200 If off is not INTEGER_CST, issue a may not be aligned warning
21201 rather than isn't aligned. Use isn%'t rather than isn't.
21202 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21204 <case MULT_EXPR>: Improve the case when bottom and one of the
21205 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21206 operand, in that case check if the other operand is multiple of
21207 bottom divided by the INTEGER_CST operand.
21209 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21212 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21213 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21214 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21215 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21216 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21217 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21218 (pa_function_arg_size): New function.
21220 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21222 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21223 in a separate statement.
21225 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21227 PR tree-optimization/83847
21228 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21229 group gathers and scatters.
21231 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21233 PR rtl-optimization/86620
21234 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21236 PR rtl-optimization/83213
21237 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21238 to last if both are JUMP_INSNs.
21240 PR tree-optimization/83843
21241 * gimple-ssa-store-merging.c
21242 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21243 store_immediate_info for bswap/nop orig_stores.
21245 2018-01-15 Andrew Waterman <andrew@sifive.com>
21247 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21249 <UDIV>: Increase cost if !TARGET_DIV.
21251 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21253 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21254 (define_attr "cr_logical_3op"): New.
21255 (cceq_ior_compare): Adjust.
21256 (cceq_ior_compare_complement): Adjust.
21257 (*cceq_rev_compare): Adjust.
21258 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21259 (is_cracked_insn): Adjust.
21260 (insn_must_be_first_in_group): Adjust.
21261 * config/rs6000/40x.md: Adjust.
21262 * config/rs6000/440.md: Adjust.
21263 * config/rs6000/476.md: Adjust.
21264 * config/rs6000/601.md: Adjust.
21265 * config/rs6000/603.md: Adjust.
21266 * config/rs6000/6xx.md: Adjust.
21267 * config/rs6000/7450.md: Adjust.
21268 * config/rs6000/7xx.md: Adjust.
21269 * config/rs6000/8540.md: Adjust.
21270 * config/rs6000/cell.md: Adjust.
21271 * config/rs6000/e300c2c3.md: Adjust.
21272 * config/rs6000/e500mc.md: Adjust.
21273 * config/rs6000/e500mc64.md: Adjust.
21274 * config/rs6000/e5500.md: Adjust.
21275 * config/rs6000/e6500.md: Adjust.
21276 * config/rs6000/mpc.md: Adjust.
21277 * config/rs6000/power4.md: Adjust.
21278 * config/rs6000/power5.md: Adjust.
21279 * config/rs6000/power6.md: Adjust.
21280 * config/rs6000/power7.md: Adjust.
21281 * config/rs6000/power8.md: Adjust.
21282 * config/rs6000/power9.md: Adjust.
21283 * config/rs6000/rs64.md: Adjust.
21284 * config/rs6000/titan.md: Adjust.
21286 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21288 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21289 ix86_indirect_branch_register logic.
21291 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21293 * config/i386/constraints.md (Bs): Update
21294 ix86_indirect_branch_register check. Don't check
21295 ix86_indirect_branch_register with GOT_memory_operand.
21297 * config/i386/predicates.md (GOT_memory_operand): Don't check
21298 ix86_indirect_branch_register here.
21299 (GOT32_symbol_operand): Likewise.
21301 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21303 * config/i386/predicates.md (constant_call_address_operand):
21304 Rewrite ix86_indirect_branch_register logic.
21305 (sibcall_insn_operand): Likewise.
21307 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21309 * config/i386/constraints.md (Bs): Replace
21310 ix86_indirect_branch_thunk_register with
21311 ix86_indirect_branch_register.
21313 * config/i386/i386.md (indirect_jump): Likewise.
21314 (tablejump): Likewise.
21315 (*sibcall_memory): Likewise.
21316 (*sibcall_value_memory): Likewise.
21317 Peepholes of indirect call and jump via memory: Likewise.
21318 * config/i386/i386.opt: Likewise.
21319 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21320 (GOT_memory_operand): Likewise.
21321 (call_insn_operand): Likewise.
21322 (sibcall_insn_operand): Likewise.
21323 (GOT32_symbol_operand): Likewise.
21325 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21327 PR middle-end/83837
21328 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21329 type rather than type addr's type points to.
21330 (expand_omp_atomic_mutex): Likewise.
21331 (expand_omp_atomic): Likewise.
21333 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21336 * config/i386/i386.c (output_indirect_thunk_function): Use
21337 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21338 for __x86_return_thunk.
21340 2018-01-15 Richard Biener <rguenther@suse.de>
21342 PR middle-end/83850
21343 * expmed.c (extract_bit_field_1): Fix typo.
21345 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21348 * config/arm/iterators.md (VF): New mode iterator.
21349 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21350 Remove integer-related logic from pattern.
21351 (neon_vabd<mode>_3): Likewise.
21353 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21355 PR middle-end/82694
21356 * common.opt (fstrict-overflow): No longer an alias.
21357 (fwrapv-pointer): New option.
21358 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21359 also for pointer types based on flag_wrapv_pointer.
21360 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21361 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21362 opts->x_flag_wrapv got set.
21363 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21364 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21365 POINTER_TYPE_OVERFLOW_UNDEFINED.
21366 * match.pd: Likewise in address comparison pattern.
21367 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21369 2018-01-15 Richard Biener <rguenther@suse.de>
21372 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21373 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21374 Reset type names to their identifier if their TYPE_DECL doesn't
21375 have linkage (and thus is used for ODR and devirt).
21376 (save_debug_info_for_decl): Remove.
21377 (save_debug_info_for_type): Likewise.
21378 (add_tree_to_fld_list): Adjust.
21379 * tree-pretty-print.c (dump_generic_node): Make dumping of
21380 type names more robust.
21382 2018-01-15 Richard Biener <rguenther@suse.de>
21384 * BASE-VER: Bump to 8.0.1.
21386 2018-01-14 Martin Sebor <msebor@redhat.com>
21389 * builtins.c (check_access): Avoid warning when the no-warning bit
21392 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21394 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21395 * ira-color (allocno_hard_regs_compare): Likewise.
21397 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21400 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21401 Use .pushsection/.popsection.
21403 2018-01-14 Martin Sebor <msebor@redhat.com>
21406 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21408 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21410 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21411 entry from extra_headers.
21412 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21413 extra_headers, make the list bitwise identical to the i?86-*-* one.
21415 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21417 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21418 -mcmodel=large with -mindirect-branch=thunk,
21419 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21420 -mfunction-return=thunk-extern.
21421 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21422 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21423 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21425 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21427 * config/i386/i386.c (print_reg): Print the name of the full
21428 integer register without '%'.
21429 (ix86_print_operand): Handle 'V'.
21430 * doc/extend.texi: Document 'V' modifier.
21432 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21434 * config/i386/constraints.md (Bs): Disallow memory operand for
21435 -mindirect-branch-register.
21437 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21438 (GOT_memory_operand): Likewise.
21439 (call_insn_operand): Likewise.
21440 (sibcall_insn_operand): Likewise.
21441 (GOT32_symbol_operand): Likewise.
21442 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21443 for -mindirect-branch-register.
21444 (tablejump): Likewise.
21445 (*sibcall_memory): Likewise.
21446 (*sibcall_value_memory): Likewise.
21447 Disallow peepholes of indirect call and jump via memory for
21448 -mindirect-branch-register.
21449 (*call_pop): Replace m with Bw.
21450 (*call_value_pop): Likewise.
21451 (*sibcall_pop_memory): Replace m with Bs.
21452 * config/i386/i386.opt (mindirect-branch-register): New option.
21453 * doc/invoke.texi: Document -mindirect-branch-register option.
21455 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21457 * config/i386/i386-protos.h (ix86_output_function_return): New.
21458 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21459 set function_return_type.
21460 (indirect_thunk_name): Add ret_p to indicate thunk for function
21462 (output_indirect_thunk_function): Pass false to
21463 indirect_thunk_name.
21464 (ix86_output_indirect_branch_via_reg): Likewise.
21465 (ix86_output_indirect_branch_via_push): Likewise.
21466 (output_indirect_thunk_function): Create alias for function
21467 return thunk if regno < 0.
21468 (ix86_output_function_return): New function.
21469 (ix86_handle_fndecl_attribute): Handle function_return.
21470 (ix86_attribute_table): Add function_return.
21471 * config/i386/i386.h (machine_function): Add
21472 function_return_type.
21473 * config/i386/i386.md (simple_return_internal): Use
21474 ix86_output_function_return.
21475 (simple_return_internal_long): Likewise.
21476 * config/i386/i386.opt (mfunction-return=): New option.
21477 (indirect_branch): Mention -mfunction-return=.
21478 * doc/extend.texi: Document function_return function attribute.
21479 * doc/invoke.texi: Document -mfunction-return= option.
21481 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21483 * config/i386/i386-opts.h (indirect_branch): New.
21484 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21485 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21486 with local indirect jump when converting indirect call and jump.
21487 (ix86_set_indirect_branch_type): New.
21488 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21489 (indirectlabelno): New.
21490 (indirect_thunk_needed): Likewise.
21491 (indirect_thunk_bnd_needed): Likewise.
21492 (indirect_thunks_used): Likewise.
21493 (indirect_thunks_bnd_used): Likewise.
21494 (INDIRECT_LABEL): Likewise.
21495 (indirect_thunk_name): Likewise.
21496 (output_indirect_thunk): Likewise.
21497 (output_indirect_thunk_function): Likewise.
21498 (ix86_output_indirect_branch_via_reg): Likewise.
21499 (ix86_output_indirect_branch_via_push): Likewise.
21500 (ix86_output_indirect_branch): Likewise.
21501 (ix86_output_indirect_jmp): Likewise.
21502 (ix86_code_end): Call output_indirect_thunk_function if needed.
21503 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21505 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21506 (ix86_attribute_table): Add indirect_branch.
21507 * config/i386/i386.h (machine_function): Add indirect_branch_type
21508 and has_local_indirect_jump.
21509 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21511 (tablejump): Likewise.
21512 (*indirect_jump): Use ix86_output_indirect_jmp.
21513 (*tablejump_1): Likewise.
21514 (simple_return_indirect_internal): Likewise.
21515 * config/i386/i386.opt (mindirect-branch=): New option.
21516 (indirect_branch): New.
21519 (thunk-inline): Likewise.
21520 (thunk-extern): Likewise.
21521 * doc/extend.texi: Document indirect_branch function attribute.
21522 * doc/invoke.texi: Document -mindirect-branch= option.
21524 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21527 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21529 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21531 * ipa-inline.c (want_inline_small_function_p): Return false if
21532 inlining has already failed with CIF_FINAL_ERROR.
21533 (update_caller_keys): Call want_inline_small_function_p before
21535 (update_callee_keys): Likewise.
21537 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21539 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21541 (rs6000_quadword_masked_address_p): Likewise.
21542 (quad_aligned_load_p): Likewise.
21543 (quad_aligned_store_p): Likewise.
21544 (const_load_sequence_p): Add comment to describe the outer-most loop.
21545 (mimic_memory_attributes_and_flags): New function.
21546 (rs6000_gen_stvx): Likewise.
21547 (replace_swapped_aligned_store): Likewise.
21548 (rs6000_gen_lvx): Likewise.
21549 (replace_swapped_aligned_load): Likewise.
21550 (replace_swapped_load_constant): Capitalize argument name in
21551 comment describing this function.
21552 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21553 and stores that access quad-word aligned addresses and replace
21554 with stvx or lvx instructions when appropriate.
21555 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21556 New function prototype.
21557 (rs6000_quadword_masked_address_p): Likewise.
21558 (rs6000_gen_lvx): Likewise.
21559 (rs6000_gen_stvx): Likewise.
21560 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21561 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21562 when memory address is aligned.
21563 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21564 this split to select lvx instruction when memory address is aligned.
21565 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21566 instruction when memory address is aligned.
21567 (*vsx_le_perm_load_v16qi): Likewise.
21568 (four unnamed splitters): Modify to select the stvx instruction
21569 when memory is aligned.
21571 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21573 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21574 which appears in the queue multiple times.
21576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21577 Alan Hayward <alan.hayward@arm.com>
21578 David Sherwood <david.sherwood@arm.com>
21580 * tree-vectorizer.h (vec_lower_bound): New structure.
21581 (_loop_vec_info): Add check_nonzero and lower_bounds.
21582 (LOOP_VINFO_CHECK_NONZERO): New macro.
21583 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21584 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21585 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21586 fields. Make seg_len the distance travelled, not including the
21588 (dr_direction_indicator): Declare.
21589 (dr_zero_step_indicator): Likewise.
21590 (dr_known_forward_stride_p): Likewise.
21591 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21593 (runtime_alias_check_p): Allow runtime alias checks with
21595 (operator ==): Compare access_size and align.
21596 (prune_runtime_alias_test_list): Rework for new distinction between
21597 the access_size and seg_len.
21598 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21600 (get_segment_min_max): New function.
21601 (create_intersect_range_checks): Use it.
21602 (dr_step_indicator): New function.
21603 (dr_direction_indicator): Likewise.
21604 (dr_zero_step_indicator): Likewise.
21605 (dr_known_forward_stride_p): Likewise.
21606 * tree-loop-distribution.c (data_ref_segment_size): Return
21607 DR_STEP * (niters - 1).
21608 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21610 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21611 (vect_preserves_scalar_order_p): New function, split out from...
21612 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21613 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21614 (vect_vfa_access_size): New function.
21615 (vect_vfa_align): Likewise.
21616 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21617 (dump_lower_bound): New function.
21618 (vect_check_lower_bound): Likewise.
21619 (vect_small_gap_p): Likewise.
21620 (vectorizable_with_step_bound_p): Likewise.
21621 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21622 depencies if the vectorization factor is 1. Convert the checks
21623 for nonzero steps into checks on the bounds of DR_STEP. Try using
21624 a bunds check for variable steps if the minimum required step is
21625 relatively small. Update calls to the dr_with_seg_len
21626 constructor and to vect_compile_time_alias.
21627 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21629 (vect_loop_versioning): Call it.
21630 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21632 (vect_estimate_min_profitable_iters): Account for any bounds checks.
21634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21635 Alan Hayward <alan.hayward@arm.com>
21636 David Sherwood <david.sherwood@arm.com>
21638 * doc/sourcebuild.texi (vect_scatter_store): Document.
21639 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21641 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21643 * genopinit.c (main): Add supports_vec_scatter_store and
21644 supports_vec_scatter_store_cached to target_optabs.
21645 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21646 IFN_MASK_SCATTER_STORE.
21647 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21649 * internal-fn.h (internal_store_fn_p): Declare.
21650 (internal_fn_stored_value_index): Likewise.
21651 * internal-fn.c (scatter_store_direct): New macro.
21652 (expand_scatter_store_optab_fn): New function.
21653 (direct_scatter_store_optab_supported_p): New macro.
21654 (internal_store_fn_p): New function.
21655 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21656 IFN_MASK_SCATTER_STORE.
21657 (internal_fn_mask_index): Likewise.
21658 (internal_fn_stored_value_index): New function.
21659 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21660 for scatter stores.
21661 * optabs-query.h (supports_vec_scatter_store_p): Declare.
21662 * optabs-query.c (supports_vec_scatter_store_p): New function.
21663 * tree-vectorizer.h (vect_get_store_rhs): Declare.
21664 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21665 true for scatter stores.
21666 (vect_gather_scatter_fn_p): Handle scatter stores too.
21667 (vect_check_gather_scatter): Consider using scatter stores if
21668 supports_vec_scatter_store_p.
21669 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
21670 scatter stores too.
21671 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21672 internal_fn_stored_value_index.
21673 (check_load_store_masking): Handle scatter stores too.
21674 (vect_get_store_rhs): Make public.
21675 (vectorizable_call): Use internal_store_fn_p.
21676 (vectorizable_store): Handle scatter store internal functions.
21677 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
21678 when deciding whether the end of the group has been reached.
21679 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
21680 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
21681 (mask_scatter_store<mode>): New insns.
21683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21684 Alan Hayward <alan.hayward@arm.com>
21685 David Sherwood <david.sherwood@arm.com>
21687 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
21688 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
21689 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
21691 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
21692 Use vect_truncate_gather_scatter_offset if we can't treat the
21693 operation as a normal gather load or scatter store.
21694 (get_group_load_store_type): Take the gather_scatter_info
21695 as argument. Try using a gather load or scatter store for
21696 single-element groups.
21697 (get_load_store_type): Update calls to get_group_load_store_type
21698 and vect_use_strided_gather_scatters_p.
21700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21701 Alan Hayward <alan.hayward@arm.com>
21702 David Sherwood <david.sherwood@arm.com>
21704 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
21705 optional tree argument.
21706 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
21708 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
21709 but continue to use the current value as a fallback.
21710 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
21711 to compare the updates.
21712 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
21713 (get_load_store_type): Use it when handling a strided access.
21714 (vect_get_strided_load_store_ops): New function.
21715 (vect_get_data_ptr_increment): Likewise.
21716 (vectorizable_load): Handle strided gather loads. Always pass
21717 a step to vect_create_data_ref_ptr and bump_vector_ptr.
21719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21720 Alan Hayward <alan.hayward@arm.com>
21721 David Sherwood <david.sherwood@arm.com>
21723 * doc/md.texi (gather_load@var{m}): Document.
21724 (mask_gather_load@var{m}): Likewise.
21725 * genopinit.c (main): Add supports_vec_gather_load and
21726 supports_vec_gather_load_cached to target_optabs.
21727 * optabs-tree.c (init_tree_optimization_optabs): Use
21728 ggc_cleared_alloc to allocate target_optabs.
21729 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
21730 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
21732 * internal-fn.h (internal_load_fn_p): Declare.
21733 (internal_gather_scatter_fn_p): Likewise.
21734 (internal_fn_mask_index): Likewise.
21735 (internal_gather_scatter_fn_supported_p): Likewise.
21736 * internal-fn.c (gather_load_direct): New macro.
21737 (expand_gather_load_optab_fn): New function.
21738 (direct_gather_load_optab_supported_p): New macro.
21739 (direct_internal_fn_optab): New function.
21740 (internal_load_fn_p): Likewise.
21741 (internal_gather_scatter_fn_p): Likewise.
21742 (internal_fn_mask_index): Likewise.
21743 (internal_gather_scatter_fn_supported_p): Likewise.
21744 * optabs-query.c (supports_at_least_one_mode_p): New function.
21745 (supports_vec_gather_load_p): Likewise.
21746 * optabs-query.h (supports_vec_gather_load_p): Declare.
21747 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
21748 and memory_type field.
21749 (NUM_PATTERNS): Bump to 15.
21750 * tree-vect-data-refs.c: Include internal-fn.h.
21751 (vect_gather_scatter_fn_p): New function.
21752 (vect_describe_gather_scatter_call): Likewise.
21753 (vect_check_gather_scatter): Try using internal functions for
21754 gather loads. Recognize existing calls to a gather load function.
21755 (vect_analyze_data_refs): Consider using gather loads if
21756 supports_vec_gather_load_p.
21757 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
21758 (vect_get_gather_scatter_offset_type): Likewise.
21759 (vect_convert_mask_for_vectype): Likewise.
21760 (vect_add_conversion_to_patterm): Likewise.
21761 (vect_try_gather_scatter_pattern): Likewise.
21762 (vect_recog_gather_scatter_pattern): New pattern recognizer.
21763 (vect_vect_recog_func_ptrs): Add it.
21764 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21765 internal_fn_mask_index and internal_gather_scatter_fn_p.
21766 (check_load_store_masking): Take the gather_scatter_info as an
21767 argument and handle gather loads.
21768 (vect_get_gather_scatter_ops): New function.
21769 (vectorizable_call): Check internal_load_fn_p.
21770 (vectorizable_load): Likewise. Handle gather load internal
21772 (vectorizable_store): Update call to check_load_store_masking.
21773 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
21774 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
21775 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
21776 (aarch64_gather_scale_operand_d): New predicates.
21777 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
21778 (mask_gather_load<mode>): New insns.
21780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21781 Alan Hayward <alan.hayward@arm.com>
21782 David Sherwood <david.sherwood@arm.com>
21784 * optabs.def (fold_left_plus_optab): New optab.
21785 * doc/md.texi (fold_left_plus_@var{m}): Document.
21786 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21787 * internal-fn.c (fold_left_direct): Define.
21788 (expand_fold_left_optab_fn): Likewise.
21789 (direct_fold_left_optab_supported_p): Likewise.
21790 * fold-const-call.c (fold_const_fold_left): New function.
21791 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21792 * tree-parloops.c (valid_reduction_p): New function.
21793 (gather_scalar_reductions): Use it.
21794 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21795 (vect_finish_replace_stmt): Declare.
21796 * tree-vect-loop.c (fold_left_reduction_fn): New function.
21797 (needs_fold_left_reduction_p): New function, split out from...
21798 (vect_is_simple_reduction): ...here. Accept reductions that
21799 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21800 (vect_force_simple_reduction): Also store the reduction type in
21801 the assignment's STMT_VINFO_REDUC_TYPE.
21802 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21803 (merge_with_identity): New function.
21804 (vect_expand_fold_left): Likewise.
21805 (vectorize_fold_left_reduction): Likewise.
21806 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21807 scalar phi in place for it. Check for target support and reject
21808 cases that would reassociate the operation. Defer the transform
21809 phase to vectorize_fold_left_reduction.
21810 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21811 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21812 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21814 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21816 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21817 call to ifc_temp_var.
21819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21820 Alan Hayward <alan.hayward@arm.com>
21821 David Sherwood <david.sherwood@arm.com>
21823 * target.def (legitimize_address_displacement): Take the original
21824 offset as a poly_int.
21825 * targhooks.h (default_legitimize_address_displacement): Update
21827 * targhooks.c (default_legitimize_address_displacement): Likewise.
21828 * doc/tm.texi: Regenerate.
21829 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21830 as an argument, moving assert of ad->disp == ad->disp_term to...
21831 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21832 Try calling targetm.legitimize_address_displacement before expanding
21833 the address rather than afterwards, and adjust for the new interface.
21834 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21835 Match the new hook interface. Handle SVE addresses.
21836 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21837 new hook interface.
21839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21841 * Makefile.in (OBJS): Add early-remat.o.
21842 * target.def (select_early_remat_modes): New hook.
21843 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21844 * doc/tm.texi: Regenerate.
21845 * targhooks.h (default_select_early_remat_modes): Declare.
21846 * targhooks.c (default_select_early_remat_modes): New function.
21847 * timevar.def (TV_EARLY_REMAT): New timevar.
21848 * passes.def (pass_early_remat): New pass.
21849 * tree-pass.h (make_pass_early_remat): Declare.
21850 * early-remat.c: New file.
21851 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21853 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21856 Alan Hayward <alan.hayward@arm.com>
21857 David Sherwood <david.sherwood@arm.com>
21859 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21860 vfm1 with a bound_epilog parameter.
21861 (vect_do_peeling): Update calls accordingly, and move the prologue
21862 call earlier in the function. Treat the base bound_epilog as 0 for
21863 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21864 this base when peeling for gaps.
21865 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21866 with fully-masked loops.
21867 (vect_estimate_min_profitable_iters): Handle the single peeled
21868 iteration in that case.
21870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21871 Alan Hayward <alan.hayward@arm.com>
21872 David Sherwood <david.sherwood@arm.com>
21874 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21875 single-element interleaving even if the size is not a power of 2.
21876 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21877 accesses for single-element interleaving if the group size is
21880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21881 Alan Hayward <alan.hayward@arm.com>
21882 David Sherwood <david.sherwood@arm.com>
21884 * doc/md.texi (fold_extract_last_@var{m}): Document.
21885 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21886 * optabs.def (fold_extract_last_optab): New optab.
21887 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21888 * internal-fn.c (fold_extract_direct): New macro.
21889 (expand_fold_extract_optab_fn): Likewise.
21890 (direct_fold_extract_optab_supported_p): Likewise.
21891 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21892 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21893 EXTRACT_LAST_REDUCTION.
21894 (get_initial_def_for_reduction): Do not create an initial vector
21895 for EXTRACT_LAST_REDUCTION reductions.
21896 (vectorizable_reduction): Leave the scalar phi in place for
21897 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21898 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21899 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21900 transform phase to vectorizable_condition.
21901 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21903 (vect_finish_stmt_generation): ...here.
21904 (vect_finish_replace_stmt): New function.
21905 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21906 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21908 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21911 Alan Hayward <alan.hayward@arm.com>
21912 David Sherwood <david.sherwood@arm.com>
21914 * doc/md.texi (extract_last_@var{m}): Document.
21915 * optabs.def (extract_last_optab): New optab.
21916 * internal-fn.def (EXTRACT_LAST): New internal function.
21917 * internal-fn.c (cond_unary_direct): New macro.
21918 (expand_cond_unary_optab_fn): Likewise.
21919 (direct_cond_unary_optab_supported_p): Likewise.
21920 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21921 loops using EXTRACT_LAST.
21922 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21923 (extract_last_<mode>): ...this optab.
21924 (vec_extract<mode><Vel>): Update accordingly.
21926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21927 Alan Hayward <alan.hayward@arm.com>
21928 David Sherwood <david.sherwood@arm.com>
21930 * target.def (empty_mask_is_expensive): New hook.
21931 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21932 * doc/tm.texi: Regenerate.
21933 * targhooks.h (default_empty_mask_is_expensive): Declare.
21934 * targhooks.c (default_empty_mask_is_expensive): New function.
21935 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21936 if the target says that empty masks are expensive.
21937 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21939 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21941 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21942 Alan Hayward <alan.hayward@arm.com>
21943 David Sherwood <david.sherwood@arm.com>
21945 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21946 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21947 (vect_use_loop_mask_for_alignment_p): New function.
21948 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21949 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21950 niters_skip argument. Make sure that the first niters_skip elements
21951 of the first iteration are inactive.
21952 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21953 Update call to vect_set_loop_masks_directly.
21954 (get_misalign_in_elems): New function, split out from...
21955 (vect_gen_prolog_loop_niters): ...here.
21956 (vect_update_init_of_dr): Take a code argument that specifies whether
21957 the adjustment should be added or subtracted.
21958 (vect_update_init_of_drs): Likewise.
21959 (vect_prepare_for_masked_peels): New function.
21960 (vect_do_peeling): Skip prologue peeling if we're using a mask
21961 instead. Update call to vect_update_inits_of_drs.
21962 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21964 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21965 alignment. Do not include the number of peeled iterations in
21966 the minimum threshold in that case.
21967 (vectorizable_induction): Adjust the start value down by
21968 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21969 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21970 Take the number of skipped iterations into account when calculating
21972 * tree-vect-stmts.c (vect_gen_while_not): New function.
21974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21975 Alan Hayward <alan.hayward@arm.com>
21976 David Sherwood <david.sherwood@arm.com>
21978 * doc/sourcebuild.texi (vect_fully_masked): Document.
21979 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21980 default value to 0.
21981 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21983 (vect_analyze_loop_2): ...here. Don't check the vectorization
21984 factor against the number of loop iterations if the loop is
21987 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21988 Alan Hayward <alan.hayward@arm.com>
21989 David Sherwood <david.sherwood@arm.com>
21991 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21992 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21993 (dump_groups): Update accordingly.
21994 (iv_use::mem_type): New member variable.
21995 (address_p): New function.
21996 (record_use): Add a mem_type argument and initialize the new
21998 (record_group_use): Add a mem_type argument. Use address_p.
21999 Remove obsolete null checks of base_object. Update call to record_use.
22000 (find_interesting_uses_op): Update call to record_group_use.
22001 (find_interesting_uses_cond): Likewise.
22002 (find_interesting_uses_address): Likewise.
22003 (get_mem_type_for_internal_fn): New function.
22004 (find_address_like_use): Likewise.
22005 (find_interesting_uses_stmt): Try find_address_like_use before
22006 calling find_interesting_uses_op.
22007 (addr_offset_valid_p): Use the iv mem_type field as the type
22008 of the addressed memory.
22009 (add_autoinc_candidates): Likewise.
22010 (get_address_cost): Likewise.
22011 (split_small_address_groups_p): Use address_p.
22012 (split_address_groups): Likewise.
22013 (add_iv_candidate_for_use): Likewise.
22014 (autoinc_possible_for_pair): Likewise.
22015 (rewrite_groups): Likewise.
22016 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22017 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22018 (get_alias_ptr_type_for_ptr_address): New function.
22019 (rewrite_use_address): Rewrite address uses in calls that were
22020 identified by find_address_like_use.
22022 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22023 Alan Hayward <alan.hayward@arm.com>
22024 David Sherwood <david.sherwood@arm.com>
22026 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22028 * gimple-expr.h (is_gimple_addressable: Likewise.
22029 * gimple-expr.c (is_gimple_address): Likewise.
22030 * internal-fn.c (expand_call_mem_ref): New function.
22031 (expand_mask_load_optab_fn): Use it.
22032 (expand_mask_store_optab_fn): Likewise.
22034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22035 Alan Hayward <alan.hayward@arm.com>
22036 David Sherwood <david.sherwood@arm.com>
22038 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22039 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22040 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22041 (cond_umax@var{mode}): Document.
22042 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22043 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22044 (cond_umin_optab, cond_umax_optab): New optabs.
22045 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22046 (COND_IOR, COND_XOR): New internal functions.
22047 * internal-fn.h (get_conditional_internal_fn): Declare.
22048 * internal-fn.c (cond_binary_direct): New macro.
22049 (expand_cond_binary_optab_fn): Likewise.
22050 (direct_cond_binary_optab_supported_p): Likewise.
22051 (get_conditional_internal_fn): New function.
22052 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22053 Cope with reduction statements that are vectorized as calls rather
22055 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22056 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22057 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22058 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22059 (UNSPEC_COND_EOR): New unspecs.
22060 (optab): Add mappings for them.
22061 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22062 (sve_int_op, sve_fp_op): New int attributes.
22064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22065 Alan Hayward <alan.hayward@arm.com>
22066 David Sherwood <david.sherwood@arm.com>
22068 * optabs.def (while_ult_optab): New optab.
22069 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22070 * internal-fn.def (WHILE_ULT): New internal function.
22071 * internal-fn.h (direct_internal_fn_supported_p): New override
22072 that takes two types as argument.
22073 * internal-fn.c (while_direct): New macro.
22074 (expand_while_optab_fn): New function.
22075 (convert_optab_supported_p): Likewise.
22076 (direct_while_optab_supported_p): New macro.
22077 * wide-int.h (wi::udiv_ceil): New function.
22078 * tree-vectorizer.h (rgroup_masks): New structure.
22079 (vec_loop_masks): New typedef.
22080 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22081 and fully_masked_p.
22082 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22083 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22084 (vect_max_vf): New function.
22085 (slpeel_make_loop_iterate_ntimes): Delete.
22086 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22087 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22088 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22089 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22090 internal-fn.h, stor-layout.h and optabs-query.h.
22091 (vect_set_loop_mask): New function.
22092 (add_preheader_seq): Likewise.
22093 (add_header_seq): Likewise.
22094 (interleave_supported_p): Likewise.
22095 (vect_maybe_permute_loop_masks): Likewise.
22096 (vect_set_loop_masks_directly): Likewise.
22097 (vect_set_loop_condition_masked): Likewise.
22098 (vect_set_loop_condition_unmasked): New function, split out from
22099 slpeel_make_loop_iterate_ntimes.
22100 (slpeel_make_loop_iterate_ntimes): Rename to..
22101 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22102 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22103 (vect_do_peeling): Update call accordingly.
22104 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22106 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22107 mask_compare_type, can_fully_mask_p and fully_masked_p.
22108 (release_vec_loop_masks): New function.
22109 (_loop_vec_info): Use it to free the loop masks.
22110 (can_produce_all_loop_masks_p): New function.
22111 (vect_get_max_nscalars_per_iter): Likewise.
22112 (vect_verify_full_masking): Likewise.
22113 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22114 retries, and free the mask rgroups before retrying. Check loop-wide
22115 reasons for disallowing fully-masked loops. Make the final decision
22116 about whether use a fully-masked loop or not.
22117 (vect_estimate_min_profitable_iters): Do not assume that peeling
22118 for the number of iterations will be needed for fully-masked loops.
22119 (vectorizable_reduction): Disable fully-masked loops.
22120 (vectorizable_live_operation): Likewise.
22121 (vect_halve_mask_nunits): New function.
22122 (vect_double_mask_nunits): Likewise.
22123 (vect_record_loop_mask): Likewise.
22124 (vect_get_loop_mask): Likewise.
22125 (vect_transform_loop): Handle the case in which the final loop
22126 iteration might handle a partial vector. Call vect_set_loop_condition
22127 instead of slpeel_make_loop_iterate_ntimes.
22128 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22129 (check_load_store_masking): New function.
22130 (prepare_load_store_mask): Likewise.
22131 (vectorizable_store): Handle fully-masked loops.
22132 (vectorizable_load): Likewise.
22133 (supportable_widening_operation): Use vect_halve_mask_nunits for
22135 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22136 (vect_gen_while): New function.
22137 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22138 (aarch64_uqdec<mode>): New insn.
22140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22141 Alan Hayward <alan.hayward@arm.com>
22142 David Sherwood <david.sherwood@arm.com>
22144 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22145 (reduc_xor_scal_optab): New optabs.
22146 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22147 (reduc_xor_scal_@var{m}): Document.
22148 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22149 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22150 internal functions.
22151 * fold-const-call.c (fold_const_call): Handle them.
22152 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22153 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22154 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22155 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22156 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22157 (UNSPEC_XORV): New unspecs.
22158 (optab): Add entries for them.
22159 (BITWISEV): New int iterator.
22160 (bit_reduc_op): New int attributes.
22162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22163 Alan Hayward <alan.hayward@arm.com>
22164 David Sherwood <david.sherwood@arm.com>
22166 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22167 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22168 * optabs.def (vec_shl_insert_optab): New optab.
22169 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22170 (duplicate_and_interleave): Likewise.
22171 * tree-vect-loop.c: Include internal-fn.h.
22172 (neutral_op_for_slp_reduction): New function, split out from
22173 get_initial_defs_for_reduction.
22174 (get_initial_def_for_reduction): Handle option 2 for variable-length
22175 vectors by loading the neutral value into a vector and then shifting
22176 the initial value into element 0.
22177 (get_initial_defs_for_reduction): Replace the code argument with
22178 the neutral value calculated by neutral_op_for_slp_reduction.
22179 Use gimple_build_vector for constant-length vectors.
22180 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22181 but the first group_size elements have a neutral value.
22182 Use duplicate_and_interleave otherwise.
22183 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22184 Update call to get_initial_defs_for_reduction. Handle SLP
22185 reductions for variable-length vectors by creating one vector
22186 result for each scalar result, with the elements associated
22187 with other scalar results stubbed out with the neutral value.
22188 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22189 Require IFN_VEC_SHL_INSERT for double reductions on
22190 variable-length vectors, or SLP reductions that have
22191 a neutral value. Require can_duplicate_and_interleave_p
22192 support for variable-length unchained SLP reductions if there
22193 is no neutral value, such as for MIN/MAX reductions. Also require
22194 the number of vector elements to be a multiple of the number of
22195 SLP statements when doing variable-length unchained SLP reductions.
22196 Update call to vect_create_epilog_for_reduction.
22197 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22198 and remove initial values.
22199 (duplicate_and_interleave): Make public.
22200 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22201 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22204 Alan Hayward <alan.hayward@arm.com>
22205 David Sherwood <david.sherwood@arm.com>
22207 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22208 (can_duplicate_and_interleave_p): New function.
22209 (vect_get_and_check_slp_defs): Take the vector of statements
22210 rather than just the current one. Remove excess parentheses.
22211 Restriction rejectinon of vect_constant_def and vect_external_def
22212 for variable-length vectors to boolean types, or types for which
22213 can_duplicate_and_interleave_p is false.
22214 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22215 (duplicate_and_interleave): New function.
22216 (vect_get_constant_vectors): Use gimple_build_vector for
22217 constant-length vectors and suitable variable-length constant
22218 vectors. Use duplicate_and_interleave for other variable-length
22219 vectors. Don't defer the update when inserting new statements.
22221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22222 Alan Hayward <alan.hayward@arm.com>
22223 David Sherwood <david.sherwood@arm.com>
22225 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22226 min_profitable_iters doesn't go negative.
22228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22229 Alan Hayward <alan.hayward@arm.com>
22230 David Sherwood <david.sherwood@arm.com>
22232 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22233 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22234 * optabs.def (vec_mask_load_lanes_optab): New optab.
22235 (vec_mask_store_lanes_optab): Likewise.
22236 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22237 (MASK_STORE_LANES): Likewise.
22238 * internal-fn.c (mask_load_lanes_direct): New macro.
22239 (mask_store_lanes_direct): Likewise.
22240 (expand_mask_load_optab_fn): Handle masked operations.
22241 (expand_mask_load_lanes_optab_fn): New macro.
22242 (expand_mask_store_optab_fn): Handle masked operations.
22243 (expand_mask_store_lanes_optab_fn): New macro.
22244 (direct_mask_load_lanes_optab_supported_p): Likewise.
22245 (direct_mask_store_lanes_optab_supported_p): Likewise.
22246 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22248 (vect_load_lanes_supported): Likewise.
22249 * tree-vect-data-refs.c (strip_conversion): New function.
22250 (can_group_stmts_p): Likewise.
22251 (vect_analyze_data_ref_accesses): Use it instead of checking
22252 for a pair of assignments.
22253 (vect_store_lanes_supported): Take a masked_p parameter.
22254 (vect_load_lanes_supported): Likewise.
22255 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22256 vect_store_lanes_supported and vect_load_lanes_supported.
22257 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22258 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22259 parameter. Don't allow gaps for masked accesses.
22260 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22261 and vect_load_lanes_supported.
22262 (get_load_store_type): Take a masked_p parameter and update
22263 call to get_group_load_store_type.
22264 (vectorizable_store): Update call to get_load_store_type.
22265 Handle IFN_MASK_STORE_LANES.
22266 (vectorizable_load): Update call to get_load_store_type.
22267 Handle IFN_MASK_LOAD_LANES.
22269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22270 Alan Hayward <alan.hayward@arm.com>
22271 David Sherwood <david.sherwood@arm.com>
22273 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22275 * config/aarch64/aarch64-protos.h
22276 (aarch64_sve_struct_memory_operand_p): Declare.
22277 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22278 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22279 (VPRED, vpred): Handle SVE structure modes.
22280 * config/aarch64/constraints.md (Utx): New constraint.
22281 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22282 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22283 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22284 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22285 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22286 structure modes. Split into pieces after RA.
22287 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22288 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22290 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22291 SVE structure modes.
22292 (aarch64_classify_address): Likewise.
22293 (sizetochar): Move earlier in file.
22294 (aarch64_print_operand): Handle SVE register lists.
22295 (aarch64_array_mode): New function.
22296 (aarch64_sve_struct_memory_operand_p): Likewise.
22297 (TARGET_ARRAY_MODE): Redefine.
22299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22300 Alan Hayward <alan.hayward@arm.com>
22301 David Sherwood <david.sherwood@arm.com>
22303 * target.def (array_mode): New target hook.
22304 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22305 * doc/tm.texi: Regenerate.
22306 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22307 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22308 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22309 targetm.array_mode.
22310 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22314 Alan Hayward <alan.hayward@arm.com>
22315 David Sherwood <david.sherwood@arm.com>
22317 * fold-const.c (fold_binary_loc): Check the argument types
22318 rather than the result type when testing for a vector operation.
22320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22322 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22323 * doc/tm.texi: Regenerate.
22325 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22326 Alan Hayward <alan.hayward@arm.com>
22327 David Sherwood <david.sherwood@arm.com>
22329 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22330 (sve): Document new AArch64 extension.
22331 * doc/md.texi (w): Extend the description of the AArch64
22332 constraint to include SVE vectors.
22333 (Upl, Upa): Document new AArch64 predicate constraints.
22334 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22336 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22337 (msve-vector-bits=): New option.
22338 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22339 SVE when these are disabled.
22340 (sve): New extension.
22341 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22342 modes. Adjust their number of units based on aarch64_sve_vg.
22343 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22344 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22345 aarch64_addr_query_type.
22346 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22347 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22348 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22349 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22350 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22351 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22352 (aarch64_simd_imm_zero_p): Delete.
22353 (aarch64_check_zero_based_sve_index_immediate): Declare.
22354 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22355 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22356 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22357 (aarch64_sve_float_mul_immediate_p): Likewise.
22358 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22359 rather than an rtx.
22360 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22361 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22362 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22363 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22364 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22365 (aarch64_regmode_natural_size): Likewise.
22366 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22367 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22369 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22370 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22371 for VG and the SVE predicate registers.
22372 (V_ALIASES): Add a "z"-prefixed alias.
22373 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22374 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22375 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22376 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22377 (REG_CLASS_NAMES): Add entries for them.
22378 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22379 and the predicate registers.
22380 (aarch64_sve_vg): Declare.
22381 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22382 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22383 (REGMODE_NATURAL_SIZE): Define.
22384 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22386 * config/aarch64/aarch64.c: Include cfgrtl.h.
22387 (simd_immediate_info): Add a constructor for series vectors,
22388 and an associated step field.
22389 (aarch64_sve_vg): New variable.
22390 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22391 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22392 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22393 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22394 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22395 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22396 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22397 (aarch64_get_mask_mode): New functions.
22398 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22399 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22400 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22401 predicate modes and predicate registers. Explicitly restrict
22402 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22403 to store a vector mode if it is recognized by
22404 aarch64_classify_vector_mode.
22405 (aarch64_regmode_natural_size): New function.
22406 (aarch64_hard_regno_caller_save_mode): Return the original mode
22408 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22409 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22410 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22411 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22413 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22414 does not overlap dest if the function is frame-related. Handle
22416 (aarch64_split_add_offset): New function.
22417 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22418 them aarch64_add_offset.
22419 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22420 and update call to aarch64_sub_sp.
22421 (aarch64_add_cfa_expression): New function.
22422 (aarch64_expand_prologue): Pass extra temporary registers to the
22423 functions above. Handle the case in which we need to emit new
22424 DW_CFA_expressions for registers that were originally saved
22425 relative to the stack pointer, but now have to be expressed
22426 relative to the frame pointer.
22427 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22429 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22430 IP0 and IP1 values for SVE frames.
22431 (aarch64_expand_vec_series): New function.
22432 (aarch64_expand_sve_widened_duplicate): Likewise.
22433 (aarch64_expand_sve_const_vector): Likewise.
22434 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22435 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22436 into the register, rather than emitting a SET directly.
22437 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22438 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22439 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22440 (offset_9bit_signed_scaled_p): New functions.
22441 (aarch64_replicate_bitmask_imm): New function.
22442 (aarch64_bitmask_imm): Use it.
22443 (aarch64_cannot_force_const_mem): Reject expressions involving
22444 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22445 (aarch64_classify_index): Handle SVE indices, by requiring
22446 a plain register index with a scale that matches the element size.
22447 (aarch64_classify_address): Handle SVE addresses. Assert that
22448 the mode of the address is VOIDmode or an integer mode.
22449 Update call to aarch64_classify_symbol.
22450 (aarch64_classify_symbolic_expression): Update call to
22451 aarch64_classify_symbol.
22452 (aarch64_const_vec_all_in_range_p): New function.
22453 (aarch64_print_vector_float_operand): Likewise.
22454 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22455 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22456 and the FP immediates 1.0 and 0.5.
22457 (aarch64_print_address_internal): Handle SVE addresses.
22458 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22459 (aarch64_regno_regclass): Handle predicate registers.
22460 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22462 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22463 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22464 (aarch64_convert_sve_vector_bits): New function.
22465 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22466 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22467 rather than an rtx.
22468 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22469 Handle SVE vector and predicate modes. Accept VL-based constants
22470 that need only one temporary register, and VL offsets that require
22471 no temporary registers.
22472 (aarch64_conditional_register_usage): Mark the predicate registers
22473 as fixed if SVE isn't available.
22474 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22475 Return true for SVE vector and predicate modes.
22476 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22477 rather than an unsigned int. Handle SVE modes.
22478 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22480 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22482 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22483 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22484 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22485 (aarch64_sve_float_mul_immediate_p): New functions.
22486 (aarch64_sve_valid_immediate): New function.
22487 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22488 Explicitly reject structure modes. Check for INDEX constants.
22489 Handle PTRUE and PFALSE constants.
22490 (aarch64_check_zero_based_sve_index_immediate): New function.
22491 (aarch64_simd_imm_zero_p): Delete.
22492 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22493 vector modes. Accept constants in the range of CNT[BHWD].
22494 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22495 ask for an Advanced SIMD mode.
22496 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22497 (aarch64_simd_vector_alignment): Handle SVE predicates.
22498 (aarch64_vectorize_preferred_vector_alignment): New function.
22499 (aarch64_simd_vector_alignment_reachable): Use it instead of
22501 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22502 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22504 (MAX_VECT_LEN): Delete.
22505 (expand_vec_perm_d): Add a vec_flags field.
22506 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22507 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22508 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22510 (aarch64_evpc_rev): Rename to...
22511 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22512 (aarch64_evpc_rev_global): New function.
22513 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22514 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22516 (aarch64_evpc_sve_tbl): New function.
22517 (aarch64_expand_vec_perm_const_1): Update after rename of
22518 aarch64_evpc_rev. Handle SVE permutes too, trying
22519 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22520 than aarch64_evpc_tbl.
22521 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22522 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22523 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22524 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22525 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22526 (aarch64_expand_sve_vcond): New functions.
22527 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22528 of aarch64_vector_mode_p.
22529 (aarch64_dwarf_poly_indeterminate_value): New function.
22530 (aarch64_compute_pressure_classes): Likewise.
22531 (aarch64_can_change_mode_class): Likewise.
22532 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22533 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22534 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22535 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22536 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22537 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22538 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22539 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22541 (Dn, Dl, Dr): Accept const as well as const_vector.
22542 (Dz): Likewise. Compare against CONST0_RTX.
22543 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22544 of "vector" where appropriate.
22545 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22546 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22547 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22548 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22549 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22550 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22551 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22552 (v_int_equiv): Extend to SVE modes.
22553 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22555 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22556 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22557 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22558 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22559 (SVE_COND_FP_CMP): New int iterators.
22560 (perm_hilo): Handle the new unpack unspecs.
22561 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22563 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22564 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22565 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22566 (aarch64_equality_operator, aarch64_constant_vector_operand)
22567 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22568 (aarch64_sve_nonimmediate_operand): Likewise.
22569 (aarch64_sve_general_operand): Likewise.
22570 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22571 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22572 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22573 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22574 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22575 (aarch64_sve_float_arith_immediate): Likewise.
22576 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22577 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22578 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22579 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22580 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22581 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22582 (aarch64_sve_float_arith_operand): Likewise.
22583 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22584 (aarch64_sve_float_mul_operand): Likewise.
22585 (aarch64_sve_vec_perm_operand): Likewise.
22586 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22587 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22588 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22589 as well as const_vector.
22590 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22591 in file. Use CONST0_RTX and CONSTM1_RTX.
22592 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22593 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22594 Use aarch64_simd_imm_zero.
22595 * config/aarch64/aarch64-sve.md: New file.
22596 * config/aarch64/aarch64.md: Include it.
22597 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22598 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22599 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22600 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22601 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22602 (sve): New attribute.
22603 (enabled): Disable instructions with the sve attribute unless
22605 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22606 aarch64_expand_mov_immediate.
22607 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22608 CNT[BHSD] immediates.
22609 (movti): Split CONST_POLY_INT moves into two halves.
22610 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22611 Split additions that need a temporary here if the destination
22612 is the stack pointer.
22613 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22614 (*add<mode>3_poly_1): New instruction.
22615 (set_clobber_cc): New expander.
22617 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22619 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22620 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22621 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22622 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22623 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22624 Change innermode from fixed_mode_size to machine_mode.
22625 (simplify_subreg): Update call accordingly. Handle a constant-sized
22626 subreg of a variable-length CONST_VECTOR.
22628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22629 Alan Hayward <alan.hayward@arm.com>
22630 David Sherwood <david.sherwood@arm.com>
22632 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22633 (add_offset_to_base): New function, split out from...
22634 (create_mem_ref): ...here. When handling a scale other than 1,
22635 check first whether the address is valid without the offset.
22636 Add it into the base if so, leaving the index and scale as-is.
22638 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22641 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22642 fold_for_warn before checking if arg2 is INTEGER_CST.
22644 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
22646 * config/rs6000/predicates.md (load_multiple_operation): Delete.
22647 (store_multiple_operation): Delete.
22648 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22649 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22650 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22651 guarded by TARGET_STRING.
22652 (rs6000_output_load_multiple): Delete.
22653 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22654 OPTION_MASK_STRING / TARGET_STRING handling.
22655 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22656 (const rs6000_opt_masks) <"string">: Change mask to 0.
22657 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22658 (MASK_STRING): Delete.
22659 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22661 (load_multiple): Delete.
22668 (store_multiple): Delete.
22675 (movmemsi_8reg): Delete.
22676 (corresponding unnamed define_insn): Delete.
22677 (movmemsi_6reg): Delete.
22678 (corresponding unnamed define_insn): Delete.
22679 (movmemsi_4reg): Delete.
22680 (corresponding unnamed define_insn): Delete.
22681 (movmemsi_2reg): Delete.
22682 (corresponding unnamed define_insn): Delete.
22683 (movmemsi_1reg): Delete.
22684 (corresponding unnamed define_insn): Delete.
22685 * config/rs6000/rs6000.opt (mno-string): New.
22686 (mstring): Replace by deprecation warning stub.
22687 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
22689 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22691 * regrename.c (regrename_do_replace): If replacing the same
22692 reg multiple times, try to reuse last created gen_raw_REG.
22695 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
22696 main to workaround a bug in GDB.
22698 2018-01-12 Tom de Vries <tom@codesourcery.com>
22701 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
22703 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
22705 PR rtl-optimization/80481
22706 * ira-color.c (get_cap_member): New function.
22707 (allocnos_conflict_by_live_ranges_p): Use it.
22708 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
22709 (setup_slot_coalesced_allocno_live_ranges): Ditto.
22711 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
22714 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
22715 (*saddl_se_1): Ditto.
22716 (*ssubsi_1): Ditto.
22717 (*ssubl_se_1): Ditto.
22719 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22721 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
22722 rather than wi::to_widest for DR_INITs.
22723 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
22724 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
22725 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
22727 (vect_analyze_group_access_1): Note that here.
22729 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22731 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
22732 polynomial type sizes.
22734 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22736 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
22737 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
22738 (gimple_add_tmp_var): Likewise.
22740 2018-01-12 Martin Liska <mliska@suse.cz>
22742 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
22743 (gimple_alloc_sizes): Likewise.
22744 (dump_gimple_statistics): Use PRIu64 in printf format.
22745 * gimple.h: Change uint64_t to int.
22747 2018-01-12 Martin Liska <mliska@suse.cz>
22749 * tree-core.h: Use uint64_t instead of int.
22750 * tree.c (tree_node_counts): Likewise.
22751 (tree_node_sizes): Likewise.
22752 (dump_tree_statistics): Use PRIu64 in printf format.
22754 2018-01-12 Martin Liska <mliska@suse.cz>
22756 * Makefile.in: As qsort_chk is implemented in vec.c, add
22757 vec.o to linkage of gencfn-macros.
22758 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
22759 passing the info to record_node_allocation_statistics.
22760 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
22762 * ggc-common.c (struct ggc_usage): Add operator== and use
22763 it in operator< and compare function.
22764 * mem-stats.h (struct mem_usage): Likewise.
22765 * vec.c (struct vec_usage): Remove operator< and compare
22766 function. Can be simply inherited.
22768 2018-01-12 Martin Jambor <mjambor@suse.cz>
22771 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
22772 * tree-ssa-math-opts.c: Include domwalk.h.
22773 (convert_mult_to_fma_1): New function.
22774 (fma_transformation_info): New type.
22775 (fma_deferring_state): Likewise.
22776 (cancel_fma_deferring): New function.
22777 (result_of_phi): Likewise.
22778 (last_fma_candidate_feeds_initial_phi): Likewise.
22779 (convert_mult_to_fma): Added deferring logic, split actual
22780 transformation to convert_mult_to_fma_1.
22781 (math_opts_dom_walker): New type.
22782 (math_opts_dom_walker::after_dom_children): New method, body moved
22783 here from pass_optimize_widening_mul::execute, added deferring logic
22785 (pass_optimize_widening_mul::execute): Moved most of code to
22786 math_opts_dom_walker::after_dom_children.
22787 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22788 * config/i386/i386.c (ix86_option_override_internal): Added
22789 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22791 2018-01-12 Richard Biener <rguenther@suse.de>
22794 * dwarf2out.c (gen_variable_die): Do not reset old_die for
22795 inline instance vars.
22797 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
22800 * config/rx/rx.c (rx_is_restricted_memory_address):
22801 Handle SUBREG case.
22803 2018-01-12 Richard Biener <rguenther@suse.de>
22805 PR tree-optimization/80846
22806 * target.def (split_reduction): New target hook.
22807 * targhooks.c (default_split_reduction): New function.
22808 * targhooks.h (default_split_reduction): Declare.
22809 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22810 target requests first reduce vectors by combining low and high
22812 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22813 (get_vectype_for_scalar_type_and_size): Export.
22814 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22815 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22816 * doc/tm.texi: Regenerate.
22817 * config/i386/i386.c (ix86_split_reduction): Implement
22818 TARGET_VECTORIZE_SPLIT_REDUCTION.
22820 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22823 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22824 in PIC mode except for TARGET_VXWORKS_RTP.
22825 * config/sparc/sparc.c: Include cfgrtl.h.
22826 (TARGET_INIT_PIC_REG): Define.
22827 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22828 (sparc_pic_register_p): New predicate.
22829 (sparc_legitimate_address_p): Use it.
22830 (sparc_legitimize_pic_address): Likewise.
22831 (sparc_delegitimize_address): Likewise.
22832 (sparc_mode_dependent_address_p): Likewise.
22833 (gen_load_pcrel_sym): Remove 4th parameter.
22834 (load_got_register): Adjust call to above. Remove obsolete stuff.
22835 (sparc_expand_prologue): Do not call load_got_register here.
22836 (sparc_flat_expand_prologue): Likewise.
22837 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22838 (sparc_use_pseudo_pic_reg): New function.
22839 (sparc_init_pic_reg): Likewise.
22840 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22841 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22843 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22845 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22846 Add item for branch_cost.
22848 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22850 PR rtl-optimization/83565
22851 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22852 not extend the result to a larger mode for rotate operations.
22853 (num_sign_bit_copies1): Likewise.
22855 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22858 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22860 Use values-Xc.o for -pedantic.
22861 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22863 2018-01-12 Martin Liska <mliska@suse.cz>
22866 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22868 (possible_polymorphic_call_targets): Use it.
22869 (ipa_devirt): Likewise.
22871 2018-01-12 Martin Liska <mliska@suse.cz>
22873 * profile-count.h (enum profile_quality): Use 0 as invalid
22874 enum value of profile_quality.
22876 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22878 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22879 -mext-string options.
22881 2018-01-12 Richard Biener <rguenther@suse.de>
22883 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22884 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22885 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22887 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22889 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22891 * configure.ac (--with-long-double-format): Add support for the
22892 configuration option to change the default long double format on
22894 * config.gcc (powerpc*-linux*-*): Likewise.
22895 * configure: Regenerate.
22896 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22897 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22898 used without modification.
22900 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22902 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22903 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22904 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22905 MISC_BUILTIN_SPEC_BARRIER.
22906 (rs6000_init_builtins): Likewise.
22907 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22909 (speculation_barrier): New define_insn.
22910 * doc/extend.texi: Document __builtin_speculation_barrier.
22912 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22915 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22916 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22917 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22919 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22920 integral modes instead of "ss" and "sd".
22921 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22922 vectors with 32-bit and 64-bit elements.
22923 (vecdupssescalarmodesuffix): New mode attribute.
22924 (vec_dup<mode>): Use it.
22926 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22929 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22930 frame if argument is passed on stack.
22932 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22935 * ree.c (combine_reaching_defs): Optimize also
22936 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22937 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22939 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22941 PR middle-end/83189
22942 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22944 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22946 PR middle-end/83718
22947 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22948 after they are computed.
22950 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22952 PR tree-optimization/83695
22953 * gimple-loop-linterchange.cc
22954 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22955 reset cached scev information after interchange.
22956 (pass_linterchange::execute): Remove call to scev_reset_htab.
22958 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22960 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22961 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22962 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22963 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22964 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22965 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22966 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22967 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22968 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22969 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22970 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22971 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22972 (V_lane_reg): Likewise.
22973 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22975 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22976 (vfmal_lane_low<mode>_intrinsic,
22977 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22978 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22979 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22980 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22981 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22982 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22984 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22986 * config/arm/arm-cpus.in (fp16fml): New feature.
22987 (ALL_SIMD): Add fp16fml.
22988 (armv8.2-a): Add fp16fml as an option.
22989 (armv8.3-a): Likewise.
22990 (armv8.4-a): Add fp16fml as part of fp16.
22991 * config/arm/arm.h (TARGET_FP16FML): Define.
22992 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22994 * config/arm/arm-modes.def (V2HF): Define.
22995 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22996 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22997 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22998 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22999 vfmsl_low, vfmsl_high): New set of builtins.
23000 * config/arm/iterators.md (PLUSMINUS): New code iterator.
23001 (vfml_op): New code attribute.
23002 (VFMLHALVES): New int iterator.
23003 (VFML, VFMLSEL): New mode attributes.
23004 (V_reg): Define mapping for V2HF.
23005 (V_hi, V_lo): New mode attributes.
23006 (VF_constraint): Likewise.
23007 (vfml_half, vfml_half_selector): New int attributes.
23008 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23010 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23011 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23013 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23014 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23015 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23016 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23018 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23019 Document new effective target and option set.
23021 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23023 * config/arm/arm-cpus.in (armv8_4): New feature.
23024 (ARMv8_4a): New fgroup.
23025 (armv8.4-a): New arch.
23026 * config/arm/arm-tables.opt: Regenerate.
23027 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23028 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23029 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23030 Add matching rules for -march=armv8.4-a and extensions.
23031 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23033 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23036 * config/rx/rx.md (BW): New mode attribute.
23037 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23039 2018-01-11 Richard Biener <rguenther@suse.de>
23041 PR tree-optimization/83435
23042 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23043 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23044 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23046 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23047 Alan Hayward <alan.hayward@arm.com>
23048 David Sherwood <david.sherwood@arm.com>
23050 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23052 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23053 (aarch64_print_address_internal): Use it to check for a zero offset.
23055 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23056 Alan Hayward <alan.hayward@arm.com>
23057 David Sherwood <david.sherwood@arm.com>
23059 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23060 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23061 Return a poly_int64 rather than a HOST_WIDE_INT.
23062 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23063 rather than a HOST_WIDE_INT.
23064 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23065 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23066 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23067 final_offset from HOST_WIDE_INT to poly_int64.
23068 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23069 to_constant when getting the number of units in an Advanced SIMD
23071 (aarch64_builtin_vectorized_function): Check for a constant number
23073 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23075 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23076 attribute instead of GET_MODE_NUNITS.
23077 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23078 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23079 GET_MODE_SIZE for fixed-size registers.
23080 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23081 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23082 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23083 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23084 (aarch64_print_operand, aarch64_print_address_internal)
23085 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23086 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23087 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23088 Handle polynomial GET_MODE_SIZE.
23089 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23090 wider than SImode without modification.
23091 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23092 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23093 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23094 passing and returning SVE modes.
23095 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23096 rather than GEN_INT.
23097 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23098 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23099 (aarch64_allocate_and_probe_stack_space): Likewise.
23100 (aarch64_layout_frame): Cope with polynomial offsets.
23101 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23102 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23103 polynomial offsets.
23104 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23105 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23106 poly_int64 rather than a HOST_WIDE_INT.
23107 (aarch64_get_separate_components, aarch64_process_components)
23108 (aarch64_expand_prologue, aarch64_expand_epilogue)
23109 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23110 (aarch64_anchor_offset): New function, split out from...
23111 (aarch64_legitimize_address): ...here.
23112 (aarch64_builtin_vectorization_cost): Handle polynomial
23113 TYPE_VECTOR_SUBPARTS.
23114 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23116 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23117 number of elements from the PARALLEL rather than the mode.
23118 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23119 rather than GET_MODE_BITSIZE.
23120 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23121 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23122 (aarch64_expand_vec_perm_const_1): Handle polynomial
23123 d->perm.length () and d->perm elements.
23124 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23125 Apply to_constant to d->perm elements.
23126 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23127 polynomial CONST_VECTOR_NUNITS.
23128 (aarch64_move_pointer): Take amount as a poly_int64 rather
23130 (aarch64_progress_pointer): Avoid temporary variable.
23131 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23132 the mode attribute instead of GET_MODE.
23134 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23135 Alan Hayward <alan.hayward@arm.com>
23136 David Sherwood <david.sherwood@arm.com>
23138 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23139 x exists before using it.
23140 (aarch64_add_constant_internal): Rename to...
23141 (aarch64_add_offset_1): ...this. Replace regnum with separate
23142 src and dest rtxes. Handle the case in which they're different,
23143 including when the offset is zero. Replace scratchreg with an rtx.
23144 Use 2 additions if there is no spare register into which we can
23145 move a 16-bit constant.
23146 (aarch64_add_constant): Delete.
23147 (aarch64_add_offset): Replace reg with separate src and dest
23148 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23149 Use aarch64_add_offset_1.
23150 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23151 an rtx rather than an int. Take the delta as a poly_int64
23152 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23153 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23154 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23155 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23156 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23157 and aarch64_add_sp.
23158 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23159 aarch64_add_constant.
23161 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23163 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23164 Use scalar_float_mode.
23166 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23168 * config/aarch64/aarch64-simd.md
23169 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23170 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23171 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23172 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23173 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23174 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23175 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23176 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23177 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23178 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23180 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23183 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23184 targ_options->x_arm_arch_string is non NULL.
23186 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23188 * config/aarch64/aarch64.h
23189 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23191 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23194 * expmed.c (emit_store_flag_force): Swap if const op0
23195 and change VOIDmode to mode of op0.
23197 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23199 PR rtl-optimization/83761
23200 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23201 than bytes to mode_for_size.
23203 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23205 PR middle-end/83189
23206 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23207 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23210 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23212 PR middle-end/83575
23213 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23214 when in layout mode.
23215 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23216 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23219 2018-01-10 Michael Collison <michael.collison@arm.com>
23221 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23222 * config/aarch64/aarch64-option-extension.def: Add
23223 AARCH64_OPT_EXTENSION of 'fp16fml'.
23224 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23225 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23226 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23227 * config/aarch64/constraints.md (Ui7): New constraint.
23228 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23229 (VFMLA_SEL_W): Ditto.
23232 (VFMLA16_LOW): New int iterator.
23233 (VFMLA16_HIGH): Ditto.
23234 (UNSPEC_FMLAL): New unspec.
23235 (UNSPEC_FMLSL): Ditto.
23236 (UNSPEC_FMLAL2): Ditto.
23237 (UNSPEC_FMLSL2): Ditto.
23238 (f16mac): New code attribute.
23239 * config/aarch64/aarch64-simd-builtins.def
23240 (aarch64_fmlal_lowv2sf): Ditto.
23241 (aarch64_fmlsl_lowv2sf): Ditto.
23242 (aarch64_fmlalq_lowv4sf): Ditto.
23243 (aarch64_fmlslq_lowv4sf): Ditto.
23244 (aarch64_fmlal_highv2sf): Ditto.
23245 (aarch64_fmlsl_highv2sf): Ditto.
23246 (aarch64_fmlalq_highv4sf): Ditto.
23247 (aarch64_fmlslq_highv4sf): Ditto.
23248 (aarch64_fmlal_lane_lowv2sf): Ditto.
23249 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23250 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23251 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23252 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23253 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23254 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23255 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23256 (aarch64_fmlal_lane_highv2sf): Ditto.
23257 (aarch64_fmlsl_lane_highv2sf): Ditto.
23258 (aarch64_fmlal_laneq_highv2sf): Ditto.
23259 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23260 (aarch64_fmlalq_lane_highv4sf): Ditto.
23261 (aarch64_fmlsl_lane_highv4sf): Ditto.
23262 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23263 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23264 * config/aarch64/aarch64-simd.md:
23265 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23266 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23267 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23268 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23269 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23270 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23271 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23272 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23273 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23274 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23275 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23276 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23277 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23278 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23279 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23280 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23281 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23282 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23283 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23284 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23285 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23286 (vfmlsl_low_u32): Ditto.
23287 (vfmlalq_low_u32): Ditto.
23288 (vfmlslq_low_u32): Ditto.
23289 (vfmlal_high_u32): Ditto.
23290 (vfmlsl_high_u32): Ditto.
23291 (vfmlalq_high_u32): Ditto.
23292 (vfmlslq_high_u32): Ditto.
23293 (vfmlal_lane_low_u32): Ditto.
23294 (vfmlsl_lane_low_u32): Ditto.
23295 (vfmlal_laneq_low_u32): Ditto.
23296 (vfmlsl_laneq_low_u32): Ditto.
23297 (vfmlalq_lane_low_u32): Ditto.
23298 (vfmlslq_lane_low_u32): Ditto.
23299 (vfmlalq_laneq_low_u32): Ditto.
23300 (vfmlslq_laneq_low_u32): Ditto.
23301 (vfmlal_lane_high_u32): Ditto.
23302 (vfmlsl_lane_high_u32): Ditto.
23303 (vfmlal_laneq_high_u32): Ditto.
23304 (vfmlsl_laneq_high_u32): Ditto.
23305 (vfmlalq_lane_high_u32): Ditto.
23306 (vfmlslq_lane_high_u32): Ditto.
23307 (vfmlalq_laneq_high_u32): Ditto.
23308 (vfmlslq_laneq_high_u32): Ditto.
23309 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23310 (AARCH64_FL_FOR_ARCH8_4): New.
23311 (AARCH64_ISA_F16FML): New ISA flag.
23312 (TARGET_F16FML): New feature flag for fp16fml.
23313 (doc/invoke.texi): Document new fp16fml option.
23315 2018-01-10 Michael Collison <michael.collison@arm.com>
23317 * config/aarch64/aarch64-builtins.c:
23318 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23319 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23320 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23321 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23322 (AARCH64_ISA_SHA3): New ISA flag.
23323 (TARGET_SHA3): New feature flag for sha3.
23324 * config/aarch64/iterators.md (sha512_op): New int attribute.
23325 (CRYPTO_SHA512): New int iterator.
23326 (UNSPEC_SHA512H): New unspec.
23327 (UNSPEC_SHA512H2): Ditto.
23328 (UNSPEC_SHA512SU0): Ditto.
23329 (UNSPEC_SHA512SU1): Ditto.
23330 * config/aarch64/aarch64-simd-builtins.def
23331 (aarch64_crypto_sha512hqv2di): New builtin.
23332 (aarch64_crypto_sha512h2qv2di): Ditto.
23333 (aarch64_crypto_sha512su0qv2di): Ditto.
23334 (aarch64_crypto_sha512su1qv2di): Ditto.
23335 (aarch64_eor3qv8hi): Ditto.
23336 (aarch64_rax1qv2di): Ditto.
23337 (aarch64_xarqv2di): Ditto.
23338 (aarch64_bcaxqv8hi): Ditto.
23339 * config/aarch64/aarch64-simd.md:
23340 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23341 (aarch64_crypto_sha512su0qv2di): Ditto.
23342 (aarch64_crypto_sha512su1qv2di): Ditto.
23343 (aarch64_eor3qv8hi): Ditto.
23344 (aarch64_rax1qv2di): Ditto.
23345 (aarch64_xarqv2di): Ditto.
23346 (aarch64_bcaxqv8hi): Ditto.
23347 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23348 (vsha512h2q_u64): Ditto.
23349 (vsha512su0q_u64): Ditto.
23350 (vsha512su1q_u64): Ditto.
23351 (veor3q_u16): Ditto.
23352 (vrax1q_u64): Ditto.
23353 (vxarq_u64): Ditto.
23354 (vbcaxq_u16): Ditto.
23355 * config/arm/types.md (crypto_sha512): New type attribute.
23356 (crypto_sha3): Ditto.
23357 (doc/invoke.texi): Document new sha3 option.
23359 2018-01-10 Michael Collison <michael.collison@arm.com>
23361 * config/aarch64/aarch64-builtins.c:
23362 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23363 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23364 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23365 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23366 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23367 (AARCH64_ISA_SM4): New ISA flag.
23368 (TARGET_SM4): New feature flag for sm4.
23369 * config/aarch64/aarch64-simd-builtins.def
23370 (aarch64_sm3ss1qv4si): Ditto.
23371 (aarch64_sm3tt1aq4si): Ditto.
23372 (aarch64_sm3tt1bq4si): Ditto.
23373 (aarch64_sm3tt2aq4si): Ditto.
23374 (aarch64_sm3tt2bq4si): Ditto.
23375 (aarch64_sm3partw1qv4si): Ditto.
23376 (aarch64_sm3partw2qv4si): Ditto.
23377 (aarch64_sm4eqv4si): Ditto.
23378 (aarch64_sm4ekeyqv4si): Ditto.
23379 * config/aarch64/aarch64-simd.md:
23380 (aarch64_sm3ss1qv4si): Ditto.
23381 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23382 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23383 (aarch64_sm4eqv4si): Ditto.
23384 (aarch64_sm4ekeyqv4si): Ditto.
23385 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23386 (sm3part_op): Ditto.
23387 (CRYPTO_SM3TT): Ditto.
23388 (CRYPTO_SM3PART): Ditto.
23389 (UNSPEC_SM3SS1): New unspec.
23390 (UNSPEC_SM3TT1A): Ditto.
23391 (UNSPEC_SM3TT1B): Ditto.
23392 (UNSPEC_SM3TT2A): Ditto.
23393 (UNSPEC_SM3TT2B): Ditto.
23394 (UNSPEC_SM3PARTW1): Ditto.
23395 (UNSPEC_SM3PARTW2): Ditto.
23396 (UNSPEC_SM4E): Ditto.
23397 (UNSPEC_SM4EKEY): Ditto.
23398 * config/aarch64/constraints.md (Ui2): New constraint.
23399 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23400 * config/arm/types.md (crypto_sm3): New type attribute.
23401 (crypto_sm4): Ditto.
23402 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23403 (vsm3tt1aq_u32): Ditto.
23404 (vsm3tt1bq_u32): Ditto.
23405 (vsm3tt2aq_u32): Ditto.
23406 (vsm3tt2bq_u32): Ditto.
23407 (vsm3partw1q_u32): Ditto.
23408 (vsm3partw2q_u32): Ditto.
23409 (vsm4eq_u32): Ditto.
23410 (vsm4ekeyq_u32): Ditto.
23411 (doc/invoke.texi): Document new sm4 option.
23413 2018-01-10 Michael Collison <michael.collison@arm.com>
23415 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23416 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23417 (AARCH64_FL_FOR_ARCH8_4): New.
23418 (AARCH64_FL_V8_4): New flag.
23419 (doc/invoke.texi): Document new armv8.4-a option.
23421 2018-01-10 Michael Collison <michael.collison@arm.com>
23423 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23424 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23425 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23426 * config/aarch64/aarch64-option-extension.def: Add
23427 AARCH64_OPT_EXTENSION of 'sha2'.
23428 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23429 (crypto): Disable sha2 and aes if crypto disabled.
23430 (crypto): Enable aes and sha2 if enabled.
23431 (simd): Disable sha2 and aes if simd disabled.
23432 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23434 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23435 (TARGET_SHA2): New feature flag for sha2.
23436 (TARGET_AES): New feature flag for aes.
23437 * config/aarch64/aarch64-simd.md:
23438 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23439 conditional on TARGET_AES.
23440 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23441 (aarch64_crypto_sha1hsi): Make pattern conditional
23443 (aarch64_crypto_sha1hv4si): Ditto.
23444 (aarch64_be_crypto_sha1hv4si): Ditto.
23445 (aarch64_crypto_sha1su1v4si): Ditto.
23446 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23447 (aarch64_crypto_sha1su0v4si): Ditto.
23448 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23449 (aarch64_crypto_sha256su0v4si): Ditto.
23450 (aarch64_crypto_sha256su1v4si): Ditto.
23451 (doc/invoke.texi): Document new aes and sha2 options.
23453 2018-01-10 Martin Sebor <msebor@redhat.com>
23455 PR tree-optimization/83781
23456 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23459 2018-01-11 Martin Sebor <msebor@gmail.com>
23460 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23462 PR tree-optimization/83501
23463 PR tree-optimization/81703
23465 * tree-ssa-strlen.c (get_string_cst): Rename...
23466 (get_string_len): ...to this. Handle global constants.
23467 (handle_char_store): Adjust.
23469 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23470 Jim Wilson <jimw@sifive.com>
23472 * config/riscv/riscv-protos.h (riscv_output_return): New.
23473 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23474 (riscv_attribute_table, riscv_output_return),
23475 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23476 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23477 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23479 (riscv_expand_prologue): Add early return for naked function.
23480 (riscv_expand_epilogue): Likewise.
23481 (riscv_function_ok_for_sibcall): Return false for naked function.
23482 (riscv_set_current_function): New.
23483 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23484 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23485 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23486 * doc/extend.texi (RISC-V Function Attributes): New.
23488 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23490 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23491 check for 128-bit long double before checking TCmode.
23492 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23493 128-bit long doubles before checking TFmode or TCmode.
23494 (FLOAT128_IBM_P): Likewise.
23496 2018-01-10 Martin Sebor <msebor@redhat.com>
23498 PR tree-optimization/83671
23499 * builtins.c (c_strlen): Unconditionally return zero for the empty
23501 Use -Warray-bounds for warnings.
23502 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23503 for non-constant array indices with COMPONENT_REF, arrays of
23504 arrays, and pointers to arrays.
23505 (gimple_fold_builtin_strlen): Determine and set length range for
23506 non-constant character arrays.
23508 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23510 PR middle-end/81897
23511 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23514 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23516 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23518 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23521 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23522 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23523 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23524 indexed_or_indirect_operand predicate.
23525 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23526 (*vsx_le_perm_load_v8hi): Likewise.
23527 (*vsx_le_perm_load_v16qi): Likewise.
23528 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23529 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23530 (*vsx_le_perm_store_v8hi): Likewise.
23531 (*vsx_le_perm_store_v16qi): Likewise.
23532 (eight unnamed splitters): Likewise.
23534 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23536 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23537 * config/rs6000/emmintrin.h: Likewise.
23538 * config/rs6000/mmintrin.h: Likewise.
23539 * config/rs6000/xmmintrin.h: Likewise.
23541 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23544 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23546 * tree.c (tree_nop_conversion): Return true for location wrapper
23548 (maybe_wrap_with_location): New function.
23549 (selftest::check_strip_nops): New function.
23550 (selftest::test_location_wrappers): New function.
23551 (selftest::tree_c_tests): Call it.
23552 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23553 (maybe_wrap_with_location): New decl.
23554 (EXPR_LOCATION_WRAPPER_P): New macro.
23555 (location_wrapper_p): New inline function.
23556 (tree_strip_any_location_wrapper): New inline function.
23558 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23561 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23562 stack_realign_offset for the largest alignment of stack slot
23564 (ix86_find_max_used_stack_alignment): New function.
23565 (ix86_finalize_stack_frame_flags): Use it. Set
23566 max_used_stack_alignment if we don't realign stack.
23567 * config/i386/i386.h (machine_function): Add
23568 max_used_stack_alignment.
23570 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23572 * config/arm/arm.opt (-mbranch-cost): New option.
23573 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23576 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23579 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23580 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23582 2018-01-10 Richard Biener <rguenther@suse.de>
23585 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23586 early out so it also covers the case where we have a non-NULL
23589 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23591 PR tree-optimization/83753
23592 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23593 for non-strided grouped accesses if the number of elements is 1.
23595 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23598 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23599 * i386.h (TARGET_USE_GATHER): Define.
23600 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23602 2018-01-10 Martin Liska <mliska@suse.cz>
23605 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23606 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23608 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23609 CLEANUP_NO_PARTITIONING is not set.
23611 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23613 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23614 for vectors, as a partial revert of r254296.
23615 * rtl.h (const_vec_p): Delete.
23616 (const_vec_duplicate_p): Don't test for vector CONSTs.
23617 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23618 * expmed.c (make_tree): Likewise.
23621 * common.md (E, F): Use CONSTANT_P instead of checking for
23623 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23624 checking for CONST_VECTOR.
23626 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23628 PR middle-end/83575
23629 * predict.c (force_edge_cold): Handle in more sane way edges
23630 with no prediction.
23632 2018-01-09 Carl Love <cel@us.ibm.com>
23634 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23636 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23637 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23638 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23639 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
23640 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23641 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
23642 * config/rs6000/rs6000-protos.h: Add extern defition for
23643 rs6000_generate_float2_double_code.
23644 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23646 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23647 (float2_v2df): Add define_expand.
23649 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
23652 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23653 op_mode in the force_to_mode call.
23655 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23657 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23658 instead of checking each element individually.
23659 (aarch64_evpc_uzp): Likewise.
23660 (aarch64_evpc_zip): Likewise.
23661 (aarch64_evpc_ext): Likewise.
23662 (aarch64_evpc_rev): Likewise.
23663 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23664 instead of checking each element individually. Return true without
23666 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23667 whether all selected elements come from the same input, instead of
23668 checking each element individually. Remove calls to gen_rtx_REG,
23669 start_sequence and end_sequence and instead assert that no rtl is
23672 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23674 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
23675 order of HIGH and CONST checks.
23677 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23679 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
23680 if the destination isn't an SSA_NAME.
23682 2018-01-09 Richard Biener <rguenther@suse.de>
23684 PR tree-optimization/83668
23685 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
23687 (canonicalize_loop_form): ... here, renamed from ...
23688 (canonicalize_loop_closed_ssa_form): ... this and amended to
23689 swap successor edges for loop exit blocks to make us use
23690 the RPO order we need for initial schedule generation.
23692 2018-01-09 Joseph Myers <joseph@codesourcery.com>
23694 PR tree-optimization/64811
23695 * match.pd: When optimizing comparisons with Inf, avoid
23696 introducing or losing exceptions from comparisons with NaN.
23698 2018-01-09 Martin Liska <mliska@suse.cz>
23701 * asan.c (shadow_mem_size): Add gcc_assert.
23703 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
23705 Don't save registers in main().
23708 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
23709 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
23710 * config/avr/avr.c (avr_set_current_function): Don't error if
23711 naked, OS_task or OS_main are specified at the same time.
23712 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
23714 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
23716 * common/config/avr/avr-common.c (avr_option_optimization_table):
23717 Switch on -mmain-is-OS_task for optimizing compilations.
23719 2018-01-09 Richard Biener <rguenther@suse.de>
23721 PR tree-optimization/83572
23722 * graphite.c: Include cfganal.h.
23723 (graphite_transform_loops): Connect infinite loops to exit
23724 and remove fake edges at the end.
23726 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23728 * ipa-inline.c (edge_badness): Revert accidental checkin.
23730 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23733 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
23734 symbols; not inline clones.
23736 2018-01-09 Jakub Jelinek <jakub@redhat.com>
23739 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
23740 hard registers. Formatting fixes.
23742 PR preprocessor/83722
23743 * gcc.c (try_generate_repro): Pass
23744 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
23745 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
23748 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
23749 Kito Cheng <kito.cheng@gmail.com>
23751 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
23752 (riscv_leaf_function_p): Delete.
23753 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
23755 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23757 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
23759 (do_ifelse): New function.
23760 (do_isel): New function.
23761 (do_sub3): New function.
23762 (do_add3): New function.
23763 (do_load_mask_compare): New function.
23764 (do_overlap_load_compare): New function.
23765 (expand_compare_loop): New function.
23766 (expand_block_compare): Call expand_compare_loop() when appropriate.
23767 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
23768 option description.
23769 (-mblock-compare-inline-loop-limit): New option.
23771 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23774 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
23775 Reverse order of second and third operands in first alternative.
23776 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
23777 of first and second elements in UNSPEC_VPERMR vector.
23778 (altivec_expand_vec_perm_le): Likewise.
23780 2018-01-08 Jeff Law <law@redhat.com>
23782 PR rtl-optimizatin/81308
23783 * tree-switch-conversion.c (cfg_altered): New file scoped static.
23784 (process_switch): If group_case_labels makes a change, then set
23786 (pass_convert_switch::execute): If a switch is converted, then
23787 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
23789 PR rtl-optimization/81308
23790 * recog.c (split_all_insns): Conditionally cleanup the CFG after
23793 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
23795 PR target/83663 - Revert r255946
23796 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23797 generation for cases where splatting a value is not useful.
23798 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23799 across a vec_duplicate and a paradoxical subreg forming a vector
23800 mode to a vec_concat.
23802 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23804 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23805 -march=armv8.3-a variants.
23806 * config/arm/t-multilib: Likewise.
23807 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23809 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23811 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23813 (cceq_ior_compare_complement): Give it a name so I can use it, and
23814 change boolean_or_operator predicate to boolean_operator so it can
23815 be used to generate a crand.
23816 (eqne): New code iterator.
23817 (bd/bd_neg): New code_attrs.
23818 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23819 a single define_insn.
23820 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23821 decrement (bdnzt/bdnzf/bdzt/bdzf).
23822 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23823 with the new names of the branch decrement patterns, and added the
23824 names of the branch decrement conditional patterns.
23826 2018-01-08 Richard Biener <rguenther@suse.de>
23828 PR tree-optimization/83563
23829 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23832 2018-01-08 Richard Biener <rguenther@suse.de>
23834 PR middle-end/83713
23835 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23837 2018-01-08 Richard Biener <rguenther@suse.de>
23839 PR tree-optimization/83685
23840 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23841 references to abnormals.
23843 2018-01-08 Richard Biener <rguenther@suse.de>
23846 * dwarf2out.c (output_indirect_strings): Handle empty
23847 skeleton_debug_str_hash.
23848 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23850 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23852 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23853 (emit_store_direct): Likewise.
23854 (arc_trampoline_adjust_address): Likewise.
23855 (arc_asm_trampoline_template): New function.
23856 (arc_initialize_trampoline): Use asm_trampoline_template.
23857 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23858 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23859 * config/arc/arc.md (flush_icache): Delete pattern.
23861 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23863 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23864 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23867 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23870 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23871 by not USED_FOR_TARGET.
23872 (make_pass_resolve_sw_modes): Likewise.
23874 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23876 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23879 2018-01-08 Richard Biener <rguenther@suse.de>
23881 PR middle-end/83580
23882 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23884 2018-01-08 Richard Biener <rguenther@suse.de>
23886 PR middle-end/83517
23887 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23889 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23891 PR middle-end/81897
23892 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23893 basic blocks with a small number of successors.
23894 (convert_control_dep_chain_into_preds): Improve handling of
23896 (dump_predicates): Split apart into...
23897 (dump_pred_chain): ...here...
23898 (dump_pred_info): ...and here.
23899 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23900 (can_chain_union_be_invalidated_p): Improve check for invalidation
23902 (uninit_uses_cannot_happen): Avoid unnecessary if
23903 convert_control_dep_chain_into_preds yielded nothing.
23905 2018-01-06 Martin Sebor <msebor@redhat.com>
23907 PR tree-optimization/83640
23908 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23909 subtracting negative offset from size.
23910 (builtin_access::overlap): Adjust offset bounds of the access to fall
23911 within the size of the object if possible.
23913 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23915 PR rtl-optimization/83699
23916 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23917 extract_bit_field_as_subreg to cases in which the extracted
23918 value is also a vector.
23920 * lra-constraints.c (process_alt_operands): Test for the equivalence
23921 substitutions when detecting a possible reload cycle.
23923 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23926 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23927 by default if flag_selective_schedling{,2}. Formatting fixes.
23929 PR rtl-optimization/83682
23930 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23931 if it has non-VECTOR_MODE element mode.
23932 (vec_duplicate_p): Likewise.
23934 PR middle-end/83694
23935 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23936 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23938 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23941 * config/i386/i386-builtin.def
23942 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23943 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23944 Require also OPTION_MASK_ISA_AVX512F in addition to
23945 OPTION_MASK_ISA_GFNI.
23946 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23947 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23948 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23949 to OPTION_MASK_ISA_GFNI.
23950 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23951 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23952 OPTION_MASK_ISA_AVX512BW.
23953 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23954 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23955 addition to OPTION_MASK_ISA_GFNI.
23956 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23957 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23958 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23959 to OPTION_MASK_ISA_GFNI.
23960 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23961 a requirement for all ISAs rather than any of them with a few
23963 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23965 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23966 bitmasks to be enabled with 3 exceptions, instead of requiring any
23967 enabled ISA with lots of exceptions.
23968 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23969 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23970 Change avx512bw in isa attribute to avx512f.
23971 * config/i386/sgxintrin.h: Add license boilerplate.
23972 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23973 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23974 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23975 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23977 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23978 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23979 temporarily sse2 rather than sse if not enabled already.
23982 * config/i386/sse.md (VI248_VLBW): Rename to ...
23983 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23984 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23985 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23986 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23987 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23988 mode iterator instead of VI248_VLBW.
23990 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23992 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23993 (record_modified): Skip clobbers; add debug output.
23994 (param_change_prob): Use sreal frequencies.
23996 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23998 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23999 punt for user-aligned variables.
24001 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24003 * tree-chrec.c (chrec_contains_symbols): Return true for
24006 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24009 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24010 of (x|y) == x for BICS pattern.
24012 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24014 PR tree-optimization/83605
24015 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24016 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24019 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24021 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24022 * config/epiphany/rtems.h: New file.
24024 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24025 Uros Bizjak <ubizjak@gmail.com>
24028 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24029 QIreg_operand instead of register_operand predicate.
24030 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24031 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24032 comments instead of -fmitigate[-_]rop.
24034 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24037 * cgraphunit.c (symbol_table::compile): Switch to text_section
24038 before calling assembly_start debug hook.
24039 * run-rtl-passes.c (run_rtl_passes): Likewise.
24042 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24044 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24045 range_int_cst_p rather than !symbolic_range_p before calling
24046 extract_range_from_multiplicative_op_1.
24048 2018-01-04 Jeff Law <law@redhat.com>
24050 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24051 redundant test in assertion.
24053 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24055 * doc/rtl.texi: Document machine_mode wrapper classes.
24057 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24059 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24060 using tree_to_uhwi.
24062 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24064 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24065 the VEC_PERM_EXPR fold to fail.
24067 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24070 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24071 to switched_sections.
24073 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24076 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24077 test for d.testing.
24079 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24082 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24083 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24085 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24088 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24089 is BLKmode and bitpos not zero or mode change is needed.
24091 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24094 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24097 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24100 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24101 instead of MULT rtx. Update all corresponding splitters.
24102 (*saddl_se): Ditto.
24103 (*ssub<modesuffix>): Ditto.
24104 (*ssubl_se): Ditto.
24105 (*cmp_sadd_di): Update split patterns.
24106 (*cmp_sadd_si): Ditto.
24107 (*cmp_sadd_sidi): Ditto.
24108 (*cmp_ssub_di): Ditto.
24109 (*cmp_ssub_si): Ditto.
24110 (*cmp_ssub_sidi): Ditto.
24111 * config/alpha/predicates.md (const23_operand): New predicate.
24112 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24113 Look for ASHIFT, not MULT inner operand.
24114 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24116 2018-01-04 Martin Liska <mliska@suse.cz>
24118 PR gcov-profile/83669
24119 * gcov.c (output_intermediate_file): Add version to intermediate
24121 * doc/gcov.texi: Document new field 'version' in intermediate
24122 file format. Fix location of '-k' option of gcov command.
24124 2018-01-04 Martin Liska <mliska@suse.cz>
24127 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24129 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24131 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24133 2018-01-03 Martin Sebor <msebor@redhat.com>
24135 PR tree-optimization/83655
24136 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24137 checking calls with invalid arguments.
24139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24141 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24142 (vectorizable_mask_load_store): Delete.
24143 (vectorizable_call): Return false for masked loads and stores.
24144 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24145 instead of gimple_assign_rhs1.
24146 (vectorizable_load): Handle IFN_MASK_LOAD.
24147 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24151 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24153 (vectorizable_mask_load_store): ...here.
24154 (vectorizable_load): ...and here.
24156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24158 * tree-vect-stmts.c (vect_build_all_ones_mask)
24159 (vect_build_zero_merge_argument): New functions, split out from...
24160 (vectorizable_load): ...here.
24162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24164 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24166 (vectorizable_mask_load_store): ...here.
24167 (vectorizable_store): ...and here.
24169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24171 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24173 (vectorizable_mask_load_store): ...here.
24175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24177 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24178 (vect_model_store_cost): Take a vec_load_store_type instead of a
24180 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24181 (vect_model_store_cost): Take a vec_load_store_type instead of a
24183 (vectorizable_mask_load_store): Update accordingly.
24184 (vectorizable_store): Likewise.
24185 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24189 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24190 IFN_MASK_LOAD calls here rather than...
24191 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24194 Alan Hayward <alan.hayward@arm.com>
24195 David Sherwood <david.sherwood@arm.com>
24197 * expmed.c (extract_bit_field_1): For vector extracts,
24198 fall back to extract_bit_field_as_subreg if vec_extract
24201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24202 Alan Hayward <alan.hayward@arm.com>
24203 David Sherwood <david.sherwood@arm.com>
24205 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24206 they are variable or constant sized.
24207 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24208 slots for constant-sized data.
24210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24211 Alan Hayward <alan.hayward@arm.com>
24212 David Sherwood <david.sherwood@arm.com>
24214 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24215 handling COND_EXPRs with boolean comparisons, try to find a better
24216 basis for the mask type than the boolean itself.
24218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24220 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24221 is calculated and how it can be overridden.
24222 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24223 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24225 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24229 Alan Hayward <alan.hayward@arm.com>
24230 David Sherwood <david.sherwood@arm.com>
24232 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24233 Remove the mode argument.
24234 (aarch64_simd_valid_immediate): Remove the mode and inverse
24236 * config/aarch64/iterators.md (bitsize): New iterator.
24237 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24238 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24239 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24240 aarch64_simd_valid_immediate.
24241 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24242 (aarch64_reg_or_bic_imm): Likewise.
24243 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24244 with an insn_type enum and msl with a modifier_type enum.
24245 Replace element_width with a scalar_mode. Change the shift
24246 to unsigned int. Add constructors for scalar_float_mode and
24247 scalar_int_mode elements.
24248 (aarch64_vect_float_const_representable_p): Delete.
24249 (aarch64_can_const_movi_rtx_p)
24250 (aarch64_simd_scalar_immediate_valid_for_move)
24251 (aarch64_simd_make_constant): Update call to
24252 aarch64_simd_valid_immediate.
24253 (aarch64_advsimd_valid_immediate_hs): New function.
24254 (aarch64_advsimd_valid_immediate): Likewise.
24255 (aarch64_simd_valid_immediate): Remove mode and inverse
24256 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24257 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24258 and aarch64_float_const_representable_p on the result.
24259 (aarch64_output_simd_mov_immediate): Remove mode argument.
24260 Update call to aarch64_simd_valid_immediate and use of
24261 simd_immediate_info.
24262 (aarch64_output_scalar_simd_mov_immediate): Update call
24265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24266 Alan Hayward <alan.hayward@arm.com>
24267 David Sherwood <david.sherwood@arm.com>
24269 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24270 (mode_nunits): Likewise CONST_MODE_NUNITS.
24271 * machmode.def (ADJUST_NUNITS): Document.
24272 * genmodes.c (mode_data::need_nunits_adj): New field.
24273 (blank_mode): Update accordingly.
24274 (adj_nunits): New variable.
24275 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24277 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24278 listed in adj_nunits.
24279 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24280 listed in adj_nunits. Don't emit case statements for such modes.
24281 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24282 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24283 nothing if adj_nunits is nonnull.
24284 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24285 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24286 (emit_mode_fbit): Update use of print_maybe_const_decl.
24287 (emit_move_size): Likewise. Treat the array as non-const
24289 (emit_mode_adjustments): Handle adj_nunits.
24291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24293 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24294 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24295 (VECTOR_MODES): Use it.
24296 (make_vector_modes): Take the prefix as an argument.
24298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24299 Alan Hayward <alan.hayward@arm.com>
24300 David Sherwood <david.sherwood@arm.com>
24302 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24303 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24304 for MODE_VECTOR_BOOL.
24305 * machmode.def (VECTOR_BOOL_MODE): Document.
24306 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24307 (make_vector_bool_mode): New function.
24308 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24310 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24311 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24313 * stor-layout.c (int_mode_for_mode): Likewise.
24314 * tree.c (build_vector_type_for_mode): Likewise.
24315 * varasm.c (output_constant_pool_2): Likewise.
24316 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24317 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24318 for MODE_VECTOR_BOOL.
24319 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24320 of mode class checks.
24321 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24322 instead of a list of mode class checks.
24323 (expand_vector_scalar_condition): Likewise.
24324 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24327 Alan Hayward <alan.hayward@arm.com>
24328 David Sherwood <david.sherwood@arm.com>
24330 * machmode.h (mode_size): Change from unsigned short to
24332 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24333 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24334 or if measurement_type is not polynomial.
24335 (fixed_size_mode::includes_p): Check for constant-sized modes.
24336 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24337 return a poly_uint16 rather than an unsigned short.
24338 (emit_mode_size): Change the type of mode_size from unsigned short
24339 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24340 (emit_mode_adjustments): Cope with polynomial vector sizes.
24341 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24343 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24345 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24346 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24347 * caller-save.c (setup_save_areas): Likewise.
24348 (replace_reg_with_saved_mem): Likewise.
24349 * calls.c (emit_library_call_value_1): Likewise.
24350 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24351 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24352 (gen_lowpart_for_combine): Likewise.
24353 * convert.c (convert_to_integer_1): Likewise.
24354 * cse.c (equiv_constant, cse_insn): Likewise.
24355 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24356 (cselib_subst_to_values): Likewise.
24357 * dce.c (word_dce_process_block): Likewise.
24358 * df-problems.c (df_word_lr_mark_ref): Likewise.
24359 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24360 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24361 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24362 (rtl_for_decl_location): Likewise.
24363 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24364 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24365 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24366 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24367 (expand_expr_real_1): Likewise.
24368 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24369 (pad_below): Likewise.
24370 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24371 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24372 * ira.c (get_subreg_tracking_sizes): Likewise.
24373 * ira-build.c (ira_create_allocno_objects): Likewise.
24374 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24375 (ira_sort_regnos_for_alter_reg): Likewise.
24376 * ira-costs.c (record_operand_costs): Likewise.
24377 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24378 (resolve_simple_move): Likewise.
24379 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24380 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24381 (lra_constraints): Likewise.
24382 (CONST_POOL_OK_P): Reject variable-sized modes.
24383 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24384 (add_pseudo_to_slot, lra_spill): Likewise.
24385 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24386 * optabs-query.c (get_best_extraction_insn): Likewise.
24387 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24388 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24389 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24390 * recog.c (offsettable_address_addr_space_p): Likewise.
24391 * regcprop.c (maybe_mode_change): Likewise.
24392 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24393 * regrename.c (build_def_use): Likewise.
24394 * regstat.c (dump_reg_info): Likewise.
24395 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24396 (find_reloads, find_reloads_subreg_address): Likewise.
24397 * reload1.c (eliminate_regs_1): Likewise.
24398 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24399 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24400 (simplify_binary_operation_1, simplify_subreg): Likewise.
24401 * targhooks.c (default_function_arg_padding): Likewise.
24402 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24403 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24404 (verify_gimple_assign_ternary): Likewise.
24405 * tree-inline.c (estimate_move_cost): Likewise.
24406 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24407 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24408 (get_address_cost_ainc): Likewise.
24409 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24410 (vect_supportable_dr_alignment): Likewise.
24411 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24412 (vectorizable_reduction): Likewise.
24413 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24414 (vectorizable_operation, vectorizable_load): Likewise.
24415 * tree.c (build_same_sized_truth_vector_type): Likewise.
24416 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24417 * var-tracking.c (emit_note_insn_var_location): Likewise.
24418 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24419 (ADDR_VEC_ALIGN): Likewise.
24421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24422 Alan Hayward <alan.hayward@arm.com>
24423 David Sherwood <david.sherwood@arm.com>
24425 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24427 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24428 or if measurement_type is polynomial.
24429 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24430 * combine.c (make_extraction): Likewise.
24431 * dse.c (find_shift_sequence): Likewise.
24432 * dwarf2out.c (mem_loc_descriptor): Likewise.
24433 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24434 (extract_bit_field, extract_low_bits): Likewise.
24435 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24436 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24437 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24438 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24439 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24440 * reload.c (find_reloads): Likewise.
24441 * reload1.c (alter_reg): Likewise.
24442 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24443 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24444 * tree-if-conv.c (predicate_mem_writes): Likewise.
24445 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24446 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24447 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24448 * valtrack.c (dead_debug_insert_temp): Likewise.
24449 * varasm.c (mergeable_constant_section): Likewise.
24450 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24453 Alan Hayward <alan.hayward@arm.com>
24454 David Sherwood <david.sherwood@arm.com>
24456 * expr.c (expand_assignment): Cope with polynomial mode sizes
24457 when assigning to a CONCAT.
24459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24460 Alan Hayward <alan.hayward@arm.com>
24461 David Sherwood <david.sherwood@arm.com>
24463 * machmode.h (mode_precision): Change from unsigned short to
24465 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24467 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24468 or if measurement_type is not polynomial.
24469 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24470 in which the mode is already known to be a scalar_int_mode.
24471 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24472 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24474 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24475 for GET_MODE_PRECISION.
24476 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24477 for GET_MODE_PRECISION.
24478 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24480 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24481 (expand_field_assignment, make_extraction): Likewise.
24482 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24483 (get_last_value): Likewise.
24484 * convert.c (convert_to_integer_1): Likewise.
24485 * cse.c (cse_insn): Likewise.
24486 * expr.c (expand_expr_real_1): Likewise.
24487 * lra-constraints.c (simplify_operand_subreg): Likewise.
24488 * optabs-query.c (can_atomic_load_p): Likewise.
24489 * optabs.c (expand_atomic_load): Likewise.
24490 (expand_atomic_store): Likewise.
24491 * ree.c (combine_reaching_defs): Likewise.
24492 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24493 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24494 * tree.h (type_has_mode_precision_p): Likewise.
24495 * ubsan.c (instrument_si_overflow): Likewise.
24497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24498 Alan Hayward <alan.hayward@arm.com>
24499 David Sherwood <david.sherwood@arm.com>
24501 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24502 polynomial numbers of units.
24503 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24504 (valid_vector_subparts_p): New function.
24505 (build_vector_type): Remove temporary shim and take the number
24506 of units as a poly_uint64 rather than an int.
24507 (build_opaque_vector_type): Take the number of units as a
24508 poly_uint64 rather than an int.
24509 * tree.c (build_vector_from_ctor): Handle polynomial
24510 TYPE_VECTOR_SUBPARTS.
24511 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24512 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24513 (build_vector_from_val): If the number of units is variable,
24514 use build_vec_duplicate_cst for constant operands and
24515 VEC_DUPLICATE_EXPR otherwise.
24516 (make_vector_type): Remove temporary is_constant ().
24517 (build_vector_type, build_opaque_vector_type): Take the number of
24518 units as a poly_uint64 rather than an int.
24519 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24521 * cfgexpand.c (expand_debug_expr): Likewise.
24522 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24523 (store_constructor, expand_expr_real_1): Likewise.
24524 (const_scalar_mask_from_tree): Likewise.
24525 * fold-const-call.c (fold_const_reduction): Likewise.
24526 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24527 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24528 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24529 (fold_relational_const): Likewise.
24530 (native_interpret_vector): Likewise. Change the size from an
24531 int to an unsigned int.
24532 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24533 TYPE_VECTOR_SUBPARTS.
24534 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24535 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24536 duplicating a non-constant operand into a variable-length vector.
24537 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24538 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24539 * ipa-icf.c (sem_variable::equals): Likewise.
24540 * match.pd: Likewise.
24541 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24542 * print-tree.c (print_node): Likewise.
24543 * stor-layout.c (layout_type): Likewise.
24544 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24545 * tree-cfg.c (verify_gimple_comparison): Likewise.
24546 (verify_gimple_assign_binary): Likewise.
24547 (verify_gimple_assign_ternary): Likewise.
24548 (verify_gimple_assign_single): Likewise.
24549 * tree-pretty-print.c (dump_generic_node): Likewise.
24550 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24551 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24552 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24553 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24554 (vect_shift_permute_load_chain): Likewise.
24555 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24556 (expand_vector_condition, optimize_vector_constructor): Likewise.
24557 (lower_vec_perm, get_compute_type): Likewise.
24558 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24559 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24560 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24561 (vect_recog_mask_conversion_pattern): Likewise.
24562 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24563 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24564 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24565 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24566 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24567 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24568 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24569 (supportable_widening_operation): Likewise.
24570 (supportable_narrowing_operation): Likewise.
24571 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24573 * varasm.c (output_constant): Likewise.
24575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24576 Alan Hayward <alan.hayward@arm.com>
24577 David Sherwood <david.sherwood@arm.com>
24579 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24580 so that both the length == 3 and length != 3 cases set up their
24581 own permute vectors. Add comments explaining why we know the
24582 number of elements is constant.
24583 (vect_permute_load_chain): Likewise.
24585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24586 Alan Hayward <alan.hayward@arm.com>
24587 David Sherwood <david.sherwood@arm.com>
24589 * machmode.h (mode_nunits): Change from unsigned char to
24591 (ONLY_FIXED_SIZE_MODES): New macro.
24592 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24593 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24594 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24596 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24597 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24598 or if measurement_type is not polynomial.
24599 * genmodes.c (ZERO_COEFFS): New macro.
24600 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24602 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24603 Use ZERO_COEFFS when emitting initializers.
24604 * data-streamer.h (bp_pack_poly_value): New function.
24605 (bp_unpack_poly_value): Likewise.
24606 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24607 for GET_MODE_NUNITS.
24608 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24609 for GET_MODE_NUNITS.
24610 * tree.c (make_vector_type): Remove temporary shim and make
24611 the real function take the number of units as a poly_uint64
24612 rather than an int.
24613 (build_vector_type_for_mode): Handle polynomial nunits.
24614 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24615 * emit-rtl.c (const_vec_series_p_1): Likewise.
24616 (gen_rtx_CONST_VECTOR): Likewise.
24617 * fold-const.c (test_vec_duplicate_folding): Likewise.
24618 * genrecog.c (validate_pattern): Likewise.
24619 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24620 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24621 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24622 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24623 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24624 * rtlanal.c (subreg_get_info): Likewise.
24625 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24626 (vect_grouped_load_supported): Likewise.
24627 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24628 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24629 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24630 (simplify_const_unary_operation, simplify_binary_operation_1)
24631 (simplify_const_binary_operation, simplify_ternary_operation)
24632 (test_vector_ops_duplicate, test_vector_ops): Likewise.
24633 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24634 instead of CONST_VECTOR_NUNITS.
24635 * varasm.c (output_constant_pool_2): Likewise.
24636 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24637 explicit-encoded elements in the XVEC for variable-length vectors.
24639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24641 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24644 Alan Hayward <alan.hayward@arm.com>
24645 David Sherwood <david.sherwood@arm.com>
24647 * coretypes.h (fixed_size_mode): Declare.
24648 (fixed_size_mode_pod): New typedef.
24649 * builtins.h (target_builtins::x_apply_args_mode)
24650 (target_builtins::x_apply_result_mode): Change type to
24651 fixed_size_mode_pod.
24652 * builtins.c (apply_args_size, apply_result_size, result_vector)
24653 (expand_builtin_apply_args_1, expand_builtin_apply)
24654 (expand_builtin_return): Update accordingly.
24656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24658 * cse.c (hash_rtx_cb): Hash only the encoded elements.
24659 * cselib.c (cselib_hash_rtx): Likewise.
24660 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24661 CONST_VECTOR encoding.
24663 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24664 Jeff Law <law@redhat.com>
24667 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
24668 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
24669 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
24670 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
24673 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
24674 explicitly probe *sp in a noreturn function if there were any callee
24675 register saves or frame pointer is needed.
24677 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24680 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24681 BLKmode for ternary, binary or unary expressions.
24684 * var-tracking.c (delete_vta_debug_insn): New inline function.
24685 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
24686 insns from get_insns () to NULL instead of each bb separately.
24687 Use delete_vta_debug_insn. No longer static.
24688 (vt_debug_insns_local, variable_tracking_main_1): Adjust
24689 delete_vta_debug_insns callers.
24690 * rtl.h (delete_vta_debug_insns): Declare.
24691 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
24692 instead of variable_tracking_main.
24694 2018-01-03 Martin Sebor <msebor@redhat.com>
24696 PR tree-optimization/83603
24697 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
24698 arguments past the endof the argument list in functions declared
24699 without a prototype.
24700 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
24701 Avoid checking when arguments are null.
24703 2018-01-03 Martin Sebor <msebor@redhat.com>
24706 * doc/extend.texi (attribute const): Fix a typo.
24707 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
24708 issuing -Wsuggest-attribute for void functions.
24710 2018-01-03 Martin Sebor <msebor@redhat.com>
24712 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
24713 offset_int::from instead of wide_int::to_shwi.
24714 (maybe_diag_overlap): Remove assertion.
24715 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
24716 * gimple-ssa-sprintf.c (format_directive): Same.
24717 (parse_directive): Same.
24718 (sprintf_dom_walker::compute_format_length): Same.
24719 (try_substitute_return_value): Same.
24721 2018-01-03 Jeff Law <law@redhat.com>
24723 PR middle-end/83654
24724 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
24725 non-constant residual for zero at runtime and avoid probing in
24726 that case. Reorganize code for trailing problem to mirror handling
24729 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24731 PR tree-optimization/83501
24732 * tree-ssa-strlen.c (get_string_cst): New.
24733 (handle_char_store): Call get_string_cst.
24735 2018-01-03 Martin Liska <mliska@suse.cz>
24737 PR tree-optimization/83593
24738 * tree-ssa-strlen.c: Include tree-cfg.h.
24739 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
24740 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
24741 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
24743 (strlen_dom_walker::before_dom_children): Call
24744 gimple_purge_dead_eh_edges. Dump tranformation with details
24746 (strlen_dom_walker::before_dom_children): Update call by adding
24747 new argument cleanup_eh.
24748 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
24750 2018-01-03 Martin Liska <mliska@suse.cz>
24753 * cif-code.def (VARIADIC_THUNK): New enum value.
24754 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
24757 2018-01-03 Jan Beulich <jbeulich@suse.com>
24759 * sse.md (mov<mode>_internal): Tighten condition for when to use
24760 vmovdqu<ssescalarsize> for TI and OI modes.
24762 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24764 Update copyright years.
24766 2018-01-03 Martin Liska <mliska@suse.cz>
24769 * ipa-visibility.c (function_and_variable_visibility): Skip
24770 functions with noipa attribure.
24772 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24774 * gcc.c (process_command): Update copyright notice dates.
24775 * gcov-dump.c (print_version): Ditto.
24776 * gcov.c (print_version): Ditto.
24777 * gcov-tool.c (print_version): Ditto.
24778 * gengtype.c (create_file): Ditto.
24779 * doc/cpp.texi: Bump @copying's copyright year.
24780 * doc/cppinternals.texi: Ditto.
24781 * doc/gcc.texi: Ditto.
24782 * doc/gccint.texi: Ditto.
24783 * doc/gcov.texi: Ditto.
24784 * doc/install.texi: Ditto.
24785 * doc/invoke.texi: Ditto.
24787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24789 * vector-builder.h (vector_builder::m_full_nelts): Change from
24790 unsigned int to poly_uint64.
24791 (vector_builder::full_nelts): Update prototype accordingly.
24792 (vector_builder::new_vector): Likewise.
24793 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24794 (vector_builder::operator ==): Likewise.
24795 (vector_builder::finalize): Likewise.
24796 * int-vector-builder.h (int_vector_builder::int_vector_builder):
24797 Take the number of elements as a poly_uint64 rather than an
24799 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24800 from unsigned int to poly_uint64.
24801 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24802 (vec_perm_indices::new_vector): Likewise.
24803 (vec_perm_indices::length): Likewise.
24804 (vec_perm_indices::nelts_per_input): Likewise.
24805 (vec_perm_indices::input_nelts): Likewise.
24806 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24807 number of elements per input as a poly_uint64 rather than an
24808 unsigned int. Use the original encoding for variable-length
24809 vectors, rather than clamping each individual element.
24810 For the second and subsequent elements in each pattern,
24811 clamp the step and base before clamping their sum.
24812 (vec_perm_indices::series_p): Handle polynomial element counts.
24813 (vec_perm_indices::all_in_range_p): Likewise.
24814 (vec_perm_indices_to_tree): Likewise.
24815 (vec_perm_indices_to_rtx): Likewise.
24816 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24817 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24818 (tree_vector_builder::new_binary_operation): Handle polynomial
24819 element counts. Return false if we need to know the number
24820 of elements at compile time.
24821 * fold-const.c (fold_vec_perm): Punt if the number of elements
24822 isn't known at compile time.
24824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24826 * vec-perm-indices.h (vec_perm_builder): Change element type
24827 from HOST_WIDE_INT to poly_int64.
24828 (vec_perm_indices::element_type): Update accordingly.
24829 (vec_perm_indices::clamp): Handle polynomial element_types.
24830 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24831 (vec_perm_indices::all_in_range_p): Likewise.
24832 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24834 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24835 polynomial vec_perm_indices element types.
24836 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24837 * fold-const.c (fold_vec_perm): Likewise.
24838 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24839 * tree-vect-generic.c (lower_vec_perm): Likewise.
24840 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24841 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24842 element type to HOST_WIDE_INT.
24844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24845 Alan Hayward <alan.hayward@arm.com>
24846 David Sherwood <david.sherwood@arm.com>
24848 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24849 rather than an int. Use plus_constant.
24850 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24851 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24854 Alan Hayward <alan.hayward@arm.com>
24855 David Sherwood <david.sherwood@arm.com>
24857 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24858 a HOST_WIDE_INT to a poly_int64.
24860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24861 Alan Hayward <alan.hayward@arm.com>
24862 David Sherwood <david.sherwood@arm.com>
24864 * calls.c (load_register_parameters): Cope with polynomial
24865 mode sizes. Require a constant size for BLKmode parameters
24866 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24867 forces a parameter to be padded at the lsb end in order to
24868 fill a complete number of words, require the parameter size
24869 to be ordered wrt UNITS_PER_WORD.
24871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24872 Alan Hayward <alan.hayward@arm.com>
24873 David Sherwood <david.sherwood@arm.com>
24875 * reload1.c (spill_stack_slot_width): Change element type
24876 from unsigned int to poly_uint64_pod.
24877 (alter_reg): Treat mode sizes as polynomial.
24879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24880 Alan Hayward <alan.hayward@arm.com>
24881 David Sherwood <david.sherwood@arm.com>
24883 * reload.c (complex_word_subreg_p): New function.
24884 (reload_inner_reg_of_subreg, push_reload): Use it.
24886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24887 Alan Hayward <alan.hayward@arm.com>
24888 David Sherwood <david.sherwood@arm.com>
24890 * lra-constraints.c (process_alt_operands): Reject matched
24891 operands whose sizes aren't ordered.
24892 (match_reload): Refer to this check here.
24894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24895 Alan Hayward <alan.hayward@arm.com>
24896 David Sherwood <david.sherwood@arm.com>
24898 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24899 that the mode size is in the set {1, 2, 4, 8, 16}.
24901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24902 Alan Hayward <alan.hayward@arm.com>
24903 David Sherwood <david.sherwood@arm.com>
24905 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24906 Use plus_constant instead of gen_rtx_PLUS.
24908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24909 Alan Hayward <alan.hayward@arm.com>
24910 David Sherwood <david.sherwood@arm.com>
24912 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24913 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24914 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24915 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24916 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24917 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24918 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24919 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24920 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24921 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24923 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24924 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24925 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24926 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24927 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24928 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24929 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24930 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24931 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24932 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24934 * expr.c (emit_move_resolve_push): Treat the input and result
24935 of PUSH_ROUNDING as a poly_int64.
24936 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24937 (emit_push_insn): Likewise.
24938 * lra-eliminations.c (mark_not_eliminable): Likewise.
24939 * recog.c (push_operand): Likewise.
24940 * reload1.c (elimination_effects): Likewise.
24941 * rtlanal.c (nonzero_bits1): Likewise.
24942 * calls.c (store_one_arg): Likewise. Require the padding to be
24943 known at compile time.
24945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24946 Alan Hayward <alan.hayward@arm.com>
24947 David Sherwood <david.sherwood@arm.com>
24949 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24950 Use plus_constant instead of gen_rtx_PLUS.
24952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24953 Alan Hayward <alan.hayward@arm.com>
24954 David Sherwood <david.sherwood@arm.com>
24956 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24957 rather than an int.
24959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24960 Alan Hayward <alan.hayward@arm.com>
24961 David Sherwood <david.sherwood@arm.com>
24963 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24964 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24965 via stack temporaries. Treat the mode size as polynomial too.
24967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24968 Alan Hayward <alan.hayward@arm.com>
24969 David Sherwood <david.sherwood@arm.com>
24971 * expr.c (expand_expr_real_2): When handling conversions involving
24972 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24973 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24974 as a poly_uint64 too.
24976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24977 Alan Hayward <alan.hayward@arm.com>
24978 David Sherwood <david.sherwood@arm.com>
24980 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24983 Alan Hayward <alan.hayward@arm.com>
24984 David Sherwood <david.sherwood@arm.com>
24986 * combine.c (can_change_dest_mode): Handle polynomial
24987 REGMODE_NATURAL_SIZE.
24988 * expmed.c (store_bit_field_1): Likewise.
24989 * expr.c (store_constructor): Likewise.
24990 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24991 and polynomial REGMODE_NATURAL_SIZE.
24992 (gen_lowpart_common): Likewise.
24993 * reginfo.c (record_subregs_of_mode): Likewise.
24994 * rtlanal.c (read_modify_subreg_p): Likewise.
24996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24997 Alan Hayward <alan.hayward@arm.com>
24998 David Sherwood <david.sherwood@arm.com>
25000 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25001 numbers of elements.
25003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25004 Alan Hayward <alan.hayward@arm.com>
25005 David Sherwood <david.sherwood@arm.com>
25007 * match.pd: Cope with polynomial numbers of vector elements.
25009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25010 Alan Hayward <alan.hayward@arm.com>
25011 David Sherwood <david.sherwood@arm.com>
25013 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25014 in a POINTER_PLUS_EXPR.
25016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25017 Alan Hayward <alan.hayward@arm.com>
25018 David Sherwood <david.sherwood@arm.com>
25020 * omp-simd-clone.c (simd_clone_subparts): New function.
25021 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25022 (ipa_simd_modify_function_body): Likewise.
25024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25025 Alan Hayward <alan.hayward@arm.com>
25026 David Sherwood <david.sherwood@arm.com>
25028 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25029 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25030 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25031 (expand_vector_condition, vector_element): Likewise.
25032 (subparts_gt): New function.
25033 (get_compute_type): Use subparts_gt.
25034 (count_type_subparts): Delete.
25035 (expand_vector_operations_1): Use subparts_gt instead of
25036 count_type_subparts.
25038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25039 Alan Hayward <alan.hayward@arm.com>
25040 David Sherwood <david.sherwood@arm.com>
25042 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25043 (vect_compile_time_alias): ...this new function. Do the calculation
25044 on poly_ints rather than trees.
25045 (vect_prune_runtime_alias_test_list): Update call accordingly.
25047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25048 Alan Hayward <alan.hayward@arm.com>
25049 David Sherwood <david.sherwood@arm.com>
25051 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25053 (vect_schedule_slp_instance): Likewise.
25055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25056 Alan Hayward <alan.hayward@arm.com>
25057 David Sherwood <david.sherwood@arm.com>
25059 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25060 constant and extern definitions for variable-length vectors.
25061 (vect_get_constant_vectors): Note that the number of units
25062 is known to be constant.
25064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25065 Alan Hayward <alan.hayward@arm.com>
25066 David Sherwood <david.sherwood@arm.com>
25068 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25069 of units as polynomial. Choose between WIDE and NARROW based
25072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25073 Alan Hayward <alan.hayward@arm.com>
25074 David Sherwood <david.sherwood@arm.com>
25076 * tree-vect-stmts.c (simd_clone_subparts): New function.
25077 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25080 Alan Hayward <alan.hayward@arm.com>
25081 David Sherwood <david.sherwood@arm.com>
25083 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25084 vectors as polynomial. Use build_index_vector for
25085 IFN_GOMP_SIMD_LANE.
25087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25088 Alan Hayward <alan.hayward@arm.com>
25089 David Sherwood <david.sherwood@arm.com>
25091 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25092 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25093 for variable-length vectors.
25094 (vectorizable_mask_load_store): Treat the number of units as
25095 polynomial, asserting that it is constant if the condition has
25096 already been enforced.
25097 (vectorizable_store, vectorizable_load): Likewise.
25099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25100 Alan Hayward <alan.hayward@arm.com>
25101 David Sherwood <david.sherwood@arm.com>
25103 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25104 of units as polynomial. Punt if we can't tell at compile time
25105 which vector contains the final result.
25107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25108 Alan Hayward <alan.hayward@arm.com>
25109 David Sherwood <david.sherwood@arm.com>
25111 * tree-vect-loop.c (vectorizable_induction): Treat the number
25112 of units as polynomial. Punt on SLP inductions. Use an integer
25113 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25114 cast of such a series for variable-length floating-point
25117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25118 Alan Hayward <alan.hayward@arm.com>
25119 David Sherwood <david.sherwood@arm.com>
25121 * tree.h (build_index_vector): Declare.
25122 * tree.c (build_index_vector): New function.
25123 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25124 of units as polynomial, forcibly converting it to a constant if
25125 vectorizable_reduction has already enforced the condition.
25126 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25127 to create a {1,2,3,...} vector.
25128 (vectorizable_reduction): Treat the number of units as polynomial.
25129 Choose vectype_in based on the largest scalar element size rather
25130 than the smallest number of units. Enforce the restrictions
25133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25134 Alan Hayward <alan.hayward@arm.com>
25135 David Sherwood <david.sherwood@arm.com>
25137 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25138 number of units as polynomial.
25140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25141 Alan Hayward <alan.hayward@arm.com>
25142 David Sherwood <david.sherwood@arm.com>
25144 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25145 * target.def (autovectorize_vector_sizes): Return the vector sizes
25146 by pointer, using vector_sizes rather than a bitmask.
25147 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25148 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25149 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25151 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25152 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25153 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25154 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25155 * omp-general.c (omp_max_vf): Likewise.
25156 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25157 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25158 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25159 * tree-vect-slp.c (vect_slp_bb): Likewise.
25160 * doc/tm.texi: Regenerate.
25161 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25163 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25164 the vector size as a poly_uint64 rather than an unsigned int.
25165 (current_vector_size): Change from an unsigned int to a poly_uint64.
25166 (get_vectype_for_scalar_type): Update accordingly.
25167 * tree.h (build_truth_vector_type): Take the size and number of
25168 units as a poly_uint64 rather than an unsigned int.
25169 (build_vector_type): Add a temporary overload that takes
25170 the number of units as a poly_uint64 rather than an unsigned int.
25171 * tree.c (make_vector_type): Likewise.
25172 (build_truth_vector_type): Take the number of units as a poly_uint64
25173 rather than an unsigned int.
25175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25176 Alan Hayward <alan.hayward@arm.com>
25177 David Sherwood <david.sherwood@arm.com>
25179 * target.def (get_mask_mode): Take the number of units and length
25180 as poly_uint64s rather than unsigned ints.
25181 * targhooks.h (default_get_mask_mode): Update accordingly.
25182 * targhooks.c (default_get_mask_mode): Likewise.
25183 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25184 * doc/tm.texi: Regenerate.
25186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25187 Alan Hayward <alan.hayward@arm.com>
25188 David Sherwood <david.sherwood@arm.com>
25190 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25191 * omp-general.c (omp_max_vf): Likewise.
25192 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25193 (expand_omp_simd): Handle polynomial safelen.
25194 * omp-low.c (omplow_simd_context): Add a default constructor.
25195 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25196 (lower_rec_simd_input_clauses): Update accordingly.
25197 (lower_rec_input_clauses): Likewise.
25199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25200 Alan Hayward <alan.hayward@arm.com>
25201 David Sherwood <david.sherwood@arm.com>
25203 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25204 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25205 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25206 (vect_analyze_slp_cost): Likewise.
25207 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25208 (vect_model_load_cost): Likewise.
25210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25211 Alan Hayward <alan.hayward@arm.com>
25212 David Sherwood <david.sherwood@arm.com>
25214 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25215 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25216 from an unsigned int * to a poly_uint64_pod *.
25217 (calculate_unrolling_factor): New function.
25218 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25221 Alan Hayward <alan.hayward@arm.com>
25222 David Sherwood <david.sherwood@arm.com>
25224 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25225 from an unsigned int to a poly_uint64.
25226 (_loop_vec_info::slp_unrolling_factor): Likewise.
25227 (_loop_vec_info::vectorization_factor): Change from an int
25229 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25230 (vect_get_num_vectors): New function.
25231 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25232 (vect_get_num_copies): Use vect_get_num_vectors.
25233 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25234 to an unsigned int *.
25235 (vect_analyze_data_refs): Change min_vf from an int * to a
25237 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25238 than an unsigned HOST_WIDE_INT.
25239 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25240 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25241 to an unsigned int *.
25242 (vect_analyze_data_ref_dependences): Likewise.
25243 (vect_compute_data_ref_alignment): Handle polynomial vf.
25244 (vect_enhance_data_refs_alignment): Likewise.
25245 (vect_prune_runtime_alias_test_list): Likewise.
25246 (vect_shift_permute_load_chain): Likewise.
25247 (vect_supportable_dr_alignment): Likewise.
25248 (dependence_distance_ge_vf): Take the vectorization factor as a
25249 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25250 (vect_analyze_data_refs): Change min_vf from an int * to a
25252 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25253 vfm1 as a poly_uint64 rather than an int. Make the same change
25254 for the returned bound_scalar.
25255 (vect_gen_vector_loop_niters): Handle polynomial vf.
25256 (vect_do_peeling): Likewise. Update call to
25257 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25258 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25260 * tree-vect-loop.c (vect_determine_vectorization_factor)
25261 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25262 (vect_get_known_peeling_cost): Likewise.
25263 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25264 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25265 (vect_transform_loop): Likewise. Use the lowest possible VF when
25266 updating the upper bounds of the loop.
25267 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25268 rather than an int.
25269 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25270 polynomial unroll factors.
25271 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25272 (vect_make_slp_decision): Likewise.
25273 (vect_supported_load_permutation_p): Likewise, and polynomial
25275 (vect_analyze_slp_cost): Handle polynomial vf.
25276 (vect_slp_analyze_node_operations): Likewise.
25277 (vect_slp_analyze_bb_1): Likewise.
25278 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25279 than an unsigned HOST_WIDE_INT.
25280 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25281 (vectorizable_load): Handle polynomial vf.
25282 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25284 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25287 Alan Hayward <alan.hayward@arm.com>
25288 David Sherwood <david.sherwood@arm.com>
25290 * match.pd: Handle bit operations involving three constants
25291 and try to fold one pair.
25293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25295 * tree-vect-loop-manip.c: Include gimple-fold.h.
25296 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25297 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25298 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25299 Add a path that uses a step of VF instead of 1, but disable it
25301 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25302 and niters_no_overflow parameters. Update calls to
25303 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25304 Create a new SSA name if the latter choses to use a ste other
25305 than zero, and return it via niters_vector_mult_vf_var.
25306 * tree-vect-loop.c (vect_transform_loop): Update calls to
25307 vect_do_peeling, vect_gen_vector_loop_niters and
25308 slpeel_make_loop_iterate_ntimes.
25309 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25310 (vect_gen_vector_loop_niters): Update declarations after above changes.
25312 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25314 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25315 128-bit round to integer instructions.
25316 (ceil<mode>2): Likewise.
25317 (btrunc<mode>2): Likewise.
25318 (round<mode>2): Likewise.
25320 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25322 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25323 unaligned VSX load/store on P8/P9.
25324 (expand_block_clear): Allow the use of unaligned VSX
25325 load/store on P8/P9.
25327 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25329 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25331 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25332 swap associated with both a load and a store.
25334 2018-01-02 Andrew Waterman <andrew@sifive.com>
25336 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25337 * config/riscv/riscv.md (clear_cache): Use it.
25339 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25341 * web.c: Remove out-of-date comment.
25343 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25345 * expr.c (fixup_args_size_notes): Check that any existing
25346 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25347 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25348 (emit_single_push_insn): ...here.
25350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25352 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25353 (const_vector_encoded_nelts): New function.
25354 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25355 (const_vector_int_elt, const_vector_elt): Declare.
25356 * emit-rtl.c (const_vector_int_elt_1): New function.
25357 (const_vector_elt): Likewise.
25358 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25359 of CONST_VECTOR_ELT.
25361 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25363 * expr.c: Include rtx-vector-builder.h.
25364 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25365 directly on the tree encoding.
25366 (const_vector_from_tree): Likewise.
25367 * optabs.c: Include rtx-vector-builder.h.
25368 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25369 sequence of "u" values.
25370 * vec-perm-indices.c: Include rtx-vector-builder.h.
25371 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25372 directly on the vec_perm_indices encoding.
25374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25376 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25377 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25378 * rtx-vector-builder.h: New file.
25379 * rtx-vector-builder.c: Likewise.
25380 * rtl.h (rtx_def::u2): Add a const_vector field.
25381 (CONST_VECTOR_NPATTERNS): New macro.
25382 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25383 (CONST_VECTOR_DUPLICATE_P): Likewise.
25384 (CONST_VECTOR_STEPPED_P): Likewise.
25385 (CONST_VECTOR_ENCODED_ELT): Likewise.
25386 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25387 (unwrap_const_vec_duplicate): Likewise.
25388 (const_vec_series_p): Check for a non-duplicated vector encoding.
25389 Say that the function only returns true for integer vectors.
25390 * emit-rtl.c: Include rtx-vector-builder.h.
25391 (gen_const_vec_duplicate_1): Delete.
25392 (gen_const_vector): Call gen_const_vec_duplicate instead of
25393 gen_const_vec_duplicate_1.
25394 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25395 (gen_const_vec_duplicate): Use rtx_vector_builder.
25396 (gen_const_vec_series): Likewise.
25397 (gen_rtx_CONST_VECTOR): Likewise.
25398 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25399 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25400 Build a new vector rather than modifying a CONST_VECTOR in-place.
25401 (handle_special_swappables): Update call accordingly.
25402 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25403 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25404 Build a new vector rather than modifying a CONST_VECTOR in-place.
25405 (handle_special_swappables): Update call accordingly.
25407 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25409 * simplify-rtx.c (simplify_const_binary_operation): Use
25410 CONST_VECTOR_ELT instead of XVECEXP.
25412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25414 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25415 the selector elements to be different from the data elements
25416 if the selector is a VECTOR_CST.
25417 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25418 ssizetype for the selector.
25420 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25422 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25423 before testing each element individually.
25424 * tree-vect-generic.c (lower_vec_perm): Likewise.
25426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25428 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25429 * selftest-run-tests.c (selftest::run_tests): Call it.
25430 * vector-builder.h (vector_builder::operator ==): New function.
25431 (vector_builder::operator !=): Likewise.
25432 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25433 (vec_perm_indices::all_from_input_p): New function.
25434 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25435 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25436 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25437 instead of reading the VECTOR_CST directly. Detect whether both
25438 vector inputs are the same before constructing the vec_perm_indices,
25439 and update the number of inputs argument accordingly. Use the
25440 utility functions added above. Only construct sel2 if we need to.
25442 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25444 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25445 the broadcast of the low byte.
25446 (expand_mult_highpart): Use an explicit encoding for the permutes.
25447 * optabs-query.c (can_mult_highpart_p): Likewise.
25448 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25449 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25450 (vectorizable_bswap): Likewise.
25451 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25452 explicit encoding for the power-of-2 permutes.
25453 (vect_permute_store_chain): Likewise.
25454 (vect_grouped_load_supported): Likewise.
25455 (vect_permute_load_chain): Likewise.
25457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25459 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25460 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25461 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25462 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25463 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25464 (vect_gen_perm_mask_any): Likewise.
25466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25468 * int-vector-builder.h: New file.
25469 * vec-perm-indices.h: Include int-vector-builder.h.
25470 (vec_perm_indices): Redefine as an int_vector_builder.
25471 (auto_vec_perm_indices): Delete.
25472 (vec_perm_builder): Redefine as a stand-alone class.
25473 (vec_perm_indices::vec_perm_indices): New function.
25474 (vec_perm_indices::clamp): Likewise.
25475 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25476 (vec_perm_indices::new_vector): New function.
25477 (vec_perm_indices::new_expanded_vector): Update for new
25478 vec_perm_indices class.
25479 (vec_perm_indices::rotate_inputs): New function.
25480 (vec_perm_indices::all_in_range_p): Operate directly on the
25481 encoded form, without computing elided elements.
25482 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25483 encoding. Update for new vec_perm_indices class.
25484 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25485 the given vec_perm_builder.
25486 (expand_vec_perm_var): Update vec_perm_builder constructor.
25487 (expand_mult_highpart): Use vec_perm_builder instead of
25488 auto_vec_perm_indices.
25489 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25490 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25491 or double series encoding as appropriate.
25492 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25493 vec_perm_indices instead of auto_vec_perm_indices.
25494 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25495 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25496 (vect_permute_store_chain): Likewise.
25497 (vect_grouped_load_supported): Likewise.
25498 (vect_permute_load_chain): Likewise.
25499 (vect_shift_permute_load_chain): Likewise.
25500 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25501 (vect_transform_slp_perm_load): Likewise.
25502 (vect_schedule_slp_instance): Likewise.
25503 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25504 (vectorizable_mask_load_store): Likewise.
25505 (vectorizable_bswap): Likewise.
25506 (vectorizable_store): Likewise.
25507 (vectorizable_load): Likewise.
25508 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25509 vec_perm_indices instead of auto_vec_perm_indices. Use
25510 tree_to_vec_perm_builder to read the vector from a tree.
25511 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25512 vec_perm_builder instead of a vec_perm_indices.
25513 (have_whole_vector_shift): Use vec_perm_builder and
25514 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25515 truncation to calc_vec_perm_mask_for_shift.
25516 (vect_create_epilog_for_reduction): Likewise.
25517 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25518 from auto_vec_perm_indices to vec_perm_indices.
25519 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25520 instead of changing individual elements.
25521 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25522 the vector in d.perm.
25523 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25524 from auto_vec_perm_indices to vec_perm_indices.
25525 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25526 instead of changing individual elements.
25527 (arm_vectorize_vec_perm_const): Use new_vector to install
25528 the vector in d.perm.
25529 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25530 Update vec_perm_builder constructor.
25531 (rs6000_expand_interleave): Likewise.
25532 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25533 (rs6000_expand_interleave): Likewise.
25535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25537 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25538 to qimode could truncate the indices.
25539 * optabs.c (expand_vec_perm_var): Likewise.
25541 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25543 * Makefile.in (OBJS): Add vec-perm-indices.o.
25544 * vec-perm-indices.h: New file.
25545 * vec-perm-indices.c: Likewise.
25546 * target.h (vec_perm_indices): Replace with a forward class
25548 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25549 * optabs.h: Include vec-perm-indices.h.
25550 (expand_vec_perm): Delete.
25551 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25552 (expand_vec_perm_const): Declare.
25553 * target.def (vec_perm_const_ok): Replace with...
25554 (vec_perm_const): ...this new hook.
25555 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25556 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25557 * doc/tm.texi: Regenerate.
25558 * optabs.def (vec_perm_const): Delete.
25559 * doc/md.texi (vec_perm_const): Likewise.
25560 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25561 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25562 expand_vec_perm for constant permutation vectors. Assert that
25563 the mode of variable permutation vectors is the integer equivalent
25564 of the mode that is being permuted.
25565 * optabs-query.h (selector_fits_mode_p): Declare.
25566 * optabs-query.c: Include vec-perm-indices.h.
25567 (selector_fits_mode_p): New function.
25568 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25569 is defined, instead of checking whether the vec_perm_const_optab
25570 exists. Use targetm.vectorize.vec_perm_const instead of
25571 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25572 fit in the vector mode before using a variable permute.
25573 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25574 vec_perm_indices instead of an rtx.
25575 (expand_vec_perm): Replace with...
25576 (expand_vec_perm_const): ...this new function. Take the selector
25577 as a vec_perm_indices rather than an rtx. Also take the mode of
25578 the selector. Update call to shift_amt_for_vec_perm_mask.
25579 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25580 Use vec_perm_indices::new_expanded_vector to expand the original
25581 selector into bytes. Check whether the indices fit in the vector
25582 mode before using a variable permute.
25583 (expand_vec_perm_var): Make global.
25584 (expand_mult_highpart): Use expand_vec_perm_const.
25585 * fold-const.c: Includes vec-perm-indices.h.
25586 * tree-ssa-forwprop.c: Likewise.
25587 * tree-vect-data-refs.c: Likewise.
25588 * tree-vect-generic.c: Likewise.
25589 * tree-vect-loop.c: Likewise.
25590 * tree-vect-slp.c: Likewise.
25591 * tree-vect-stmts.c: Likewise.
25592 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25594 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25595 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25596 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25597 (aarch64_vectorize_vec_perm_const): ...this new function.
25598 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25599 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25600 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25601 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25602 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25603 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25604 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25606 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25607 check for NEON modes.
25608 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25609 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25610 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25611 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25613 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25614 the old VEC_PERM_CONST conditions.
25615 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25616 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25617 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25618 (ia64_vectorize_vec_perm_const_ok): Merge into...
25619 (ia64_vectorize_vec_perm_const): ...this new function.
25620 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25621 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25622 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25623 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25624 * config/mips/mips.c (mips_expand_vec_perm_const)
25625 (mips_vectorize_vec_perm_const_ok): Merge into...
25626 (mips_vectorize_vec_perm_const): ...this new function.
25627 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25628 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25629 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25630 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25631 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25632 (rs6000_expand_vec_perm_const): Delete.
25633 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25635 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25636 (altivec_expand_vec_perm_const_le): Take each operand individually.
25637 Operate on constant selectors rather than rtxes.
25638 (altivec_expand_vec_perm_const): Likewise. Update call to
25639 altivec_expand_vec_perm_const_le.
25640 (rs6000_expand_vec_perm_const): Delete.
25641 (rs6000_vectorize_vec_perm_const_ok): Delete.
25642 (rs6000_vectorize_vec_perm_const): New function.
25643 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25644 an element count and rtx array.
25645 (rs6000_expand_extract_even): Update call accordingly.
25646 (rs6000_expand_interleave): Likewise.
25647 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25648 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25649 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25650 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25651 (rs6000_expand_vec_perm_const): Delete.
25652 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25653 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25654 (altivec_expand_vec_perm_const_le): Take each operand individually.
25655 Operate on constant selectors rather than rtxes.
25656 (altivec_expand_vec_perm_const): Likewise. Update call to
25657 altivec_expand_vec_perm_const_le.
25658 (rs6000_expand_vec_perm_const): Delete.
25659 (rs6000_vectorize_vec_perm_const_ok): Delete.
25660 (rs6000_vectorize_vec_perm_const): New function. Remove stray
25661 reference to the SPE evmerge intructions.
25662 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25663 an element count and rtx array.
25664 (rs6000_expand_extract_even): Update call accordingly.
25665 (rs6000_expand_interleave): Likewise.
25666 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25667 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
25669 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25671 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25673 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
25674 vector mode and that that mode matches the mode of the data
25676 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
25677 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
25678 directly using expand_vec_perm_1 when forcing selectors into
25680 (expand_vec_perm_var): New function, split out from expand_vec_perm.
25682 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25684 * optabs-query.h (can_vec_perm_p): Delete.
25685 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
25686 * optabs-query.c (can_vec_perm_p): Split into...
25687 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
25688 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
25689 particular selector is valid.
25690 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25691 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25692 (vect_grouped_load_supported): Likewise.
25693 (vect_shift_permute_load_chain): Likewise.
25694 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25695 (vect_transform_slp_perm_load): Likewise.
25696 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25697 (vectorizable_bswap): Likewise.
25698 (vect_gen_perm_mask_checked): Likewise.
25699 * fold-const.c (fold_ternary_loc): Likewise. Don't take
25700 implementations of variable permutation vectors into account
25701 when deciding which selector to use.
25702 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
25703 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
25704 with a false third argument.
25705 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
25706 to test whether the constant selector is valid and can_vec_perm_var_p
25707 to test whether a variable selector is valid.
25709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25711 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
25712 * optabs-query.c (can_vec_perm_p): Likewise.
25713 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
25714 instead of vec_perm_indices.
25715 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
25716 (vect_gen_perm_mask_checked): Likewise,
25717 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
25718 (vect_gen_perm_mask_checked): Likewise,
25720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25722 * optabs-query.h (qimode_for_vec_perm): Declare.
25723 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
25724 (qimode_for_vec_perm): ...this new function.
25725 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
25727 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25729 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
25730 does not have a conditional at the top.
25732 2018-01-02 Richard Biener <rguenther@suse.de>
25734 * ipa-inline.c (big_speedup_p): Fix expression.
25736 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25739 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25742 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25746 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
25747 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
25748 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
25749 cond_taken_branch_cost 3->4.
25751 2018-01-01 Jakub Jelinek <jakub@redhat.com>
25753 PR tree-optimization/83581
25754 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
25755 TODO_cleanup_cfg if any changes have been made.
25757 PR middle-end/83608
25758 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
25759 convert_modes if target mode has the right side, but different mode
25762 PR middle-end/83609
25763 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
25764 last argument when extracting from CONCAT. If either from_real or
25765 from_imag is NULL, use expansion through memory. If result is not
25766 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
25767 the parts directly to inner mode, if even that fails, use expansion
25770 PR middle-end/83623
25771 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
25772 check for bswap in mode rather than HImode and use that in expand_unop
25775 Copyright (C) 2018 Free Software Foundation, Inc.
25777 Copying and distribution of this file, with or without modification,
25778 are permitted in any medium without royalty provided the copyright
25779 notice and this notice are preserved.