1 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6 hard registers for the clobbered pseudo.
8 2018-06-27 Paul Koning <ni1d@arrl.net>
10 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11 mutually exclusive options.
12 * config/pdp11/constraints.md (h): New constraint.
13 (O): Update definition to match shift code generation.
15 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
17 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
19 (output_jump): Change arguments.
20 (pdp11_fixed_cc_regs): New function.
21 (pdp11_cc_mode): Ditto.
22 (pdp11_expand_shift): Ditto.
23 (pdp11_assemble_shift): Ditto.
24 (pdp11_small_shift): Ditto.
25 (pdp11_branch_cost): Remove.
26 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
28 (pdp11_register_move_cost): Update for CC registers.
29 (pdp11_rtx_costs): Add case for LSHIFTRT.
30 (pdp11_output_jump): Add CCNZ mode conditional branches.
31 (notice_update_cc_on_set): Remove.
32 (pdp11_cc_mode): New function.
33 (simple_memory_operand): Correct pre/post decrement case.
34 (no_side_effect_operand): New function.
35 (pdp11_regno_reg_class): Add CC_REGS class.
36 (pdp11_fixed_cc_regs): New function.
37 (pdp11_small_shift): New function.
38 (pdp11_expand_shift): New function to expand shift insns.
39 (pdp11_assemble_shift): New function to output shifts.
40 (pdp11_branch_cost): Remove.
41 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
42 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
44 (PTRDIFF_TYPE): Ditto.
45 (ADJUST_INSN_LENGTH): New macro.
46 (FIXED_REGISTERS): Add CC registers.
47 (CALL_USED_REGISTERS): Ditto.
49 (REG_CLASS_NAMES): Ditto.
50 (REG_CLASS_CONTENTS): Ditto.
51 (SELECT_CC_MODE): Use new function.
52 (TARGET_FLAGS_REGNUM): New macro.
53 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
54 (cc0_reg_rtx): Remove.
55 (CC_STATUS_MDEP): Remove.
56 (CC_STATUS_MDEFP_INIT): Remove.
58 (NOTICE_UPDATE_CC): Remove.
59 (REGISTER_NAMES): Add CC registers.
60 (BRANCH_COST): Change to constant 1.
61 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
63 * config/pdp11/pdp11.opt (mbcopy): Remove.
64 (mbcopy-builtin): Remove.
65 (mbranch-cheap): Remove.
66 (mbranch-expensive): Remove.
67 * config/pdp11/predicates.md (expand_shift_operand): Update to
68 match shift code generation.
69 (ccnz_operator): New predicate.
70 * doc/invoke.texi (PDP-11 Options): Remove deleted options
71 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
72 Remove non-existent option -mabshi, -mno-abshi. Document mutually
74 * doc/md.texi (PDP-11): Document new D and h constraints. Update
75 description of O constraint.
77 2018-06-27 Jeff Law <law@redhat.com>
78 Austin Law <austinklaw@gmail.com>
80 * config/v850/v850.md (addsi3_set_flags): New pattern.
81 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
82 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
83 (zero_extendhisi2_v850_set_flags): Likewise.
84 (zero_extendqisi2_v850_set_flags): Likewise.
85 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
86 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
87 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
89 * config/v850/v850-protos.h (notice_update_cc): Remove.
90 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
91 (v850_print_operand): Handle 'D' and "d".
92 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
93 Add handling of arithmetic/logical operations compared against zero.
94 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
95 Do not look at v850_compare_op, instead get mode from last argument.
96 (v850_gen_compare): Remove
97 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
98 after reload for prologue insns.
99 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
101 (construct_save_jarl): Likewise.
102 (TARGET_FLAGS_REGNUM): Define.
103 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
104 (NOTICE_UPDATE_CC): Remove.
105 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
106 than cc0. Conditionalize on reload_completed.
107 (cmpsi_insn, setfcc_insn): Likewise.
108 (tst1 splitter): Turn into define_and_split which sets the flags
110 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
111 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
112 (cstoresf4, cstoredf4): Clobber the flags.
113 (cmpsi, cmpsf, cmpdf): Remove expanders.
114 (setf_insn): Remove pattern.
115 (addsi3): Turn into define_and_split which clobbers the flags after
116 reload and a suitable pattern (addsi3_clobber_flags) for use after
118 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
119 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
120 (ashrsi3, ashrsi3_v850e2): Likewise.
121 (bins): Clobber the flags.
122 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
123 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
124 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
125 (call_value_internal_short, call_value_internal_long): Likewise.
126 (callt_save_interrupt, callt_return_interrupt): Likewise.
127 (save_interrupt, return_interrupt): Likewise.
128 (callt_save_all_interrupt, save_all_interrupt): Likewise.
129 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
130 (restore_all_interrupt, _restore_all_interrupt): Likewise.
131 (All FP comparisons): Only allow after reload has completed.
133 (divh, divhu): Tweak output template.
134 (branch_z_normal, branch_z_invert): Remove
135 (branch_nz_normal, branch_nz_invert): Likewise.
136 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
138 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
139 * config/v850/v850.c (notice_update_cc): Remove.
140 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
141 (CC_NO_CARRY): Likewise.
142 (NOTICE_UPDATE_CC): Define to nothing.
143 * config/v850/v850.md: Remove block comment on cc0 handling
144 Remove "cc" attribute from all patterns. Remove cc_status handling
145 from all patterns. Minor formatting fixes.
147 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
149 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
150 (cortex-a76.cortex-a55): Likewise.
151 * config/aarch64/aarch64-tune.md: Regenerate.
152 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
153 cortex-a76.cortex-a55.
155 2018-06-27 Jeff Law <law@redhat.com>
157 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
158 (MULTILIB_DIRNAMES): Similarly.
160 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
162 * gimple.h (gimple_return_retbnd): Delete.
163 (gimple_return_set_retbnd): Likewise.
164 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
165 gimple_return_set_retbnd.
166 * gimple-pretty-print.c (dump_gimple_return): Remove call to
167 gimple_return_retbnd and adjust.
168 * tree-inline.h (struct copy_body_data): Remove retbnd field.
169 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
170 Explicitly return NULL in a couple more cases. Move assertion
171 on debug statements and remove unreachable code.
172 (reset_debug_binding): Do not test id->retbnd.
173 (expand_call_inline): Do not set it.
175 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
177 * configure.ac: Add --disable-gcov option.
178 * configure: Regenerate.
179 * Makefile.in: Honour @enable_gcov@.
180 * doc/install.texi: Document --disable-gcov.
182 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
184 * config/arm/arm-cpus.in (cortex-a76): New entry.
185 (cortex-a76.cortex-a55): Likewise.
186 * config/arm/arm-tables.opt: Regenerate.
187 * config/arm/arm-tune.md: Likewise.
188 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
189 * doc/invoke.texi (ARM Options): Document cortex-a76 and
190 cortex-a76.cortex-a55.
192 2018-06-27 Tamar Christina <tamar.christina@arm.com>
195 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
197 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
199 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
201 (EPILOGUE_USES): Likewise.
203 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
205 * tree-inline.c (remap_location): New function extracted from...
206 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
207 (copy_phis_for_bb): ...here. Call remap_location.
208 (copy_cfg_body): Adjust call to copy_edges_for_bb.
210 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
212 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
213 unaligned vsx for 16B memset.
215 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
218 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
219 ieee128_float_type_node to long_double_type_node unless
220 TARGET_LONG_DOUBLE_128 is set.
222 2018-06-26 David Malcolm <dmalcolm@redhat.com>
224 * cfgloop.c (get_loop_location): Convert return type from
225 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
226 by implicit construction from rtx_insn *, and using
227 dump_user_location_t::from_function_decl for the fallback case.
228 * cfgloop.h (get_loop_location): Convert return type from
229 location_t to dump_user_location_t.
230 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
231 dump_printf_loc to pass in a dump_location_t rather than a
232 location_t, via the gimple stmt.
233 * coverage.c (get_coverage_counts): Update calls to
234 dump_printf_loc to pass in dump_location_t rather than a
236 * doc/optinfo.texi (Dump types): Convert example of
237 dump_printf_loc from taking "locus" to taking "insn". Update
238 description of the "_loc" calls to cover dump_location_t.
239 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
241 (dump_user_location_t::dump_user_location_t): New constructors,
242 from gimple *stmt and rtx_insn *.
243 (dump_user_location_t::from_function_decl): New function.
244 (dump_loc): Make static.
245 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
246 const dump_location_t &.
247 (dump_generic_expr_loc): Delete.
248 (dump_printf_loc): Convert param "loc" from location_t to
249 const dump_location_t &.
250 (selftest::test_impl_location): New function.
251 (selftest::dumpfile_c_tests): New function.
252 * dumpfile.h: Include "profile-count.h".
253 (class dump_user_location_t): New class.
254 (struct dump_impl_location_t): New struct.
255 (class dump_location_t): New class.
256 (dump_printf_loc): Convert 2nd param from source_location to
257 const dump_location_t &.
258 (dump_generic_expr_loc): Delete.
259 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
260 const dump_location_t &.
261 * gimple-fold.c (fold_gimple_assign): Update call to
262 dump_printf_loc to pass in a dump_location_t rather than a
263 location_t, via the gimple stmt.
264 (gimple_fold_call): Likewise.
265 * gimple-loop-interchange.cc
266 (loop_cand::analyze_iloop_reduction_var): Update for change to
267 check_reduction_path.
268 (tree_loop_interchange::interchange): Update for change to
270 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
271 change in return-type of find_loop_location.
272 (graphite_regenerate_ast_isl): Likewise.
273 * graphite-optimize-isl.c (optimize_isl): Likewise.
274 * graphite.c (graphite_transform_loops): Likewise.
275 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
276 pass in a dump_location_t rather than a location_t, via the
278 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
279 * ipa.c (walk_polymorphic_call_targets): Likewise.
280 * loop-unroll.c (report_unroll): Convert "locus" param from
281 location_t to dump_location_t.
282 (decide_unrolling): Update for change to get_loop_location's
284 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
285 location_t to dump_user_location_t.
286 (grid_find_single_omp_among_assignments_1): Updates calls to
287 dump_printf_loc to pass in a dump_location_t rather than a
288 location_t, via the gimple stmt.
289 (grid_parallel_clauses_gridifiable): Convert "tloc" from
290 location_t to dump_location_t. Updates calls to dump_printf_loc
291 to pass in a dump_location_t rather than a location_t, via the
293 (grid_inner_loop_gridifiable_p): Likewise.
294 (grid_dist_follows_simple_pattern): Likewise.
295 (grid_gfor_follows_tiling_pattern): Likewise.
296 (grid_target_follows_gridifiable_pattern): Likewise.
297 (grid_attempt_target_gridification): Convert initialization
298 of local "grid" from memset to zero-initialization; FIXME: does
299 this require C++11? Update call to dump_printf_loc to pass in a
300 optinfo_location rather than a location_t, via the gimple stmt.
301 * profile.c (read_profile_edge_counts): Updates call to
302 dump_printf_loc to pass in a dump_location_t rather than a
304 (compute_branch_probabilities): Likewise.
305 * selftest-run-tests.c (selftest::run_tests): Call
307 * selftest.h (dumpfile_c_tests): New decl.
308 * tree-loop-distribution.c (pass_loop_distribution::execute):
309 Update for change in return type of find_loop_location.
310 * tree-parloops.c (parallelize_loops): Likewise.
311 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
312 "locus" from location_t to dump_user_location_t.
313 (canonicalize_loop_induction_variables): Likewise.
314 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
315 for change in return type of find_loop_location.
316 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
317 to dump_printf_loc to pass in a dump_location_t rather than a
318 location_t, via the stmt.
319 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
321 * tree-vect-loop-manip.c (find_loop_location): Convert return
322 type from source_location to dump_user_location_t.
323 (vect_do_peeling): Update for above change.
324 (vect_loop_versioning): Update for change in type of
326 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
327 from location_t to dump_user_location_t.
328 (vect_estimate_min_profitable_iters): Update for change in type
330 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
331 location_t to dump_location_t.
332 (vect_slp_bb): Update for change in type of vect_location.
333 * tree-vectorizer.c (vect_location): Convert from source_location
334 to dump_user_location_t.
335 (try_vectorize_loop_1): Update for change in vect_location's type.
336 (vectorize_loops): Likewise.
337 (increase_alignment): Likewise.
338 * tree-vectorizer.h (vect_location): Convert from source_location
339 to dump_user_location_t.
340 (find_loop_location): Convert return type from source_location to
341 dump_user_location_t.
342 (check_reduction_path): Convert 1st param from location_t to
343 dump_user_location_t.
344 * value-prof.c (check_counter): Update call to dump_printf_loc to
345 pass in a dump_user_location_t rather than a location_t; update
346 call to error_at for change in type of "locus".
347 (check_ic_target): Update call to dump_printf_loc to
348 pass in a dump_user_location_t rather than a location_t, via the
351 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
353 * config/s390/s390.h (enum processor_flags): Do not use
354 default tune parameter when -march was specified.
356 2018-06-26 Jakub Jelinek <jakub@redhat.com>
359 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
360 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
363 2018-06-26 Richard Biener <rguenther@suse.de>
365 PR tree-optimization/86287
367 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
368 (vect_analyze_loop): Initialize n_stmts.
370 2018-06-26 Richard Biener <rguenther@suse.de>
373 * fold-const.c (fold_convertible_p): Pointer extension
376 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
379 * dwarf2out.c (loc_list_has_views): Adjust comments.
380 (dw_loc_list): Split single cross-partition range with
383 2018-06-25 Jeff Law <law@redhat.com>
385 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
386 on -mbig-switch by default.
388 * config/v850/predicates.md (const_float_1_operand): Fix match_code
390 (const_float_0_operand): Remove unused predicate.
391 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
392 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
393 (recipsf2): New expander. Original pattern now called
395 (recipdf2, recipdf2_insn): Similarly.
396 (rsqrtsf2, rsqrtsf2_insn): Similarly
397 (rsqrtdf2, rsqrtdf2_insn): Similarly
399 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
401 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
402 Simplify logic for FreeBSD (twice).
404 2018-06-25 Martin Sebor <msebor@redhat.com>
406 PR tree-optimization/86204
407 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
408 a strnlen result if it's less than the length of the string.
410 2018-06-25 Martin Sebor <msebor@redhat.com>
412 PR tree-optimization/85700
413 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
414 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
415 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
417 2018-06-25 Martin Sebor <msebor@redhat.com>
419 * doc/extend.texi (Zero-length arrays): Update and clarify.
421 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
423 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
424 added IEEE/IBM long double multilib support on PowerPC little
425 endian Linux systems.
426 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
427 (MULTILIB_DEFAULTS): Likewise.
428 * config/rs6000/rs6000.c (rs6000_option_override_internal):
430 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
431 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
432 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
434 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
437 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
438 (REORDER_45): Likewise.
440 2018-06-25 Jeff Law <law@redhat.com>
442 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
443 dividend to 32 bits. Adjust length.
444 (udivmodhi4): Cleanup output template. Fix length.
446 2018-06-25 Carl Love <cel@us.ibm.com>
448 * config/rs6000/vsx.md: Change word selector to prefered location.
450 2018-06-25 Richard Biener <rguenther@suse.de>
452 PR tree-optimization/86304
453 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
454 epilogue-if-converted loops as well.
456 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
458 * lto-section-out.c (lto_begin_section): Do not print section
459 name for noaddr and unnumbered dumps.
461 2018-06-25 Richard Biener <rguenther@suse.de>
463 * tree-vectorizer.h (struct vec_info_shared): New structure
464 with parts split out from struct vec_info and loop_nest from
465 struct _loop_vec_info.
466 (struct vec_info): Adjust accordingly.
467 (struct _loop_vec_info): Likewise.
468 (LOOP_VINFO_LOOP_NEST): Adjust.
469 (LOOP_VINFO_DATAREFS): Likewise.
470 (LOOP_VINFO_DDRS): Likewise.
471 (struct _bb_vec_info): Likewise.
472 (BB_VINFO_DATAREFS): Likewise.
473 (BB_VINFO_DDRS): Likewise.
474 (struct _stmt_vec_info): Add dr_aux member.
475 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
476 (DR_MISALIGNMENT_UNINITIALIZED): New.
477 (set_dr_misalignment): Adjust.
478 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
479 (vect_analyze_loop): Adjust prototype.
480 (vect_analyze_loop_form): Likewise.
481 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
482 Compute dependences lazily.
483 (vect_record_base_alignments): Use shared datarefs/ddrs.
484 (vect_verify_datarefs_alignment): Likewise.
485 (vect_analyze_data_refs_alignment): Likewise.
486 (vect_analyze_data_ref_accesses): Likewise.
487 (vect_analyze_data_refs): Likewise.
488 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
489 constructor parameter for shared part.
490 (vect_analyze_loop_form): Pass in shared part and adjust.
491 (vect_analyze_loop_2): Pass in storage for the number of
492 stmts. Move loop nest finding to the caller. Compute
494 (vect_analyze_loop): Pass in shared part.
495 (vect_transform_loop): Verify shared datarefs are unchanged.
496 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
497 constructor parameter for shared part.
498 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
499 (vect_slp_bb): Verify shared datarefs are unchanged before
501 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
503 (new_stmt_vec_info): Initialize DR_AUX misalignment to
504 DR_MISALIGNMENT_UNINITIALIZED.
505 * tree-vectorizer.c (vec_info::vec_info): Add constructor
506 parameter for shared part.
507 (vec_info::~vec_info): Adjust.
508 (vec_info_shared::vec_info_shared): New.
509 (vec_info_shared::~vec_info_shared): Likewise.
510 (vec_info_shared::save_datarefs): Likewise.
511 (vec_info_shared::check_datarefs): Likewise.
512 (try_vectorize_loop_1): Construct shared part live for analyses
513 of a single loop for multiple vector sizes.
514 * tree-parloops.c (gather_scalar_reductions): Adjust.
516 2018-06-25 Richard Biener <rguenther@suse.de>
518 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
519 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
520 (vect_analyze_data_refs): Remove similar code from here and
521 simplify accordingly.
523 2018-06-25 Richard Biener <rguenther@suse.de>
525 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
526 for reverse storage order accesses rather than asserting
527 they cannot happen here.
529 2018-06-25 Tom de Vries <tdevries@suse.de>
532 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
533 Use data16 instead of .byte for insn prefix.
535 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
538 * parser.c (make_char_string_pack): Pass this literal chars
539 through cpp_interpret_string.
540 (cp_parser_userdef_numeric_literal): Check the result of
541 make_char_string_pack.
543 2018-06-24 Maya Rashish <coypu@sdf.org>
545 * ginclude/stddef.h: Simplify conditions around avoiding
546 re-definition of __size_t.
548 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
550 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
551 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
553 2018-06-22 Maya Rashish <coypu@sdf.org>
555 * doc/invoke.texi (mno-fancy-math-387): Update for changes
556 made to OpenBSD and NetBSD through the years.
558 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
560 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
561 behavior of vec_pack (vector double, vector double) to match
562 behavior of vec_float2 (vector double, vector double).
564 2018-06-22 Olivier Hainque <hainque@adacore.com>
566 * gimplify.c (gimplify_function_tree): Prevent creation
567 of a trampoline for the address of the current function
568 passed to entry/exit instrumentation hooks.
570 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
573 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
576 2018-06-22 Martin Liska <mliska@suse.cz>
578 PR tree-optimization/86263
579 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
580 Bail out if is_enabled is false.
581 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
583 (jump_table_cluster::is_enabled): New function.
585 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
587 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
588 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
589 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
590 (lto_input_ts_binfo_tree_pointers): Likewise.
591 * tree-streamer-out.c (streamer_write_tree_bitfields,
592 write_ts_binfo_tree_pointers): Likewise.
593 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
595 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
597 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
599 2018-06-22 Martin Liska <mliska@suse.cz>
601 * symbol-summary.h (get): Make it pure and inline move
602 functionality from ::get function.
603 (get): Remove and inline into ::get and ::get_create.
604 (get_create): Move code from ::get function.
606 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
609 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
610 -x assembler-with-cpp.
612 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
614 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
615 _FILE_OFFSET_BITS=64 for C++.
617 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
619 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
620 conversion insn that shows up when pr85657-3.c is compiled using
621 IEEE 128-bit long double.
622 (neg<mode>2_internal): Use the correct mode to check whether the
623 mode is IBM extended.
624 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
625 multiply and divide external functions from being created more
628 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
630 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
632 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
633 the edge can be forwarded.
634 (cfg_layout_merge_blocks): Likewise.
636 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
638 * except.c (finish_eh_generation): Commit edge insertions only after
639 the EH edges have been redirected from post-landing to landing pads.
641 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
643 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
644 create_tmp_var_for to create the FRAME decl.
645 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
647 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
649 * tree-inline.c (copy_edges_for_bb): Minor tweak.
650 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
651 debug statement when resetting its value.
652 (expand_call_inline): Copy the locus of the call onto the assignment
653 of the return value, if any. Use local variable in more cases.
655 2018-06-21 Martin Liska <mliska@suse.cz>
657 * ipa-pure-const.c (propagate_nothrow): Use
658 funct_state_summaries->get.
659 (dump_malloc_lattice): Likewise.
660 (propagate_malloc): Likewise.
662 2018-06-21 Richard Biener <rguenther@suse.de>
664 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
665 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
666 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
668 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
669 BLOCK_ABSTRACT_ORIGIN unconditionally.
671 2018-06-21 David Malcolm <dmalcolm@redhat.com>
673 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
675 * ipa-reference.c (ipa_reference_c_finalize): Delete
676 ipa_ref_opt_sum_summaries and set it to NULL.
678 2018-06-21 Tom de Vries <tdevries@suse.de>
680 PR tree-optimization/85859
681 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
682 test with comment from bb_no_side_effects_p.
684 2018-06-21 Richard Biener <rguenther@suse.de>
686 PR tree-optimization/86232
687 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
688 max for constant niter.
690 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
692 * config/aarch64/aarch64-simd.md
693 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
695 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
697 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
698 Make opernads of the unspec commutative.
700 2018-06-21 Richard Biener <rguenther@suse.de>
702 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
703 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
704 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
705 (vect_analyze_data_ref_dependence): Re-order checks to deal with
707 (vect_record_base_alignments): Do not record base alignment
708 for gathers or scatters.
709 (vect_compute_data_ref_alignment): Drop return value that is always
710 true. Bail out early for gathers or scatters.
711 (vect_enhance_data_refs_alignment): Bail out early for gathers
713 (vect_find_same_alignment_drs): Likewise.
714 (vect_analyze_data_refs_alignment): Remove dead code.
715 (vect_slp_analyze_and_verify_node_alignment): Likewise.
716 (vect_analyze_data_refs): For possible gathers or scatters do
717 not create an alternate DR, just check their possible validity
718 and mark them. Adjust DECL_NONALIASED handling to not rely
720 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
721 update inits of gathers or scatters.
722 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
723 Also copy gather/scatter flag to pattern vinfo.
725 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
727 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
728 behavior of vec_packsu (vector unsigned long long, vector unsigned
729 long long) to match behavior of vec_packs with same signature.
731 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
732 Thomas Schwinge <thomas@codesourcery.com>
733 Cesar Philippidis <cesar@codesourcery.com>
735 * gimplify.c (gimplify_scan_omp_clauses): Add support for
736 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
737 (gimplify_adjust_omp_clauses): Likewise.
738 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
739 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
740 (gimplify_omp_target_update): Update handling of acc update and
742 * omp-low.c (install_var_field): Remove unused parameter
743 base_pointers_restrict.
744 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
745 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
747 (omp_target_base_pointers_restrict_p): Delete.
748 (scan_omp_target): Update call to scan_sharing_clauses.
749 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
751 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
752 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
753 (convert_local_omp_clauses): Likewise.
754 * tree-pretty-print.c (dump_omp_clause): Likewise.
755 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
757 (omp_clause_code_name): Likewise.
759 2018-06-20 Jakub Jelinek <jakub@redhat.com>
762 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
765 PR tree-optimization/86231
766 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
767 anti-range don't overwrite *vr0min before using it to compute *vr0max.
769 2018-06-20 Tom de Vries <tdevries@suse.de>
771 PR tree-optimization/86097
772 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
773 iv type if signedness of iv type is not the same as that of *nit.
775 2018-06-20 Jakub Jelinek <jakub@redhat.com>
777 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
778 EDGE_EH edges, verify they are all EDGE_EH.
780 2018-06-20 Maya Rashish <coypu@sdf.org>
782 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
784 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
786 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
787 * config/aarch64/aarch64.c (xgene1_tunings): Add
788 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
789 (aarch64_mode_valid_for_sched_fusion_p):
791 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
792 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
794 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
796 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
797 * config/aarch64/iterators.md (VQ2): New mode iterator.
799 2018-06-20 Martin Liska <mliska@suse.cz>
801 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
802 Change default ratio from 10 to 8.
804 2018-06-20 Martin Liska <mliska@suse.cz>
806 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
808 (bit_test_cluster::find_bit_tests): Likewise.
809 (switch_decision_tree::analyze_switch_statement): Find clusters.
810 * tree-switch-conversion.h (struct jump_table_cluster): Document
813 2018-06-20 Martin Liska <mliska@suse.cz>
815 * tree-switch-conversion.c (switch_conversion::collect):
816 Record m_uniq property.
817 (switch_conversion::expand): Bail out for special conditions.
818 (group_cluster::~group_cluster): New.
819 (group_cluster::group_cluster): Likewise.
820 (group_cluster::dump): Likewise.
821 (jump_table_cluster::emit): New.
822 (switch_decision_tree::fix_phi_operands_for_edges): New.
823 (struct case_node): Remove struct.
824 (jump_table_cluster::can_be_handled): New.
825 (case_values_threshold): Moved to header.
826 (reset_out_edges_aux): Likewise.
827 (jump_table_cluster::is_beneficial): New.
828 (bit_test_cluster::can_be_handled): Likewise.
829 (add_case_node): Remove.
830 (bit_test_cluster::is_beneficial): New.
831 (case_bit_test::cmp): New.
832 (bit_test_cluster::emit): New.
833 (expand_switch_as_decision_tree_p): Remove.
834 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
835 (fix_phi_operands_for_edge): Likewise.
836 (switch_decision_tree::analyze_switch_statement): New.
837 (compute_cases_per_edge): Move ...
838 (switch_decision_tree::compute_cases_per_edge): ... here.
839 (try_switch_expansion): Likewise.
840 (switch_decision_tree::try_switch_expansion): Likewise.
841 (record_phi_operand_mapping): Likewise.
842 (switch_decision_tree::record_phi_operand_mapping): Likewise.
843 (emit_case_decision_tree): Likewise.
844 (switch_decision_tree::emit): Likewise.
845 (balance_case_nodes): Likewise.
846 (switch_decision_tree::balance_case_nodes): Likewise.
847 (dump_case_nodes): Likewise.
848 (switch_decision_tree::dump_case_nodes): Likewise.
849 (emit_jump): Likewise.
850 (switch_decision_tree::emit_jump): Likewise.
851 (emit_cmp_and_jump_insns): Likewise.
852 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
853 (emit_case_nodes): Likewise.
854 (switch_decision_tree::emit_case_nodes): Likewise.
855 (conditional_probability): Remove.
856 * tree-switch-conversion.h (enum cluster_type): New.
858 (struct cluster): Likewise.
859 (cluster::cluster): Likewise.
860 (struct simple_cluster): Likewise.
861 (simple_cluster::simple_cluster): Likewise.
862 (struct group_cluster): Likewise.
863 (struct jump_table_cluster): Likewise.
864 (struct bit_test_cluster): Likewise.
865 (struct min_cluster_item): Likewise.
866 (struct case_tree_node): Likewise.
867 (case_tree_node::case_tree_node): Likewise.
868 (jump_table_cluster::case_values_threshold): Likewise.
869 (struct case_bit_test): Likewise.
870 (struct switch_decision_tree): Likewise.
871 (struct switch_conversion): Likewise.
872 (switch_decision_tree::reset_out_edges_aux): Likewise.
874 2018-06-20 Martin Liska <mliska@suse.cz>
876 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
877 (hoist_edge_and_branch_if_true): Likewise.
878 (expand_switch_using_bit_tests_p): Likewise.
879 (struct case_bit_test): Likewise.
880 (case_bit_test_cmp): Likewise.
881 (emit_case_bit_tests): Likewise.
882 (switch_conversion::switch_conversion): New class.
883 (struct switch_conv_info): Remove old struct.
884 (collect_switch_conv_info): More to ...
885 (switch_conversion::collect): ... this.
886 (check_range): Likewise.
887 (switch_conversion::check_range): Likewise.
888 (check_all_empty_except_final): Likewise.
889 (switch_conversion::check_all_empty_except_final): Likewise.
890 (check_final_bb): Likewise.
891 (switch_conversion::check_final_bb): Likewise.
892 (create_temp_arrays): Likewise.
893 (switch_conversion::create_temp_arrays): Likewise.
894 (free_temp_arrays): Likewise.
895 (gather_default_values): Likewise.
896 (switch_conversion::gather_default_values): Likewise.
897 (build_constructors): Likewise.
898 (switch_conversion::build_constructors): Likewise.
899 (constructor_contains_same_values_p): Likewise.
900 (switch_conversion::contains_same_values_p): Likewise.
901 (array_value_type): Likewise.
902 (switch_conversion::array_value_type): Likewise.
903 (build_one_array): Likewise.
904 (switch_conversion::build_one_array): Likewise.
905 (build_arrays): Likewise.
906 (switch_conversion::build_arrays): Likewise.
907 (gen_def_assigns): Likewise.
908 (switch_conversion::gen_def_assigns): Likewise.
909 (prune_bbs): Likewise.
910 (switch_conversion::prune_bbs): Likewise.
911 (fix_phi_nodes): Likewise.
912 (switch_conversion::fix_phi_nodes): Likewise.
913 (gen_inbound_check): Likewise.
914 (switch_conversion::gen_inbound_check): Likewise.
915 (process_switch): Use the newly created class.
916 (switch_conversion::expand): New.
917 (switch_conversion::~switch_conversion): New.
918 * tree-switch-conversion.h: New file.
920 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
922 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
923 tree-vect-patterns.c.
924 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
925 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
926 (vect_recog_sad_pattern): Likewise.
927 (vect_recog_widen_sum_pattern): Likewise.
928 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
929 (vect_recog_widen_shift_pattern): Remove the type_in argument.
930 (vect_recog_rotate_pattern): Likewise.
931 (vect_recog_mult_pattern): Likewise.
932 (vect_recog_vector_vector_shift_pattern): Likewise.
933 (vect_recog_divmod_pattern): Likewise.
934 (vect_recog_mixed_size_cond_pattern): Likewise.
935 (vect_recog_bool_pattern): Likewise.
936 (vect_recog_mask_conversion_pattern): Likewise.
937 (vect_try_gather_scatter_pattern): Likewise.
938 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
939 (vect_recog_over_widening_pattern): Likewise.
940 (vect_recog_gather_scatter_pattern): Likewise.
941 (vect_recog_func_ptr): Move from tree-vectorizer.h
942 (vect_vect_recog_func_ptrs): Move further down the file.
943 (vect_recog_func): Likewise. Remove the third argument.
944 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
945 (vect_pattern_recog_1): Expect the pattern function to do any
946 necessary target tests. Also expect it to provide a vector type.
947 Remove the type_in handling.
949 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
951 * tree-vect-patterns.c (vect_pattern_detected): New function.
952 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
953 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
954 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
955 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
956 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
957 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
958 (vect_recog_mask_conversion_pattern)
959 (vect_try_gather_scatter_pattern): Likewise.
961 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
963 * tree-vect-patterns.c (vect_get_internal_def): New function.
964 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
965 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
966 (search_type_for_mask_1): Use it.
968 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
970 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
971 redundant WIDEN_SUM_EXPR handling.
972 (vect_recog_sad_pattern): Likewise.
974 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
976 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
977 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
978 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
979 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
980 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
982 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
984 * tree-vect-stmts.c (vectorizable_call): Make sure that we
985 use the stmt_vec_info of the original bb statement for the
986 new zero assignment, even if the call is part of a pattern.
988 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
990 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
991 that the sequence is attached to the original statement rather
992 than the pattern statement.
993 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
994 PATTERN_DEF_SEQ from the original statement rather than
995 the main pattern statement.
996 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
997 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
998 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1000 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1002 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1003 definition statements before the early exit for statements that aren't
1005 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1007 (vect_transform_loop): ...here. Process pattern definition
1008 statements without first checking whether the main pattern
1009 statement is live or relevant.
1011 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1013 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1014 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1016 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1018 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1019 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1020 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1021 (expand_block_compare): Change select_block_compare_mode call.
1022 (expand_strncmp_align_check): Use new functions, fix comment.
1023 (emit_final_str_compare_gpr): New function.
1024 (expand_strn_compare): Refactor and clean up code.
1025 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1027 2018-06-19 Tony Reix <tony.reix@atos.com>
1028 Damien Bergamini <damien.bergamini@atos.com>
1029 David Edelsohn <dje.gcc@gmail.com>
1031 * collect2.c (static_obj): New variable.
1032 (static_libs): New variable.
1033 (is_in_list): Uncomment declaration.
1034 (main): Track AIX libraries linked statically.
1035 (is_in_list): Uncomment definition.
1036 (scan_prog_file): Don't add AIX shared libraries initializer
1037 to constructor list if linking statically.
1039 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1041 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1043 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1045 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1047 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1050 2018-06-19 Martin Liska <mliska@suse.cz>
1052 * config/i386/i386.c (ix86_can_inline_p): Do not use
1053 ipa_fn_summaries::get_create.
1054 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1056 (devirtualization_time_bonus): Likewise.
1057 (ipcp_propagate_stage): Likewise.
1058 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1059 (edge_set_predicate): Likewise.
1060 (evaluate_conditions_for_known_args): Likewise.
1061 (evaluate_properties_for_edge): Likewise.
1062 (ipa_call_summary::reset): Tranform to ...
1063 (ipa_call_summary::~ipa_call_summary): ... this.
1064 (ipa_fn_summary::reset): Transform to ...
1065 (ipa_fn_summary::~ipa_fn_summary): ... this.
1066 (ipa_fn_summary_t::remove): Rename to ...
1067 (ipa_fn_summary_t::remove_callees): ... this.
1068 (ipa_fn_summary_t::duplicate): Use placement new
1069 instead of memory copy.
1070 (ipa_call_summary_t::duplicate): Likewise.
1071 (ipa_call_summary_t::remove): Remove.
1072 (dump_ipa_call_summary): Change get_create to get.
1073 (ipa_dump_fn_summary): Dump only when summary exists.
1074 (analyze_function_body): Use symbol_summary::get instead
1076 (compute_fn_summary): Likewise.
1077 (estimate_edge_devirt_benefit): Likewise.
1078 (estimate_edge_size_and_time): Likewise.
1079 (inline_update_callee_summaries): Likewise.
1080 (remap_edge_change_prob): Likewise.
1081 (remap_edge_summaries): Likewise.
1082 (ipa_merge_fn_summary_after_inlining): Likewise.
1083 (write_ipa_call_summary): Likewise.
1084 (ipa_fn_summary_write): Likewise.
1085 (ipa_free_fn_summary): Likewise.
1086 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1087 (struct ipa_call_summary): Likewise.
1088 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1090 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1091 (estimate_size_after_inlining): Likewise.
1092 (estimate_growth): Likewise.
1093 (growth_likely_positive): Likewise.
1094 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1095 (inline_call): Likewise.
1096 * ipa-inline.c (caller_growth_limits): Likewise.
1097 (can_inline_edge_p): Likewise.
1098 (can_inline_edge_by_limits_p): Likewise.
1099 (compute_uninlined_call_time): Likewise.
1100 (compute_inlined_call_time): Likewise.
1101 (want_inline_small_function_p): Likewise.
1102 (edge_badness): Likewise.
1103 (update_caller_keys): Likewise.
1104 (update_callee_keys): Likewise.
1105 (inline_small_functions): Likewise.
1106 (inline_to_all_callers_1): Likewise.
1107 (dump_overall_stats): Likewise.
1108 (early_inline_small_functions): Likewise.
1109 (early_inliner): Likewise.
1110 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1111 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1112 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1113 * ipa-split.c (execute_split_functions): Likewise.
1114 * symbol-summary.h: Likewise.
1115 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1117 2018-06-19 Richard Biener <rguenther@suse.de>
1119 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1120 (vectorize_loops): ... here. Fix dbgcnt handling.
1121 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1123 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1126 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1127 ieee128 argument takes up only one (vector) register, not two (floating
1130 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1132 * gimplify.c (gimplify_init_constructor): Really never clear for an
1133 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1135 2018-06-19 Richard Biener <rguenther@suse.de>
1137 PR tree-optimization/86179
1138 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1139 after failed recognition.
1141 2018-06-18 Martin Sebor <msebor@redhat.com>
1144 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1145 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1146 Handle integer subtraction.
1147 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1148 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1150 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1152 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1153 param from rtx to rtx_insn *.
1154 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1156 (frv_ifcvt_modify_insn): Likwise.
1157 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1158 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1159 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1160 as_a <rtx_insn *> cast to local "unprotected_region" once
1161 it's been established that it's not NULL or pc_rtx.
1162 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1163 param "sethi" from rtx to rtx_insn *.
1164 (nds32_group_float_insns): Likewise for param "insn".
1165 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1167 (vax_output_int_subtract): Likewise.
1168 * config/vax/vax.c (vax_output_int_add): Likewise for param
1170 (vax_output_int_subtract): Likewise.
1171 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1172 (emit_pattern_after): Likewise for param "after".
1173 (emit_insn_after): Likewise.
1174 (emit_jump_insn_after): Likewise.
1175 (emit_call_insn_after): Likewise.
1176 (emit_debug_insn_after): Likewise.
1177 (emit_pattern_before): Likewise for param "before".
1178 (emit_insn_before): Likewise.
1179 (emit_jump_insn_before): Likewise.
1180 * final.c (get_insn_template): Likewise for param "insn", removing
1182 * output.h (get_insn_template): Likewise for 2nd param.
1183 * rtl.h (emit_insn_before): Likewise.
1184 (emit_jump_insn_before): Likewise.
1185 (emit_debug_insn_before_noloc): Likewise.
1186 (emit_insn_after): Likewise.
1187 (emit_jump_insn_after): Likewise.
1188 (emit_call_insn_after): Likewise.
1189 (emit_debug_insn_after): Likewise.
1190 (set_insn_deleted): Likewise for param.
1192 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1195 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1196 floating point modes, so that IFmode is numerically greater than
1197 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1198 to declare the ordering. This prevents IFmode from being
1199 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1200 machine. Include rs6000-modes.h to share the fractional values
1201 between genmodes* and the rest of the compiler.
1205 * config/rs6000/rs6000-modes.h: New file.
1206 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1207 meaning of rs6000_long_double_size so that 126..128 selects an
1208 appropriate 128-bit floating point type.
1209 (rs6000_option_override_internal): Likewise.
1210 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1211 (TARGET_LONG_DOUBLE_128): Change the meaning of
1212 rs6000_long_double_size so that 126..128 selects an appropriate
1213 128-bit floating point type.
1214 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1215 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1216 source and destination to match the standard usage.
1217 (truncifkf2): Likewise.
1218 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1219 ISA 2.07 to use an explicit clobber, instead of passing in a
1221 (copysign<mode>3_soft): Likewise.
1223 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1225 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1226 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1227 (vect_slp_analyze_instance_dependence): Likewise.
1228 (vect_enhance_data_refs_alignment): Likewise.
1229 (vect_analyze_data_refs_alignment): Likewise.
1230 (vect_slp_analyze_and_verify_instance_alignment
1231 (vect_analyze_data_ref_accesses): Likewise.
1232 (vect_prune_runtime_alias_test_list): Likewise.
1233 (vect_analyze_data_refs): Likewise.
1234 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1235 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1236 (vect_analyze_scalar_cycles_1): Likewise.
1237 (vect_get_loop_niters): Likewise.
1238 (vect_analyze_loop_form_1): Likewise.
1239 (vect_update_vf_for_slp): Likewise.
1240 (vect_analyze_loop_operations): Likewise.
1241 (vect_analyze_loop): Likewise.
1242 (vectorizable_induction): Likewise.
1243 (vect_transform_loop): Likewise.
1244 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1245 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1246 (vect_make_slp_decision): Likewise.
1247 (vect_detect_hybrid_slp): Likewise.
1248 (vect_slp_analyze_operations): Likewise.
1249 (vect_slp_bb): Likewise.
1250 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1251 (vectorizable_bswap): Likewise.
1252 (vectorizable_call): Likewise.
1253 (vectorizable_simd_clone_call): Likewise.
1254 (vectorizable_conversion): Likewise.
1255 (vectorizable_assignment): Likewise.
1256 (vectorizable_shift): Likewise.
1257 (vectorizable_operation): Likewise.
1258 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1260 2018-06-18 Martin Sebor <msebor@redhat.com>
1262 PR tree-optimization/81384
1263 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1264 * builtins.c (expand_builtin_strnlen): New function.
1265 (expand_builtin): Call it.
1266 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1267 * builtins.def (BUILT_IN_STRNLEN): New.
1268 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1269 Warn for bounds in excess of maximum object size.
1270 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1271 single-value ranges. Handle strnlen.
1272 (handle_builtin_strlen): Handle strnlen.
1273 (strlen_check_and_optimize_stmt): Same.
1274 * doc/extend.texi (Other Builtins): Document strnlen.
1276 2018-06-18 Maya Rashish <coypu@sdf.org>
1278 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1279 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1280 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1282 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1284 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1286 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1288 * tree.c (escaped_string::escape): Replace cast to char * by
1289 const_cast<char *> (unescaped).
1291 2018-06-18 Nick Clifton <nickc@redhat.com>
1294 * tree.c (escaped_string): New class. Converts an unescaped
1295 string into its escaped equivalent.
1296 (warn_deprecated_use): Use the new class to convert the
1297 deprecation message, if present.
1298 (test_escaped_strings): New self test.
1299 (test_c_tests): Add test_escaped_strings.
1300 * doc/extend.texi (deprecated): Add a note that the
1301 deprecation message is affected by the -fmessage-length
1302 option, and that control characters will be escaped.
1303 (#pragma GCC error): Document this pragma.
1304 (#pragma GCC warning): Likewise.
1305 * doc/invoke.texi (-fmessage-length): Document this option's
1306 effect on the #warning and #error preprocessor directives and
1307 the deprecated attribute.
1309 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1311 * tree.c (decl_value_expr_lookup): Revert latest change.
1312 (decl_value_expr_insert): Likewise.
1314 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1316 * gimplify.c (nonlocal_vlas): Delete.
1317 (nonlocal_vla_vars): Likewise.
1318 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1320 (gimplify_body): Do not create and destroy nonlocal_vlas.
1321 * tree-nested.c: Include diagnostic.h.
1322 (use_pointer_in_frame): Tweak.
1323 (lookup_field_for_decl): Add assertion and declare the transformation.
1324 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1325 internal error when the reference is in a wrong context. Do not
1326 create a debug decl by default.
1327 (note_nonlocal_block_vlas): Delete.
1328 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1329 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1330 create a debug decl by default.
1331 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1332 call is in a wrong context.
1333 (fixup_vla_decls): New function.
1334 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1335 debug variables were created.
1336 * tree.c (decl_value_expr_lookup): Add checking assertion.
1337 (decl_value_expr_insert): Likewise.
1339 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1342 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1343 * tree-scalar-evolution.c (interpret_expr): Likewise.
1344 (expression_expensive_p): Likewise.
1345 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1346 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1347 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1348 (ssa_defined_by_minus_one_stmt_p): New.
1350 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1353 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1354 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1355 * dojump.c (do_jump): Likewise.
1356 * expr.c (expand_expr_real_2): Check operand type's sign.
1357 * fold-const.c (const_unop): Handle ABSU_EXPR.
1358 (fold_abs_const): Likewise.
1359 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1360 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1361 (strip_sign_op_1): Likesise.
1362 * match.pd: Add new pattern to generate ABSU_EXPR.
1363 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1364 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1365 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1366 * tree-inline.c (estimate_operator_cost): Likewise.
1367 * tree-pretty-print.c (dump_generic_node): Likewise.
1368 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1369 * tree.def (ABSU_EXPR): New.
1371 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1374 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1375 documented as preserved for backward compatibility only.
1376 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1378 PR rtl-optimization/86108
1379 * bb-reorder.c (create_forwarder_block): Renamed to ...
1380 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1381 jump from new landing pad to the second part.
1382 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1385 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1388 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1389 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1390 Only call store_expr for halves if the mode is the same.
1393 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1394 Fix up comment formatting.
1396 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1398 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1399 (typed_splay_tree::closure,
1400 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1401 (typed_splay_tree::typed_splay_tree,
1402 typed_splay_tree::operator =): Declared private.
1403 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1404 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1405 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1406 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1407 typed_splay_tree::splay_tree_splay,
1408 typed_splay_tree::splay_tree_foreach_helper,
1409 typed_splay_tree::splay_tree_insert,
1410 typed_splay_tree::splay_tree_remove,
1411 typed_splay_tree::splay_tree_lookup,
1412 typed_splay_tree::splay_tree_predecessor,
1413 typed_splay_tree::splay_tree_successor,
1414 typed_splay_tree::splay_tree_min,
1415 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1416 (typed_splay_tree::root, typed_splay_tree::comp,
1417 typed_splay_tree::delete_key,
1418 typed_splay_tree::delete_value): New data members.
1419 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1420 typed_splay_tree::remove.
1422 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1424 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1425 -mginv and -mno-ginv to the assembler.
1426 * config/mips/mips.opt (-mcrc): New option.
1428 * doc/invoke.text (-mcrc): Document.
1431 2018-06-15 Nick Clifton <nickc@redhat.com>
1434 * tree.c (escaped_string): New class. Converts an unescaped
1435 string into its escaped equivalent.
1436 (warn_deprecated_use): Use the new class to convert the
1437 deprecation message, if present.
1438 (test_escaped_strings): New self test.
1439 (test_c_tests): Add test_escaped_strings.
1440 * doc/extend.texi (deprecated): Add a note that the
1441 deprecation message is affected by the -fmessage-length
1442 option, and that control characters will be escaped.
1443 (#pragma GCC error): Document this pragma.
1444 (#pragma GCC warning): Likewise.
1445 * doc/invoke.texi (-fmessage-length): Document this option's
1446 effect on the #warning and #error preprocessor directives and
1447 the deprecated attribute.
1449 2018-06-15 Richard Biener <rguenther@suse.de>
1451 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1452 here, also noting vector size used.
1453 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1454 size used in MSG_OPTIMIZED_LOCATIONS dump.
1455 (pass_slp_vectorize::execute): Adjust.
1457 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1460 * config/arc/arc.c (arc_return_address_register): Fix
1463 2018-06-15 Richard Biener <rguenther@suse.de>
1466 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1467 leave useless conversion stripping to force_gimple_operand_gsi.
1468 (gimplify_build2): Likewise.
1469 (gimplify_build1): Likewise.
1471 2018-06-15 Richard Biener <rguenther@suse.de>
1474 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1475 before adjusting their block.
1477 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1479 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1480 multilibs for *-*-rtems*.
1481 * config/riscv/t-rtems: New file.
1483 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1486 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1487 unsigned_type_for returns NULL.
1490 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1491 subregs of multi-word pseudos unless the float mode has word size.
1493 2018-06-14 Richard Biener <rguenther@suse.de>
1496 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1497 duplicate and harmful type_hash_canon.
1498 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1500 2018-06-14 Richard Biener <rguenther@suse.de>
1503 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1506 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1508 * config/rtems.h (STDINT_LONG32): Define.
1510 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1511 Prachi Godbole <prachi.godbole@imgtec.com>
1513 * config/mips/mips-cpus.def: Define P6600.
1514 * config/mips/mips-tables.opt: Regenerate.
1515 * config/mips/mips.c (mips_ucbranch_type): New enum.
1516 (mips_rtx_cost_data): Add support for P6600.
1517 (mips_issue_rate): Likewise.
1518 (mips_multipass_dfa_lookahead): Likewise.
1519 (mips_avoid_hazard): Likewise.
1520 (mips_reorg_process_insns): Likewise.
1521 (mips_classify_branch_p6600): New function.
1522 * config/mips/mips.h (TUNE_P6600): New define.
1523 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1524 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1525 * config/mips/mips.md: Include p6600.md.
1526 (processor): Add p6600.
1527 * config/mips/p6600.md: New file.
1528 * doc/invoke.texi: Add p6600 to supported architectures.
1530 2018-06-13 Martin Sebor <msebor@redhat.com>
1532 PR tree-optimization/86114
1533 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1535 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1537 2018-06-13 Richard Biener <rguenther@suse.de>
1539 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1540 Properly set vector type of the intermediate stmt.
1541 * tree-vect-stmts.c (vectorizable_operation): The destination
1542 var always has vectype_out type.
1544 2018-06-13 Jeff Law <law@redhat.com>
1546 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1547 integer 0 for argument to print_rtl_with_bb.
1548 (rl78_reorg): Likewise.
1550 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1552 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1553 from rtx to rtx_insn *.
1554 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1556 (add_sched_insns_for_speculation): Likewise for local "target",
1557 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1558 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1559 from rtx_insn ** to rtx_code_label **.
1560 (reorg_emit_nops): Likewise.
1561 (c6x_reorg): Likewise for local "call_labels".
1562 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1564 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1565 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1566 the loops over LABEL_REFS.
1567 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1569 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1570 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1571 (split_branches): Strengthen local "olabel" from rtx to
1572 rtx_insn *, adding a safe_as_a cast.
1573 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1575 (add_insn_after): Likewise for first two params.
1576 (add_insn_before): Likewise.
1577 (remove_insn): Likewise for param.
1578 (emit_pattern_before_noloc): Likewise for second and third params.
1579 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1580 (emit_call_insn_before_noloc): Likewise.
1581 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1583 (emit_barrier_before): Likewise.
1584 (emit_label_before): Strengthen "label" param from "rtx" to
1585 "rtx_code_label *". Strengthen "before" param from "rtx" to
1587 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1589 (emit_pattern_after_noloc): Likewise.
1590 (emit_insn_after_noloc): Likewise.
1591 (emit_jump_insn_after_noloc): Likewise.
1592 (emit_call_insn_after_noloc): Likewise.
1593 (emit_debug_insn_after_noloc): Likewise.
1594 (emit_barrier_after): Likewise.
1595 (emit_label_after): Likewise for both params.
1596 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1597 "loc" param from "int" to "location_t".
1598 (emit_insn_after_setloc): Likewise.
1599 (emit_jump_insn_after_setloc): Likewise.
1600 (emit_call_insn_after_setloc): Likewise.
1601 (emit_debug_insn_after_setloc): Likewise.
1602 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1603 "loc" param from "int" to "location_t".
1604 (emit_pattern_before): Convert NULL_RTX to NULL.
1605 (emit_insn_before_setloc): Convert "loc" param from "int" to
1607 (emit_jump_insn_before_setloc): Likewise.
1608 (emit_call_insn_before_setloc): Likewise.
1609 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1610 rtx_insn *. Convert "loc" param from "int" to "location_t".
1611 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1612 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1613 Convert 3rd param from "int" to "location_t".
1614 (emit_barrier_before, emit_barrier_after, next_real_insn):
1615 Strengthen param from rtx to rtx_insn *.
1616 (emit_label_before): Strengthen 1st param from "rtx" to
1617 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1619 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1620 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1621 Strengthen 2nd param from "rtx" to "rtx_insn *".
1622 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1623 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1624 Likewise. Convert 3rd param from "int" to "location_t".
1625 (emit_label_after): Strengthen 1st param from "rtx" to
1627 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1629 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1630 from "rtx" to "rtx_insn *".
1632 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1634 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1635 bodies streamed in with -Q.
1636 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1637 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1638 * gimple-streamer-out.c: Include gimple-pretty-print.h
1639 (output_bb): Dump stmts streamed.
1640 * lto-section-out.c: Include print-tree.h
1641 (lto_begin_section): Dump sections created.
1642 (lto_output_decl_index): Dump decl encoded.
1643 * lto-streamer-out.c: Include print-tree.h
1644 (create_output_block): Dump output block created.
1645 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1646 (output_function): Dump function output.
1647 (output_constructor): Dump constructor streamed.
1648 (write_global_stream): Output indexes encoded.
1649 (produce_asm_for_decls): Dump streams encoded.
1650 * lto-streamer.c (streamer_dump_file): New global var.
1651 * lto-streamer.h (streamer_dump_file): Declare.
1652 * passes.c (ipa_write_summaries): Initialize streamer dump.
1653 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1656 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1659 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1660 offsets for register save directives. Emit a second batch of save
1661 directives, if need be, when the function accesses prior frames.
1663 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1665 * config/arc/fpu.md (fmasf4): Force operand to register.
1666 (fnmasf4): Likewise.
1668 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1670 * config/arc/arc-protos.h (arc_pad_return): Remove.
1671 * config/arc/arc.c (machine_function): Remove force_short_suffix
1673 (arc_print_operand): Adjust printing of '&'.
1674 (arc_verify_short): Remove conditional printing of short suffix.
1675 (arc_final_prescan_insn): Remove reference to size_reason.
1676 (pad_return): New function.
1677 (arc_reorg): Call pad_return.
1678 (arc_pad_return): Remove.
1679 (arc_init_machine_status): Remove reference to force_short_suffix.
1680 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1681 (attr length): When attribute iscompact is true force to 2
1682 regardless; in the case of maybe check if we want to force the
1683 instruction to have 4 bytes length.
1684 (nopv): Change it to generate 4 byte long nop as well.
1685 (blockage): New pattern.
1686 (simple_return): Remove call to arc_pad_return.
1687 (p_return_i): Likewise.
1689 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1691 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1693 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1695 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1698 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1700 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1701 for ARC700 and ARCv2.
1703 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1706 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1707 operands[2] instead of operands[1].
1710 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1712 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1713 case, check whether the outer register overlaps an unallocatable
1714 register, not just whether it fits the required class.
1716 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1718 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1720 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1721 (memrefs_conflict_p): Likewise.
1722 (init_alias_analysis): Likewise.
1723 * cfgexpand.c (expand_debug_expr): Likewise.
1724 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1725 * cse.c (fold_rtx): Likewise.
1726 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1727 * expr.c (emit_block_move_hints): Likewise.
1728 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1729 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1730 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1731 (emit_group_store): Likewise.
1732 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1733 to read the PRE/POST_MODIFY increment.
1734 * calls.c (store_one_arg): Use strip_offset.
1735 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1737 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1739 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1740 (simplify_binary_operation_1): Extend CONST_INT handling to
1742 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1743 than a HOST_WIDE_INT.
1744 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1746 (adjust_mems, add_stores): Update accodingly.
1747 (vt_canonicalize_addr): Track polynomial offsets.
1748 (emit_note_insn_var_location): Likewise.
1749 (vt_add_function_parameter): Likewise.
1750 (vt_initialize): Likewise.
1752 2018-06-12 Jeff Law <law@redhat.com>
1754 * config.gcc (alpha*-*-freebsd*): Remove.
1755 * config/alpha/freebsd.h: Remove.
1757 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1760 * spellcheck-tree.c (levenshtein_distance): Rename to...
1761 (get_edit_distance): ...this, and update for underlying renaming.
1762 * spellcheck-tree.h (levenshtein_distance): Rename to...
1763 (get_edit_distance): ...this.
1764 * spellcheck.c (levenshtein_distance): Rename to...
1765 (get_edit_distance): ...this. Convert from Levenshtein distance
1766 to Damerau-Levenshtein distance by supporting transpositions of
1767 adjacent characters. Rename "v1" to "v_next" and "v0" to
1769 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1770 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1771 update for underlying renaming.
1772 (selftest::levenshtein_distance_unit_test): Rename to...
1773 (selftest::test_get_edit_distance_unit): ...this, and update for
1774 underlying renaming.
1775 (selftest::test_find_closest_string): Add example from PR 69968
1776 where transposition helps
1777 (selftest::test_metric_conditions): Update for renaming.
1778 (selftest::test_metric_conditions): Likewise.
1779 (selftest::spellcheck_c_tests): Likewise.
1780 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1781 (get_edit_distance): ...this.
1782 (best_match::consider): Update for renaming.
1784 2018-06-12 Martin Sebor <msebor@redhat.com>
1786 PR tree-optimization/85259
1787 * builtins.c (compute_objsize): Handle constant offsets.
1788 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1789 true iff a warning has been issued.
1790 * gimple.h (gimple_nonartificial_location): New function.
1791 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1792 gimple_nonartificial_location and handle -Wno-system-headers.
1793 (handle_builtin_stxncpy): Same.
1795 2018-06-12 Martin Sebor <msebor@redhat.com>
1798 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1800 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1802 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1803 BUILTIN_VEC_XST entries for pointer to double and long long.
1805 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1808 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1810 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1813 2018-06-12 Martin Liska <mliska@suse.cz>
1815 * doc/options.texi: Document IntegerRange.
1817 2018-06-12 Martin Liska <mliska@suse.cz>
1819 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1820 * opt-functions.awk: Handle Deprecated flag.
1821 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1823 (read_cmdline_option): Report warning for a deprecated option.
1824 * opts.h (struct cl_option): Add new field cl_deprecated.
1825 (CL_ERR_DEPRECATED): New.
1827 2018-06-12 Martin Liska <mliska@suse.cz>
1829 * doc/options.texi: Document Deprecated option flag.
1831 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1833 * config/arc/arc-arch.h (arc_extras): New enum.
1834 (arc_cpu_t):Add field extra.
1835 (arc_cpu_types): Consider the extras.
1836 * config/arc/arc-cpus.def: Add extras info.
1837 * config/arc/arc-opts.h (processor_type): Consider extra field.
1838 * config/arc/arc.c (arc_override_options): Handle extra field.
1840 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1842 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1843 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1845 * config/arc/arc.c (arc_init): Update pic warning.
1846 (irq_range): Update irq range parsing warnings.
1847 (arc_override_options): Update various warning messages.
1848 (arc_handle_aux_attribute): Likewise.
1850 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1852 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1854 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1856 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1857 compiler messages directives.
1859 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1861 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1862 * config/mips/mips-tables.opt: Regenerate.
1863 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1865 * doc/invoke.texi: Document -march=i6500.
1867 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1869 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1870 (i6400_gpmul): Add cpu_unit.
1871 (i6400_gpdiv): Likewise.
1872 (i6400_msa_add_d): Update reservations.
1873 (i6400_msa_int_add) Likewise.
1874 (i6400_msa_short_logic3) Likewise.
1875 (i6400_msa_short_logic2) Likewise.
1876 (i6400_msa_short_logic) Likewise.
1877 (i6400_msa_move) Likewise.
1878 (i6400_msa_cmp) Likewise.
1879 (i6400_msa_short_float2) Likewise.
1880 (i6400_msa_div_d) Likewise.
1881 (i6400_msa_long_logic1) Likewise.
1882 (i6400_msa_long_logic2) Likewise.
1883 (i6400_msa_mult) Likewise.
1884 (i6400_msa_long_float2) Likewise.
1885 (i6400_msa_long_float4) Likewise.
1886 (i6400_msa_long_float5) Likewise.
1887 (i6400_msa_long_float8) Likewise.
1888 (i6400_fpu_fadd): Include frint type.
1889 (i6400_fpu_store): New define_insn_reservation.
1890 (i6400_fpu_load): Likewise.
1891 (i6400_fpu_move): Likewise.
1892 (i6400_fpu_fcmp): Likewise.
1893 (i6400_fpu_fmadd): Likewise.
1894 (i6400_int_mult): Include imul3nc type and update reservation.
1895 (i6400_int_div): Include idiv3 type and update reservation.
1896 (i6400_int_load): Update to check type not move_type.
1897 (i6400_int_store): Likewise.
1898 (i6400_int_prefetch): Set zero latency.
1900 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1902 * gcc.c: Document new %@{...} sequence.
1903 (LINK_COMMAND_SPEC): Use it for the -L switches.
1904 (cpp_unique_options): Use it for the -I switches.
1905 (at_file_argbuf): New global variable.
1906 (in_at_file): Likewise.
1907 (alloc_args): Create at_file_argbuf.
1908 (clear_args): Truncate at_file_argbuf.
1909 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1910 (open_at_file): New function.
1911 (close_at_file): Likewise.
1912 (create_at_file): Delete.
1913 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1916 (validate_switches_from_spec): Deal with %@{...} sequence.
1917 (validate_switches): Likewise.
1918 (driver::finalize): Call clear_args.
1920 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1922 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1924 2018-06-11 Martin Sebor <msebor@redhat.com>
1926 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1927 (Wno-class-memaccess): Add @opindex.
1928 (Wno-templates, Wno-multiple-inheritance): Same.
1929 (Wno-virtual-inheritance, Wno-namespaces): Same.
1930 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1931 (Wno-format-overflow, Wno-format-truncation): Same.
1932 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1933 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1934 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1935 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1936 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1937 (Wno-misspelled-isr): Same.
1939 2018-06-11 Martin Sebor <msebor@redhat.com>
1941 * PR tree-optimization/86083
1942 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1944 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1946 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1948 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1951 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1952 on the correct operand.
1953 (*movdi_internal64): Ditto.
1955 2018-06-11 Martin Liska <mliska@suse.cz>
1957 PR tree-optimization/86089
1958 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1960 2018-06-11 Julia Koval <julia.koval@intel.com>
1962 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1963 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1964 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1966 2018-06-11 Olivier Hainque <hainque@adacore.com>
1968 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1969 for Ada with strict dwarf2.
1971 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1974 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1977 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1979 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1981 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1983 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1986 2018-06-08 Martin Liska <mliska@suse.cz>
1988 * tree-cfg.h (debug_function): Fix argument type to match
1991 2018-06-08 Martin Liska <mliska@suse.cz>
1993 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1994 Remove usage of MPX-related (and removed) fields.
1995 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1997 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1999 * cfg.c (debug): Use TDF_NONE rather than 0.
2000 * cfghooks.c (debug): Likewise.
2001 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2002 (struct dump_option_value_info): Convert to...
2003 (struct kv_pair): ...this template type.
2004 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2006 (optinfo_verbosity_options): Likewise.
2007 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2009 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2010 than int for "optgroup_flags" param.
2011 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2013 (dump_dec): Likewise.
2014 (dump_finish): Use TDF_NONE rather than 0.
2015 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2016 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2017 than 0. Update for change to option_ptr.
2018 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2019 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2020 0. Update for changes to optinfo_verbosity_options and
2022 (opt_info_switch_p): Convert optgroup_flags from int to
2024 (dump_basic_block): Use dump_flags_t rather than int
2025 for "dump_kind" param.
2026 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2027 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2028 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2029 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2030 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2031 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2032 TDF_NONE): Convert from macros to...
2033 (enum dump_flag): ...this new enum.
2034 (dump_flags_t): Update to use enum.
2035 (operator|, operator&, operator~, operator|=, operator&=):
2036 Implement for dump_flags_t.
2037 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2038 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2039 Convert from macros to...
2040 (enum optgroup_flag): ...this new enum.
2041 (optgroup_flags_t): New typedef.
2042 (operator|, operator|=): Implement for optgroup_flags_t.
2043 (struct dump_file_info): Convert field "alt_flags" to
2044 dump_flags_t. Convert field "optgroup_flags" to
2046 (dump_basic_block): Use dump_flags_t rather than int for param.
2047 (dump_generic_expr_loc): Likewise.
2048 (dump_dec): Likewise.
2049 (dump_register): Convert param "optgroup_flags" to
2051 (opt_info_enable_passes): Likewise.
2052 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2054 * gimple-pretty-print.c (debug): Likewise.
2055 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2056 (merged_store_group::apply_stores): Likewise.
2057 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2058 * gimple.c (verify_gimple_pp): Likewise.
2059 * graphite-poly.c (print_pbb_body): Likewise.
2060 * passes.c (pass_manager::register_one_dump_file): Convert
2061 local "optgroup_flags" to optgroup_flags_t.
2062 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2064 (debug_body): Likewise.
2065 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2066 to optgroup_flags_t.
2067 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2069 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2070 (convert_mult_to_fma): Likewise.
2071 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2072 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2073 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2074 "dump_kind" to dump_flags_t.
2076 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2078 * config/rs6000/rs6000.c (min, max): Delete.
2080 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2082 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2083 -mabi=spe and -mabi=no-spe.
2085 2018-06-08 Martin Liska <mliska@suse.cz>
2087 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2088 where we expect an existing summary.
2090 2018-06-08 Martin Liska <mliska@suse.cz>
2092 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2093 * ipa-inline.h (estimate_edge_growth): Likewise.
2095 2018-06-08 Martin Liska <mliska@suse.cz>
2097 * cgraph.c (function_version_hasher::hash): Use
2098 cgraph_node::get_uid ().
2099 (function_version_hasher::equal):
2100 * cgraph.h (cgraph_node::get_uid): New method.
2101 * ipa-inline.c (update_caller_keys): Use
2102 cgraph_node::get_uid ().
2103 (update_callee_keys): Likewise.
2104 * ipa-utils.c (searchc): Likewise.
2105 (ipa_reduced_postorder): Likewise.
2106 * lto-cgraph.c (input_node): Likewise.
2107 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2108 * symbol-summary.h (symtab_insertion): Likewise.
2109 (symtab_removal): Likewise.
2110 (symtab_duplication): Likewise.
2111 * tree-pretty-print.c (dump_function_header): Likewise.
2112 * tree-sra.c (convert_callers_for_node): Likewise.
2114 2018-06-08 Martin Liska <mliska@suse.cz>
2116 * cgraph.c (symbol_table::create_edge): Always assign a new
2118 (symbol_table::free_edge): Do not recycle numbers.
2119 * cgraph.h (cgraph_edge::get): New method.
2120 * symbol-summary.h (symtab_removal): Use it.
2121 (symtab_duplication): Likewise.
2122 (call_summary::hashable_uid): Remove.
2124 2018-06-08 Martin Liska <mliska@suse.cz>
2126 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2127 (initialize_growth_caches): Remove.
2128 (free_growth_caches): Likewise.
2129 (do_estimate_edge_time): Use edge_growth_cache.
2130 (do_estimate_edge_size): Likewise.
2131 (do_estimate_edge_hints): Likewise.
2132 * ipa-inline.c (reset_edge_caches): Likewise.
2133 (recursive_inlining): Likewise.
2134 (inline_small_functions): Likewise.
2135 * ipa-inline.h (initialize_growth_caches): Remove.
2136 (estimate_edge_size): Likewise.
2137 (estimate_edge_time): Likewise.
2138 (estimate_edge_hints): Likewise.
2139 (reset_edge_growth_cache): Likewise.
2140 * symbol-summary.h (call_summary::remove): New method.
2142 2018-06-08 Martin Liska <mliska@suse.cz>
2144 * ipa-cp.c (class edge_clone_summary): New summary.
2145 (grow_edge_clone_vectors): Remove.
2146 (ipcp_edge_duplication_hook): Remove.
2147 (class edge_clone_summary_t): New call_summary class.
2148 (ipcp_edge_removal_hook): Remove.
2149 (edge_clone_summary_t::duplicate): New function.
2150 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2151 (create_specialized_node): Likewise.
2152 (ipcp_driver): Initialize edge_clone_summaries and do not
2155 2018-06-08 Martin Liska <mliska@suse.cz>
2157 * symbol-summary.h (get): New function.
2158 (call_summary::m_initialize_when_cloning): New class member.
2160 2018-06-08 Martin Liska <mliska@suse.cz>
2162 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2163 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2164 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2165 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2166 (remove_cgraph_node_from_order): Use the removed_nodes set.
2167 (do_per_function_toporder): Likwise.
2168 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2169 instead of summary_uid.
2170 (symtab_removal): Likewise.
2171 (symtab_duplication): Likewise.
2173 2018-06-08 Martin Liska <mliska@suse.cz>
2175 * ipa-cp.c (ipcp_store_bits_results): Use
2176 ipcp_transformation_sum.
2177 (ipcp_store_vr_results): Likewise.
2178 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2180 (ipcp_transformation_initialize): ... this.
2181 (ipa_set_node_agg_value_chain):
2182 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2183 (write_ipcp_transformation_info): Likewise.
2184 (read_ipcp_transformation_info): Likewise.
2185 (ipcp_update_bits): Likewise.
2186 (ipcp_update_vr): Likewise.
2187 (ipcp_transform_function): Likewise.
2188 * ipa-prop.h: Rename ipcp_transformation_summary to
2189 ipcp_transformation.
2190 (class ipcp_transformation_t): New function summary.
2191 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2192 (ipa_get_agg_replacements_for_node): Likewise.
2194 2018-06-08 Martin Liska <mliska@suse.cz>
2196 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2198 (class funct_state_summary_t): New function_summary class.
2199 (has_function_state): Remove.
2200 (get_function_state): Likewise.
2201 (set_function_state): Likewise.
2202 (add_new_function): Likewise.
2203 (funct_state_summary_t::insert): New function.
2204 (duplicate_node_data): Remove.
2205 (remove_node_data): Remove.
2206 (funct_state_summary_t::duplicate): New function.
2207 (register_hooks): Create new funct_state_summaries.
2208 (pure_const_generate_summary): Use it.
2209 (pure_const_write_summary): Likewise.
2210 (pure_const_read_summary): Likewise.
2211 (propagate_pure_const): Likewise.
2212 (propagate_nothrow): Likewise.
2213 (dump_malloc_lattice): Likewise.
2214 (propagate_malloc): Likewise.
2215 (execute): Do not register hooks, just remove summary
2217 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2220 2018-06-08 Martin Liska <mliska@suse.cz>
2222 * ipa-reference.c (remove_node_data): Remove.
2223 (duplicate_node_data): Likewise.
2224 (class ipa_ref_var_info_summary_t): New class.
2225 (class ipa_ref_opt_summary_t): Likewise.
2226 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2227 (get_reference_optimization_summary): Use
2228 ipa_ref_opt_sum_summaries.
2229 (set_reference_vars_info): Remove.
2230 (set_reference_optimization_summary): Likewise.
2231 (ipa_init): Create summaries.
2232 (init_function_info): Use function summary.
2233 (ipa_ref_opt_summary_t::duplicate): New function.
2234 (ipa_ref_opt_summary_t::remove): New function.
2235 (get_read_write_all_from_node): Fix GNU coding style.
2236 (propagate): Use function summary.
2237 (write_node_summary_p): Fix GNU coding style.
2238 (stream_out_bitmap): Likewise.
2239 (ipa_reference_read_optimization_summary): Use function summary.
2240 (ipa_reference_c_finalize): Do not release hooks.
2242 2018-06-08 Martin Liska <mliska@suse.cz>
2244 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2245 (analyze_function_body): Extract multiple calls of get_create.
2246 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2247 * ipa-inline.c (recursive_inlining): Use ::get method.
2248 * ipa-inline.h (estimate_edge_growth): Likewise.
2250 2018-06-08 Martin Liska <mliska@suse.cz>
2252 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2254 (hsa_function_summary::hsa_function_summary): Use the new enum
2256 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2257 * hsa-gen.c (hsa_get_host_function): Likewise.
2258 (get_brig_function_name): Likewise.
2259 * ipa-hsa.c (process_hsa_functions): Likewise.
2260 (ipa_hsa_write_summary): Likewise.
2261 * symbol-summary.h (symtab_duplication): Use ::get function/
2262 (get): New function.
2264 2018-06-08 Martin Liska <mliska@suse.cz>
2266 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2268 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2269 (hsa_register_kernel): Likewise.
2270 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2271 * hsa-gen.c (hsa_get_host_function): Likewise.
2272 (get_brig_function_name): Likewise.
2273 (generate_hsa): Likewise.
2274 (pass_gen_hsail::execute): Likewise.
2275 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2276 (devirtualization_time_bonus): Likewise.
2277 (ipcp_propagate_stage): Likewise.
2278 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2279 (edge_set_predicate): Likewise.
2280 (evaluate_conditions_for_known_args): Likewise.
2281 (evaluate_properties_for_edge): Likewise.
2282 (ipa_fn_summary::reset): Likewise.
2283 (ipa_fn_summary_t::duplicate): Likewise.
2284 (dump_ipa_call_summary): Likewise.
2285 (ipa_dump_fn_summary): Likewise.
2286 (analyze_function_body): Likewise.
2287 (compute_fn_summary): Likewise.
2288 (estimate_edge_devirt_benefit): Likewise.
2289 (estimate_edge_size_and_time): Likewise.
2290 (estimate_calls_size_and_time): Likewise.
2291 (estimate_node_size_and_time): Likewise.
2292 (inline_update_callee_summaries): Likewise.
2293 (remap_edge_change_prob): Likewise.
2294 (remap_edge_summaries): Likewise.
2295 (ipa_merge_fn_summary_after_inlining): Likewise.
2296 (ipa_update_overall_fn_summary): Likewise.
2297 (read_ipa_call_summary): Likewise.
2298 (inline_read_section): Likewise.
2299 (write_ipa_call_summary): Likewise.
2300 (ipa_fn_summary_write): Likewise.
2301 (ipa_free_fn_summary): Likewise.
2302 * ipa-hsa.c (process_hsa_functions): Likewise.
2303 (ipa_hsa_write_summary): Likewise.
2304 (ipa_hsa_read_section): Likewise.
2305 * ipa-icf.c (sem_function::merge): Likewise.
2306 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2307 (do_estimate_edge_time): Likewise.
2308 (estimate_size_after_inlining): Likewise.
2309 (estimate_growth): Likewise.
2310 (growth_likely_positive): Likewise.
2311 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2312 (inline_call): Likewise.
2313 * ipa-inline.c (caller_growth_limits): Likewise.
2314 (can_inline_edge_p): Likewise.
2315 (can_inline_edge_by_limits_p): Likewise.
2316 (compute_uninlined_call_time): Likewise.
2317 (compute_inlined_call_time): Likewise.
2318 (want_inline_small_function_p): Likewise.
2319 (edge_badness): Likewise.
2320 (update_caller_keys): Likewise.
2321 (update_callee_keys): Likewise.
2322 (recursive_inlining): Likewise.
2323 (inline_small_functions): Likewise.
2324 (inline_to_all_callers_1): Likewise.
2325 (dump_overall_stats): Likewise.
2326 (early_inline_small_functions): Likewise.
2327 (early_inliner): Likewise.
2328 * ipa-inline.h (estimate_edge_growth): Likewise.
2329 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2330 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2331 * ipa-prop.h (IPA_NODE_REF): Likewise.
2332 (IPA_EDGE_REF): Likewise.
2333 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2334 (propagate_malloc): Likewise.
2335 * ipa-split.c (execute_split_functions): Likewise.
2336 * symbol-summary.h: Rename get to get_create.
2338 (get_create): Likewise.
2339 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2341 2018-06-08 Martin Liska <mliska@suse.cz>
2343 * symbol-summary.h (release): Move definition out of class
2345 (symtab_removal): Likewise.
2346 (symtab_duplication): Likewise.
2348 2018-06-08 Martin Liska <mliska@suse.cz>
2350 * symbol-summary.h (function_summary): Move constructor
2351 implementation out of class declaration.
2352 (release): Likewise.
2353 (symtab_insertion): Likewise.
2354 (symtab_removal): Likewise.
2355 (symtab_duplication): Likewise.
2358 2018-06-08 Martin Liska <mliska@suse.cz>
2360 * Makefile.in: Remove support for MPX (macros, related functions,
2361 fields in cgraph_node, ...).
2362 * builtin-types.def (BT_BND): Likewise.
2363 (BT_FN_BND_CONST_PTR): Likewise.
2364 (BT_FN_CONST_PTR_BND): Likewise.
2365 (BT_FN_VOID_PTR_BND): Likewise.
2366 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2367 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2368 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2369 (expand_builtin_mempcpy_with_bounds): Likewise.
2370 (expand_builtin_memset_with_bounds): Likewise.
2371 (expand_builtin_memset_args): Likewise.
2372 (std_expand_builtin_va_start): Likewise.
2373 (expand_builtin): Likewise.
2374 (expand_builtin_with_bounds): Likewise.
2375 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2376 (DEF_LIB_BUILTIN_CHKP): Likewise.
2377 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2378 (DEF_CHKP_BUILTIN): Likewise.
2379 (BUILT_IN_MEMCPY): Likewise.
2380 (BUILT_IN_MEMMOVE): Likewise.
2381 (BUILT_IN_MEMPCPY): Likewise.
2382 (BUILT_IN_MEMSET): Likewise.
2383 (BUILT_IN_STPCPY): Likewise.
2384 (BUILT_IN_STRCAT): Likewise.
2385 (BUILT_IN_STRCHR): Likewise.
2386 (BUILT_IN_STRCPY): Likewise.
2387 (BUILT_IN_STRLEN): Likewise.
2388 (BUILT_IN_MEMCPY_CHK): Likewise.
2389 (BUILT_IN_MEMMOVE_CHK): Likewise.
2390 (BUILT_IN_MEMPCPY_CHK): Likewise.
2391 (BUILT_IN_MEMSET_CHK): Likewise.
2392 (BUILT_IN_STPCPY_CHK): Likewise.
2393 (BUILT_IN_STRCAT_CHK): Likewise.
2394 (BUILT_IN_STRCPY_CHK): Likewise.
2395 * calls.c (store_bounds): Likewise.
2396 (emit_call_1): Likewise.
2397 (special_function_p): Likewise.
2398 (maybe_warn_nonstring_arg): Likewise.
2399 (initialize_argument_information): Likewise.
2400 (finalize_must_preallocate): Likewise.
2401 (compute_argument_addresses): Likewise.
2402 (expand_call): Likewise.
2403 * cfgexpand.c (expand_call_stmt): Likewise.
2404 (expand_return): Likewise.
2405 (expand_gimple_stmt_1): Likewise.
2406 (pass_expand::execute): Likewise.
2407 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2408 (cgraph_node::remove): Likewise.
2409 (cgraph_node::dump): Likewise.
2410 (cgraph_node::verify_node): Likewise.
2411 * cgraph.h (chkp_function_instrumented_p): Likewise.
2412 (symtab_node::get_alias_target): Likewise.
2413 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2414 (cgraph_local_p): Likewise.
2415 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2416 (cgraph_edge::rebuild_references): Likewise.
2417 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2418 (walk_polymorphic_call_targets): Likewise.
2419 (cgraph_node::expand_thunk): Likewise.
2420 (symbol_table::output_weakrefs): Likewise.
2421 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2422 (ix86_handle_option): Likewise.
2423 * config/i386/constraints.md: Likewise.
2424 * config/i386/i386-builtin-types.def (BND): Likewise.
2428 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2429 (BDESC_FIRST): Likewise.
2431 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2432 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2433 * config/i386/i386.c (enum reg_class): Likewise.
2434 (ix86_target_string): Likewise.
2435 (ix86_option_override_internal): Likewise.
2436 (ix86_conditional_register_usage): Likewise.
2437 (ix86_valid_target_attribute_inner_p): Likewise.
2438 (ix86_set_indirect_branch_type): Likewise.
2439 (ix86_set_current_function): Likewise.
2440 (ix86_function_arg_regno_p): Likewise.
2441 (init_cumulative_args): Likewise.
2442 (ix86_function_arg_advance): Likewise.
2443 (ix86_function_arg): Likewise.
2444 (ix86_pass_by_reference): Likewise.
2445 (ix86_function_value_regno_p): Likewise.
2446 (ix86_function_value_1): Likewise.
2447 (ix86_function_value_bounds): Likewise.
2448 (ix86_return_in_memory): Likewise.
2449 (ix86_setup_incoming_vararg_bounds): Likewise.
2450 (ix86_va_start): Likewise.
2451 (indirect_thunk_need_prefix): Likewise.
2452 (print_reg): Likewise.
2453 (ix86_print_operand): Likewise.
2454 (ix86_expand_call): Likewise.
2455 (ix86_output_function_return): Likewise.
2456 (reg_encoded_number): Likewise.
2457 (BDESC_VERIFYS): Likewise.
2458 (ix86_init_mpx_builtins): Likewise.
2459 (ix86_init_builtins): Likewise.
2460 (ix86_emit_cmove): Likewise.
2461 (ix86_emit_move_max): Likewise.
2462 (ix86_expand_builtin): Likewise.
2463 (ix86_builtin_mpx_function): Likewise.
2464 (ix86_get_arg_address_for_bt): Likewise.
2465 (ix86_load_bounds): Likewise.
2466 (ix86_store_bounds): Likewise.
2467 (ix86_load_returned_bounds): Likewise.
2468 (ix86_store_returned_bounds): Likewise.
2469 (ix86_class_likely_spilled_p): Likewise.
2470 (ix86_hard_regno_mode_ok): Likewise.
2471 (x86_order_regs_for_local_alloc): Likewise.
2472 (ix86_mitigate_rop): Likewise.
2473 (ix86_bnd_prefixed_insn_p): Likewise.
2474 (ix86_mpx_bound_mode): Likewise.
2475 (ix86_make_bounds_constant): Likewise.
2476 (ix86_initialize_bounds): Likewise.
2477 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2478 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2479 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2480 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2481 (TARGET_CHKP_BOUND_MODE): Likewise.
2482 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2483 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2484 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2485 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2486 * config/i386/i386.h (TARGET_MPX): Likewise.
2487 (TARGET_MPX_P): Likewise.
2488 (VALID_BND_REG_MODE): Likewise.
2489 (FIRST_BND_REG): Likewise.
2490 (LAST_BND_REG): Likewise.
2491 (enum reg_class): Likewise.
2492 (BND_REG_P): Likewise.
2493 (BND_REGNO_P): Likewise.
2494 (BNDmode): Likewise.
2495 (ADJUST_INSN_LENGTH): Likewise.
2496 * config/i386/i386.md: Likewise.
2497 * config/i386/i386.opt: Likewise.
2498 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2499 (defined): Likewise.
2500 (LINK_MPX): Likewise.
2501 (MPX_SPEC): Likewise.
2502 (LIBMPX_SPEC): Likewise.
2503 (LIBMPXWRAPPERS_SPEC): Likewise.
2504 (CHKP_SPEC): Likewise.
2505 * config/i386/predicates.md: Likewise.
2506 * dbxout.c (dbxout_type): Likewise.
2507 * doc/extend.texi: Likewise.
2508 * doc/invoke.texi: Likewise.
2509 * doc/md.texi: Likewise.
2510 * doc/tm.texi: Likewise.
2511 * doc/tm.texi.in: Likewise.
2512 * dwarf2out.c (is_base_type): Likewise.
2513 (gen_formal_types_die): Likewise.
2514 (gen_subprogram_die): Likewise.
2515 (gen_type_die_with_usage): Likewise.
2516 (gen_decl_die): Likewise.
2517 (dwarf2out_late_global_decl): Likewise.
2518 * expr.c (expand_assignment): Likewise.
2519 (emit_storent_insn): Likewise.
2520 (store_expr_with_bounds): Likewise.
2521 (store_expr): Likewise.
2522 (expand_expr_real_1): Likewise.
2523 * expr.h (store_expr_with_bounds): Likewise.
2524 * function.c (use_register_for_decl): Likewise.
2525 (struct bounds_parm_data): Likewise.
2526 (assign_parms_augmented_arg_list): Likewise.
2527 (assign_parm_find_entry_rtl): Likewise.
2528 (assign_parm_is_stack_parm): Likewise.
2529 (assign_parm_load_bounds): Likewise.
2530 (assign_bounds): Likewise.
2531 (assign_parms): Likewise.
2532 (expand_function_start): Likewise.
2533 * gcc.c (CHKP_SPEC): Likewise.
2534 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2535 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2536 (wrestrict_dom_walker::check_call): Likewise.
2537 * gimple.c (gimple_build_call_from_tree): Likewise.
2538 * gimple.h (enum gf_mask): Likewise.
2539 (gimple_call_with_bounds_p): Likewise.
2540 (gimple_call_set_with_bounds): Likewise.
2541 * gimplify.c (gimplify_init_constructor): Likewise.
2542 * ipa-cp.c (initialize_node_lattices): Likewise.
2543 (propagate_constants_across_call): Likewise.
2544 (find_more_scalar_values_for_callers_subset): Likewise.
2545 * ipa-hsa.c (process_hsa_functions): Likewise.
2546 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2547 * ipa-icf.c (sem_function::merge): Likewise.
2548 * ipa-inline.c (early_inliner): Likewise.
2549 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2550 (warn_function_cold): Likewise.
2551 (propagate_pure_const): Likewise.
2552 * ipa-ref.h (enum GTY): Likewise.
2553 * ipa-split.c (find_retbnd): Likewise.
2554 (consider_split): Likewise.
2555 (split_function): Likewise.
2556 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2557 * ipa.c (walk_polymorphic_call_targets): Likewise.
2558 (symbol_table::remove_unreachable_nodes): Likewise.
2559 (process_references): Likewise.
2560 (cgraph_build_static_cdtor_1): Likewise.
2561 * lto-cgraph.c (lto_output_node): Likewise.
2562 (output_refs): Likewise.
2563 (compute_ltrans_boundary): Likewise.
2564 (input_overwrite_node): Likewise.
2565 (input_node): Likewise.
2566 (input_cgraph_1): Likewise.
2567 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2568 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2569 (class pass_chkp_instrumentation_passes): Likewise.
2570 (make_pass_chkp_instrumentation_passes): Likewise.
2571 * passes.def: Likewise.
2572 * rtl.h (struct GTY): Likewise.
2573 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2574 * stor-layout.c (layout_type): Likewise.
2575 * symtab.c: Likewise.
2576 * target.def: Likewise.
2577 * targhooks.c (default_chkp_bound_type): Likewise.
2578 (default_chkp_bound_mode): Likewise.
2579 (default_builtin_chkp_function): Likewise.
2580 (default_chkp_function_value_bounds): Likewise.
2581 (default_chkp_make_bounds_constant): Likewise.
2582 (default_chkp_initialize_bounds): Likewise.
2583 * targhooks.h (default_chkp_bound_type): Likewise.
2584 (default_chkp_bound_mode): Likewise.
2585 (default_builtin_chkp_function): Likewise.
2586 (default_chkp_function_value_bounds): Likewise.
2587 (default_chkp_make_bounds_constant): Likewise.
2588 (default_chkp_initialize_bounds): Likewise.
2589 * toplev.c (compile_file): Likewise.
2590 (process_options): Likewise.
2591 * tree-core.h (DEF_BUILTIN): Likewise.
2592 (DEF_BUILTIN_CHKP): Likewise.
2593 * tree-inline.c (declare_return_variable): Likewise.
2594 (remap_gimple_stmt): Likewise.
2595 (copy_bb): Likewise.
2596 (initialize_inlined_parameters): Likewise.
2597 (expand_call_inline): Likewise.
2598 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2599 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2600 (make_pass_ipa_chkp_produce_thunks): Likewise.
2601 (make_pass_chkp): Likewise.
2602 (make_pass_chkp_opt): Likewise.
2603 (make_pass_chkp_instrumentation_passes): Likewise.
2604 * tree-pretty-print.c (dump_generic_node): Likewise.
2605 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2606 * tree-ssa-dce.c (propagate_necessity): Likewise.
2607 (eliminate_unnecessary_stmts): Likewise.
2608 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2609 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2610 * tree-ssa-sccvn.h: Likewise.
2611 * tree-ssa-strlen.c (get_string_length): Likewise.
2612 (valid_builtin_call): Likewise.
2613 (adjust_last_stmt): Likewise.
2614 (handle_builtin_strchr): Likewise.
2615 (handle_builtin_strcpy): Likewise.
2616 (handle_builtin_stxncpy): Likewise.
2617 (handle_builtin_memcpy): Likewise.
2618 (handle_builtin_strcat): Likewise.
2619 (strlen_check_and_optimize_stmt): Likewise.
2620 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2621 * tree-streamer-in.c: Likewise.
2622 * tree-streamer.c (record_common_node): Likewise.
2623 * tree.c (tree_code_size): Likewise.
2624 (wide_int_to_tree_1): Likewise.
2625 (type_contains_placeholder_1): Likewise.
2626 (build_common_tree_nodes): Likewise.
2627 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2628 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2629 (POINTER_BOUNDS_P): Likewise.
2630 (BOUNDED_TYPE_P): Likewise.
2631 (BOUNDED_P): Likewise.
2632 (CALL_WITH_BOUNDS_P): Likewise.
2633 (pointer_bounds_type_node): Likewise.
2634 * value-prof.c (gimple_ic): Likewise.
2635 * var-tracking.c (vt_add_function_parameters): Likewise.
2636 * varasm.c (make_decl_rtl): Likewise.
2637 (assemble_start_function): Likewise.
2638 (output_constant): Likewise.
2639 (maybe_assemble_visibility): Likewise.
2640 * varpool.c (ctor_for_folding): Likewise.
2641 * chkp-builtins.def: Remove.
2642 * ipa-chkp.c: Remove.
2643 * ipa-chkp.h: Remove.
2644 * rtl-chkp.c: Remove.
2645 * rtl-chkp.h: Remove.
2646 * tree-chkp-opt.c: Remove.
2647 * tree-chkp.c: Remove.
2648 * tree-chkp.h: Remove.
2650 2018-06-07 Carl Love <cel@us.ibm.com>
2652 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2653 vextract_fp_from_shortl): Add BE support.
2655 2018-06-07 Paul Koning <ni1d@arrl.net>
2657 * compare-elim.c (try_merge_compare): Don't merge compare if
2658 address contains a side effect.
2659 (try_eliminate_compare): Likewise.
2661 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2663 * config.gcc: Support "tremont".
2664 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2665 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2667 * config/i386/i386.c (m_TREMONT): Define.
2668 (processor_target_table): Add "tremont".
2669 (PTA_TREMONT): Define.
2670 (ix86_lea_outperforms): Add TARGET_TREMONT.
2671 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2672 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2673 and M_INTEL_GOLDMONT_PLUS.
2674 (fold_builtin_cpu): Add "tremont".
2675 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2676 (ix86_option_override_internal): Add "tremont".
2677 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2678 (processor_type): Add PROCESSOR_TREMONT.
2679 * config/i386/x86-tune.def: Add m_TREMONT.
2680 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2682 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2684 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2685 symbol defined for msp430i* devices to be lower case.
2687 gcc/testsuite/gcc.target/msp430/
2688 * msp430i-device-symbol.c: New test.
2689 * msp430f-device-symbol.c: New test.
2690 * msp430.h: New test header file.
2692 2018-06-07 Richard Biener <rguenther@suse.de>
2694 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2695 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2696 Properly wrap signed arithmetic if overflow wraps.
2698 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2700 PR tree-optimization/69615
2701 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2702 of a cast from a same precision integral SSA_NAME in a bb dominated
2703 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2704 cast to utype if rhs2 has already a compatible type.
2706 2018-06-07 Richard Biener <rguenther@suse.de>
2708 PR tree-optimization/85935
2709 * graphite-scop-detection.c (find_params_in_bb): Analyze
2710 condition operands with respect to the correct loop. Assert
2711 the analysis doesn't fail.
2713 2018-06-04 Carl Love <cel@us.ibm.com>
2715 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2716 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2717 as it is slightly cheaper.
2718 (first_match_or_eos_index_<mode>):
2719 Calculate index using natural element order.
2720 (first_match_index_<mode>):
2721 Calculate index using natural element order.
2722 (first_match_or_eos_index_<mode>):
2723 Calculate index using natural order.
2724 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2725 for BE and LE modes.
2726 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2727 P9V_BUILTIN_VCLZLSBB_V16QI.
2728 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2731 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2733 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2734 indentation and line wrap for many prototypes. Add missing
2735 @smallexample directives around block of prototypes for vec_xl and
2738 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2740 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2741 track if we pass or return IEEE 128-bit floating point.
2742 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2743 C++ mangling that is compatible with GCC 8.1.
2744 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2745 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2746 floating point types.
2747 (rs6000_function_arg_advance_1): Likewise.
2748 (rs6000_mangle_type): Optionally generate mangled names that match
2749 what GCC 8.1 generated for IEEE 128-bit floating point types.
2750 (rs6000_globalize_decl_name): If we have an external function that
2751 passes or returns IEEE 128-bit types, generate a weak reference
2752 from the mangled name used in GCC 8.1 to the current mangled
2754 (rs6000_init_builtins): Make __ibm128 use the long double type if
2755 long double is IBM extended double. Make __float128 use the long
2756 double type if long double is IEEE 128-bit.
2759 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2760 macro for __ibm128 built-in functions.
2761 (PACK_IF): Add __ibm128 pack/unpack functions.
2762 (UNPACK_IF): Likewise.
2763 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2764 enable long double built-in functions if long double is IEEE
2765 128-bit floating point.
2766 (rs6000_invalid_builtin): Update long double built-in function
2768 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2769 functions, adjust the built-in function to use the long double
2770 built-in function if __ibm128 and long double are the same type.
2771 * doc/extend.texi (PowerPC builtins): Update documention for
2772 __builtin_{,un}pack_longdouble. Add documentation for
2773 __builtin_{,un}pack_ibm128.
2775 2018-06-06 Jim Wilson <jimw@sifive.com>
2777 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2778 (struct machine_function): New field interrupt_mode.
2779 (riscv_handle_type_attribute): New function. Add forward declaration.
2780 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2781 (riscv_expand_epilogue): Check interrupt_mode field.
2782 (riscv_set_current_function): Check interrupt attribute args and
2783 set interrupt_mode field.
2784 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2785 (riscv_sret, riscv_uret): New.
2786 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2787 new arguments to interrupt attribute.
2789 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2792 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2793 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2795 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2797 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2798 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2799 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2800 several redundant entries.
2802 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2804 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2805 type from "rtx" to "rtx_insn *".
2806 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2807 for local "call_insn", removing cast.
2808 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2810 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2812 PR tree-optimization/86066
2813 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2814 for BIT_INSERT_EXPR stores.
2816 2018-06-06 Richard Biener <rguenther@suse.de>
2818 PR tree-optimization/86062
2819 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2820 component refs ontop
2821 of to be offsetted base.
2823 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2825 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2826 to be static and remove check on interrupt attribute name.
2828 gcc/testsuite/gcc.target/msp430/
2829 * function-attributes-4.c: New test.
2830 * static-interrupts.c: New test.
2832 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2834 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2835 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2837 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2840 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2842 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2843 Remove second argument, change how error is called.
2844 (aarch64_layout_arg): Remove second argument from
2845 aarch64_err_no_fpadvsimd call.
2846 (aarch64_init_cumulative_args): Ditto.
2847 (aarch64_gimplify_va_arg_expr): Ditto.
2848 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2850 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2852 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2853 (*simple_return_indirect_internal<mode>): Rename from
2854 simple_return_indirect_internal. Use W mode iterator.
2855 (rstorssp): New expander.
2856 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2857 (clrssbsy): New expander.
2858 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2860 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2862 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2864 (cmse_check_pointed_object): Likewise.
2866 2018-06-05 Martin Liska <mliska@suse.cz>
2868 PR gcov-profile/47618
2869 * doc/invoke.texi: Document how -fprofile-dir format
2872 2018-06-05 Richard Biener <rguenther@suse.de>
2874 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2875 removal pretend DOM info isn't available so we do not update
2876 it and only remove edges, not dominated blocks. Actually free
2877 DOM info in case we removed something. Remove unreachable blocks.
2878 (mfb_keep_latches): Work with either DOM info or marked backedges.
2879 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2880 first. Mark backedges if DOM info isn't available.
2881 (Re-)compute DOM info after cleanup_control_flow_pre.
2883 2018-06-05 Richard Biener <rguenther@suse.de>
2885 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2886 (locus_discrim_hasher::hash): Adjust.
2887 (locus_discrim_hasher::equal): Likewise.
2888 (next_discriminator_for_locus): Work on line directly.
2889 (same_line_p): Pass in expanded locus1 as well.
2890 (assign_discriminators): Avoid redundant location expansions.
2892 2018-06-05 Richard Biener <rguenther@suse.de>
2894 PR tree-optimization/86046
2895 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2896 if required after clearing TREE_ADDRESSABLE.
2898 2018-06-05 Richard Biener <rguenther@suse.de>
2900 PR tree-optimization/86047
2901 * tree-ssa-loop.c (for_each_index): Glob handling of all
2902 decls and constants and really handle all of them.
2904 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2907 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2908 qualifier_void_pointer and qualifier_const_void_pointer.
2909 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2910 (arm_init_builtins): Handle the above.
2911 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2912 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2913 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2916 2018-06-05 Martin Liska <mliska@suse.cz>
2918 * auto-profile.c (read_autofdo_file): Do not use
2919 gcov_ctr_summary struct.
2920 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2921 * coverage.c (struct counts_entry): Likewise.
2922 (read_counts_file): Read just single summary entry.
2923 (get_coverage_counts): Use gcov_summary struct.
2924 * coverage.h (get_coverage_counts): Likewise.
2925 * gcov-dump.c (dump_working_sets): Likewise.
2926 (tag_summary): Dump just single summary.
2927 * gcov-io.c (gcov_write_summary): Write just histogram
2929 (gcov_read_summary): Read just single summary.
2930 (compute_working_sets): Use gcov_summary struct.
2931 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2932 of GCOV_COUNTERS_SUMMABLE.
2933 (GCOV_COUNTERS_SUMMABLE): Remove.
2934 (GCOV_FIRST_VALUE_COUNTER): Replace with
2935 GCOV_COUNTER_V_INTERVAL.
2936 (struct gcov_ctr_summary): Remove.
2937 (struct gcov_summary): Directly use fields of former
2939 (compute_working_sets): Use gcov_summary struct.
2940 * gcov.c (read_count_file): Do not use ctrs fields.
2941 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2943 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2945 * profile.c: Likewise.
2946 * profile.h: Likewise.
2948 2018-06-05 Martin Liska <mliska@suse.cz>
2950 PR gcov-profile/84846
2951 * gcov.c (output_lines): Print working directory only
2952 in intermediate format.
2954 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2956 * config/s390/s390-builtin-types.def: Add void function type.
2957 * config/s390/s390-builtins.def: Use the function type for the
2960 2018-06-04 Jim Wilson <jimw@sifive.com>
2962 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2964 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2965 and maybe_eh_return. Change regno to unsigned int. Use new args to
2966 handle EH_RETURN_DATA_REGNO registers properly.
2967 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2968 (riscv_expand_epilogue): Update comment. Change argument name and
2969 type. Update code to use new name and type. Pass new args to
2970 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2972 * config/riscv/riscv.md (NORMAL_RETURN): New.
2973 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2974 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2975 (eh_return): Call gen_eh_return_internal and emit barrier.
2976 (eh_return_internal): Call riscv_expand_epilogue.
2978 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2980 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2981 bit_insertion field and declare can_be_merged_into method.
2982 (merged_store_group::can_be_merged_into): New method.
2983 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2984 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2985 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2987 2018-06-04 Richard Biener <rguenther@suse.de>
2989 PR tree-optimization/85955
2990 * builtins.c (fold_builtin_sincos): Convert pointers to
2991 destination to appropriate type before dereferencing.
2993 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2995 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2997 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2999 * expr.c (expand_expr_real_1): Force the operand into memory if
3000 its TYPE_MODE is BLKmode and if there is no integer mode for
3001 the number of bits being extracted.
3003 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3007 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3008 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3010 2018-06-04 Richard Biener <rguenther@suse.de>
3012 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3013 (cleanup_tree_cfg_noloop): ... single caller. Do
3014 start_recording_case_labels later.
3016 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3018 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3019 to _IMMINTRIN_H_INCLUDED.
3020 * config/i386/pconfigintrin.h: Ditto.
3021 * config/i386/waitpkgintrin.h: Ditto.
3022 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3023 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3024 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3025 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3026 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3027 waitpkgintrin.h and cldemoteintrin.h.
3029 2018-06-04 Richard Biener <rguenther@suse.de>
3031 PR tree-optimization/86038
3032 * tracer.c (find_best_successor): Check probability for
3033 being initialized, bail out if not.
3035 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3038 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3039 of bits to ignore when comparing architectures.
3041 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3043 PR tree-optimization/69615
3044 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3045 maximum or minimum of the type, try to merge it also as if
3046 range1 is + [-, x - 1] or + [x + 1, -].
3049 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3051 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3053 PR tree-optimization/86034
3054 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3055 the unsigned bitfield type in a bit insertion sequence if it does not
3056 have a larger precision than the bitfield size.
3057 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3059 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3061 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3063 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3065 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3066 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3067 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3068 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3070 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3072 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3073 Disable -fdelete-null-pointer-checks for ELF toolchain.
3075 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3076 Kito Cheng <kito.cheng@gmail.com>
3078 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3079 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3080 (nds32*-*-*): Add glibc and uclibc conditions.
3081 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3082 (TARGET_EXCEPT_UNWIND_INFO): Define.
3083 * config/nds32/elf.h: New file.
3084 * config/nds32/linux.h: New file.
3085 * config/nds32/nds32-elf.opt: New file.
3086 * config/nds32/nds32-linux.opt: New file.
3087 * config/nds32/nds32-fp-as-gp.c
3088 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3089 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3091 (nds32_asm_file_end): Ditto.
3092 (nds32_print_operand): Ditto.
3093 (nds32_insert_attributes): Ditto.
3094 (nds32_init_libfuncs): New function.
3095 (TARGET_HAVE_TLS): Define.
3096 (TARGET_INIT_LIBFUNCS): Define.
3097 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3099 (TARGET_ELF): Apply different mcmodel setting.
3100 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3101 been migrated into elf.h and linux.h files.
3102 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3103 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3104 (mcmodel): The content has been migrated into nds32-elf.opt and
3105 nds32-linux.opt files.
3106 * config/nds32/t-elf: New file.
3107 * config/nds32/t-linux: New file.
3109 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3110 Shiva Chen <shiva0217@gmail.com>
3112 * config/nds32/constants.md (unspec_volatile_element): Add
3113 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3114 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3116 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3117 (make_pass_nds32_fp_as_gp): Declare.
3118 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3120 (nds32_asm_function_end_prologue): Remove unused asm output.
3121 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3122 (nds32_asm_file_start): Output necessary fp_as_gp information.
3123 (nds32_option_override): Adjust register usage.
3124 (nds32_expand_prologue): Consider fp_as_gp situation.
3125 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3126 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3129 (simple_return): Ditto.
3130 (omit_fp_begin): Output special directive for fp_as_gp.
3131 (omit_fp_end): Output special directive for fp_as_gp.
3132 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3133 mforbid-fp-as-gp): New options.
3135 2018-06-01 Mark Wielaard <mark@klomp.org>
3137 * dwarf2out.c (dwarf2out_finish): Remove generation of
3138 DW_AT_loclists_base.
3140 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3142 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3143 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3144 (struct merged_store_group): Add bit_insertion field.
3145 (dump_char_array): Use standard hexadecimal format.
3146 (merged_store_group::merged_store_group): Set bit_insertion to false.
3147 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3148 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3149 also print the mask in the dump file.
3150 (pass_store_merging::gate): Minor tweak.
3151 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3152 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3153 stores with INTEGER_CST stores.
3154 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3155 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3156 and use it throughout. Generate bit insertion sequences if need be.
3157 (pass_store_merging::process_store): Remove redundant condition.
3158 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3160 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3162 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3163 the 128-bit floating point types. Fix function comment.
3165 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3167 * config/aarch64/aarch64-simd.md
3168 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3170 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3173 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3175 PR tree-optimization/85989
3176 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3178 (backprop::intersect_uses): Check it when deciding whether this
3179 is a backedge reference.
3180 (backprop::process_block): Add each phi to m_visited_phis
3181 after visiting it, then clear it at the end.
3183 2018-06-01 Richard Biener <rguenther@suse.de>
3185 * tree-vectorizer.h (vect_dr_stmt): New function.
3186 (vect_get_load_cost): Adjust.
3187 (vect_get_store_cost): Likewise.
3188 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3189 Use vect_dr_stmt instead of DR_SMTT.
3190 (vect_record_base_alignments): Likewise.
3191 (vect_calculate_target_alignment): Likewise.
3192 (vect_compute_data_ref_alignment): Likewise and make static.
3193 (vect_update_misalignment_for_peel): Likewise.
3194 (vect_verify_datarefs_alignment): Likewise.
3195 (vector_alignment_reachable_p): Likewise.
3196 (vect_get_data_access_cost): Likewise. Pass down
3197 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3198 (vect_get_peeling_costs_all_drs): Likewise.
3199 (vect_peeling_hash_get_lowest_cost): Likewise.
3200 (vect_enhance_data_refs_alignment): Likewise.
3201 (vect_find_same_alignment_drs): Likewise.
3202 (vect_analyze_data_refs_alignment): Likewise.
3203 (vect_analyze_group_access_1): Likewise.
3204 (vect_analyze_group_access): Likewise.
3205 (vect_analyze_data_ref_access): Likewise.
3206 (vect_analyze_data_ref_accesses): Likewise.
3207 (vect_vfa_segment_size): Likewise.
3208 (vect_small_gap_p): Likewise.
3209 (vectorizable_with_step_bound_p): Likewise.
3210 (vect_prune_runtime_alias_test_list): Likewise.
3211 (vect_analyze_data_refs): Likewise.
3212 (vect_supportable_dr_alignment): Likewise.
3213 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3214 (vect_gen_prolog_loop_niters): Likewise.
3215 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3216 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3218 (vect_recog_mask_conversion_pattern): Likewise.
3219 (vect_try_gather_scatter_pattern): Likewise.
3220 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3221 to vect_get_store_cost.
3222 (vect_get_store_cost): Get stmt_info instead of DR.
3223 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3224 (vect_get_load_cost): Get stmt_info instead of DR.
3226 2018-06-01 Richard Biener <rguenther@suse.de>
3229 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3230 vars wrapped in MEM_REFs.
3232 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3234 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3235 Fix subreg tests so that we only return a choice between
3236 GENERAL_REGS and FP_REGS if the original classes included both.
3238 2018-06-01 Richard Biener <rguenther@suse.de>
3241 * tree-ssa-structalias.c (get_function_part_constraint):
3242 Handle NULL fi->decl.
3243 (find_func_aliases_for_call): Properly handle indirect
3244 fi from direct call.
3245 (find_func_clobbers): Likewise.
3246 (ipa_pta_execute): Likewise.
3247 (create_variable_info_for): For functions that are ifunc_resolver
3248 resolve to a varinfo that contains the result of the resolver call.
3249 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3252 2018-05-31 Michael Collison <michael.collison@arm.com>
3254 * config/aarch64/aarch64.md:
3255 (*fix_to_zero_extenddfdi2): New pattern.
3256 * gcc.target/aarch64/fix_extend1.c: New testcase.
3258 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3262 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3263 and BUILT_IN_STRNCMP_EQ.
3264 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3265 BUILT_IN_STRNCMP_EQ.
3266 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3267 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3268 (gimple_fold_builtin): Likewise.
3269 * tree-ssa-strlen.c (compute_string_length): New function.
3270 (determine_min_obsize): New function.
3271 (handle_builtin_string_cmp): New function to handle calls to
3272 string compare functions.
3273 (strlen_optimize_stmt): Add handling to builtin string compare
3275 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3276 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3277 * tree.c (build_common_builtin_nodes): Add new defines of
3278 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3280 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3283 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3284 functions with naked attribute.
3286 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3288 * config/i386/sse.md (avx_vec_concat<mode>):
3289 Substitute concat_tg_mode mode attribute with xtg_mode.
3290 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3291 (concat_tg_mode): Remove mode attribute.
3293 2018-05-31 Martin Sebor <msebor@redhat.com>
3296 * calls.c (alloc_max_size): Correct a logic error/typo.
3297 Treat excessive arguments as infinite. Warn for invalid arguments.
3298 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3300 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3303 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3304 and movx for Haswell.
3306 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3307 Cesar Philippidis <cesar@codesourcery.com>
3310 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3311 when emitting error on private/firstprivate reductions.
3312 * omp-low.c (lower_omp_target): Avoid reference-type processing
3313 on pointers for firstprivate clause.
3315 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3317 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3320 * config/aarch64/aarch64-simd.md
3321 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3322 (aarch64_ld1_x3_<mode>): Likewise
3323 (aarch64_st1x2<VALLDIF:mode>): Likewise
3324 (aarch64_st1_x2_<mode>): Likewise
3325 (aarch64_st1x3<VALLDIF:mode>): Likewise
3326 (aarch64_st1_x3_<mode>): Likewise
3327 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3328 (vld1_s8_x3): Likewise.
3329 (vld1_u16_x3): Likewise.
3330 (vld1_s16_x3): Likewise.
3331 (vld1_u32_x3): Likewise.
3332 (vld1_s32_x3): Likewise.
3333 (vld1_u64_x3): Likewise.
3334 (vld1_s64_x3): Likewise.
3335 (vld1_f16_x3): Likewise.
3336 (vld1_f32_x3): Likewise.
3337 (vld1_f64_x3): Likewise.
3338 (vld1_p8_x3): Likewise.
3339 (vld1_p16_x3): Likewise.
3340 (vld1_p64_x3): Likewise.
3341 (vld1q_u8_x3): Likewise.
3342 (vld1q_s8_x3): Likewise.
3343 (vld1q_u16_x3): Likewise.
3344 (vld1q_s16_x3): Likewise.
3345 (vld1q_u32_x3): Likewise.
3346 (vld1q_s32_x3): Likewise.
3347 (vld1q_u64_x3): Likewise.
3348 (vld1q_s64_x3): Likewise.
3349 (vld1q_f16_x3): Likewise.
3350 (vld1q_f32_x3): Likewise.
3351 (vld1q_f64_x3): Likewise.
3352 (vld1q_p8_x3): Likewise.
3353 (vld1q_p16_x3): Likewise.
3354 (vld1q_p64_x3): Likewise.
3355 (vst1_s64_x2): Likewise.
3356 (vst1_u64_x2): Likewise.
3357 (vst1_f64_x2): Likewise.
3358 (vst1_s8_x2): Likewise.
3359 (vst1_p8_x2): Likewise.
3360 (vst1_s16_x2): Likewise.
3361 (vst1_p16_x2): Likewise.
3362 (vst1_s32_x2): Likewise.
3363 (vst1_u8_x2): Likewise.
3364 (vst1_u16_x2): Likewise.
3365 (vst1_u32_x2): Likewise.
3366 (vst1_f16_x2): Likewise.
3367 (vst1_f32_x2): Likewise.
3368 (vst1_p64_x2): Likewise.
3369 (vst1q_s8_x2): Likewise.
3370 (vst1q_p8_x2): Likewise.
3371 (vst1q_s16_x2): Likewise.
3372 (vst1q_p16_x2): Likewise.
3373 (vst1q_s32_x2): Likewise.
3374 (vst1q_s64_x2): Likewise.
3375 (vst1q_u8_x2): Likewise.
3376 (vst1q_u16_x2): Likewise.
3377 (vst1q_u32_x2): Likewise.
3378 (vst1q_u64_x2): Likewise.
3379 (vst1q_f16_x2): Likewise.
3380 (vst1q_f32_x2): Likewise.
3381 (vst1q_f64_x2): Likewise.
3382 (vst1q_p64_x2): Likewise.
3383 (vst1_s64_x3): Likewise.
3384 (vst1_u64_x3): Likewise.
3385 (vst1_f64_x3): Likewise.
3386 (vst1_s8_x3): Likewise.
3387 (vst1_p8_x3): Likewise.
3388 (vst1_s16_x3): Likewise.
3389 (vst1_p16_x3): Likewise.
3390 (vst1_s32_x3): Likewise.
3391 (vst1_u8_x3): Likewise.
3392 (vst1_u16_x3): Likewise.
3393 (vst1_u32_x3): Likewise.
3394 (vst1_f16_x3): Likewise.
3395 (vst1_f32_x3): Likewise.
3396 (vst1_p64_x3): Likewise.
3397 (vst1q_s8_x3): Likewise.
3398 (vst1q_p8_x3): Likewise.
3399 (vst1q_s16_x3): Likewise.
3400 (vst1q_p16_x3): Likewise.
3401 (vst1q_s32_x3): Likewise.
3402 (vst1q_s64_x3): Likewise.
3403 (vst1q_u8_x3): Likewise.
3404 (vst1q_u16_x3): Likewise.
3405 (vst1q_u32_x3): Likewise.
3406 (vst1q_u64_x3): Likewise.
3407 (vst1q_f16_x3): Likewise.
3408 (vst1q_f32_x3): Likewise.
3409 (vst1q_f64_x3): Likewise.
3410 (vst1q_p64_x3): Likewise.
3412 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3414 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3415 user_label_prefix to name.
3417 * tree-core.h: Update comment about the format of NAME string
3418 passed to handler in attribute_spec.
3420 * config/msp430/msp430.md: Remove erroneous subreg expression from
3421 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3422 zero_extend{q,h}isi2.
3424 2018-05-30 Borislav Petkov <bp@suse.de>
3426 * doc/extend.texi: Document some architecture specific
3427 constraints and sort entries.
3429 2018-05-30 Martin Sebor <msebor@redhat.com>
3432 * builtins.c (expand_builtin_stpcpy_1): New function.
3433 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3434 only if the former succeeds.
3436 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3438 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3441 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3443 * doc/invoke.texi (-flinker-output): Document
3445 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3447 * passes.c (ipa_write_summaries): Only modify statements if body
3449 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3450 incrementally linking.
3451 (ipa_passes): Likewise.
3452 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3453 pass down resolution info.
3454 * common.opt (flag_incremental_link): Update info.
3455 * gcc.c (plugin specs): Turn flinker-output=* to
3456 -plugin-opt=-linker-output-known
3457 * toplev.c (compile_file): Also cut compilation when doing incremental
3459 * flag-types. (enum lto_partition_model): Add
3460 LTO_LINKER_OUTPUT_NOLTOREL.
3461 (invoke.texi): Add -flinker-output docs.
3462 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3463 link same way as WPA; do not stream in dead initializers.
3465 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3466 darf2out_register_external_decl): Support incremental link.
3468 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3470 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3472 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3474 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3475 it down to simple_object_copy_lto_debug_sections.
3476 (run_gcc): Determine incremental LTO link time and configure
3477 lto1 into non-wpa mode, disable renaming of debug sections.
3479 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3481 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3482 descriptions of various incorrectly documented functions.
3484 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3487 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3488 address check not strict.
3490 2018-05-30 Richard Biener <rguenther@suse.de>
3492 PR tree-optimization/85964
3493 * tracer.c (better_p): Drop initialized count check, we only
3494 call the function with initialized counts now.
3495 (find_best_successor): Do find a best edge if one
3496 has uninitialized count.
3497 (find_best_predecessor): Likewise. Do BB frequency check only
3498 if count is initialized.
3500 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3502 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3503 (aarch64_ldrstr_offset_compare): New.
3504 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3505 load/store orderings.
3506 (aarch64_gen_adjusted_ldpstp): Likewise.
3508 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3510 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3511 Check for subset of GENERAL_REGS and FP_REGS.
3512 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3515 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3517 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3518 and wi::to_poly_offset. Add the current offset and then check
3519 whether the sum fits, rather than using an unchecked addition of
3520 a checked term. Check for a shwi rather than a uhwi.
3521 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3522 (store_constructor): Use poly_int_tree_p.
3523 (expand_expr_real_1): Likewise.
3524 * function.c (assign_temp): Likewise.
3525 * fold-const.c (const_binop): Use poly_int_tree_p and
3527 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3529 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3530 to_poly_offset for MEM offsets.
3531 * ipa-icf.c (sem_variable::equals): Likewise.
3532 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3533 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3534 wi::to_poly_offset for BIT_FIELD_REF offsets.
3535 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3537 * var-tracking.c (emit_note_insn_var_location): Use
3538 tree_to_poly_uint64.
3540 2018-05-29 Jim Wilson <jimw@sifive.com>
3542 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3544 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3547 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3548 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3550 (sse4_1_round<mode>2): Use nonimmediate_operand
3551 for operand 1 predicate.
3553 2018-05-29 Martin Sebor <msebor@redhat.com>
3554 Richard Biener <rguenther@suse.de>
3557 * calls.c (get_size_range): Call determine_value_range instead
3558 of get_value_range..
3559 * tree-vrp.h (determine_value_range): Declared new function.
3560 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3562 2018-05-29 Richard Biener <rguenther@suse.de>
3564 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3565 sure to use non-pattern stmts for get_earlier_stmt arguments.
3566 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3567 called on pattern stmts.
3568 (get_later_stmt): Likewise.
3570 2018-05-29 Martin Liska <mliska@suse.cz>
3572 PR gcov-profile/85759
3573 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3576 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3578 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3580 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3584 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3585 VEC_PACK_FLOAT_EXPR): New tree codes.
3586 * tree-pretty-print.c (op_code_prio): Handle
3587 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3588 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3589 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3590 * tree-inline.c (estimate_operator_cost): Likewise.
3591 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3592 * fold-const.c (const_binop): Likewise.
3593 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3594 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3595 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3596 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3597 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3598 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3599 * expr.c (expand_expr_real_2): Likewise.
3600 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3601 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3602 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3604 * optabs.c (expand_widen_pattern_expr): For
3605 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3606 sign from result type rather than operand's type.
3607 (expand_binop_directly): For vec_packu_float_optab and
3608 vec_packs_float_optab allow result type to be different from operand's
3610 * optabs-tree.c (optab_for_tree_code): Handle
3611 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3612 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3613 * tree-vect-generic.c (expand_vector_operations_1): Handle
3614 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3615 VEC_PACK_FLOAT_EXPR.
3616 * tree-vect-stmts.c (supportable_widening_operation): Handle
3618 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3619 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3620 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3621 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3622 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3624 (vec_pack<floatprefix>_float_<mode>): New expander.
3625 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3627 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3628 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3629 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3630 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3631 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3633 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3634 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3635 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3636 VEC_PACK_FLOAT_EXPR): Document.
3638 2018-05-29 Richard Biener <rguenther@suse.de>
3640 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3642 (stmt_vec_info_vec): Make pointer.
3643 (init_stmt_vec_info_vec): Remove.
3644 (free_stmt_vec_info_vec): Likewise.
3645 (set_stmt_vec_info_vec): New function.
3646 (free_stmt_vec_infos): Likewise.
3647 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3648 (set_vinfo_for_stmt): Likewise.
3649 (get_earlier_stmt): Likewise.
3650 (get_later_stmt): Likewise.
3651 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3652 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3653 (vec_info::~vec_info): Free stmt_vec_infos.
3654 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3655 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3656 (pass_slp_vectorize::execute): Likewise.
3657 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3658 (free_stmt_vec_info_vec): Likewise.
3659 (set_stmt_vec_info_vec): New function.
3660 (free_stmt_vec_infos): Likewise.
3661 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3662 the global stmt_vec_info_vec.
3663 * tree-parloops.c (gather_scalar_reductions): Use
3664 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3667 2018-05-29 Richard Biener <rguenther@suse.de>
3669 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3671 2018-05-29 Martin Liska <mliska@suse.cz>
3672 David Malcolm <dmalcolm@redhat.com>
3674 * vec.c (test_reverse): New.
3675 (vec_c_tests): Add new test.
3676 * vec.h (vl_ptr>::reverse): New function.
3678 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3680 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3682 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3685 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3687 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3689 2018-05-28 Richard Biener <rguenther@suse.de>
3691 PR tree-optimization/85933
3692 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3693 look at stmts marked as vectorizable.
3695 2018-05-28 Richard Biener <rguenther@suse.de>
3697 PR tree-optimization/85934
3698 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3699 vector boolean check before scalar optimization.
3701 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3703 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3706 2018-05-28 Mark Wielaard <mark@klomp.org>
3708 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3709 if it is an expression containing a minus sign.
3711 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3713 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3715 2018-05-27 Paul Koning <ni1d@arrl.net>
3717 * config/pdp11/pdp11.md (truncsihi2): Remove.
3719 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3720 Chung-Ju Wu <jasonwucj@gmail.com>
3722 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3724 (unaligned_store_dw): Ditto.
3725 * config/nds32/nds32-memory-manipulation.c
3726 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3727 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3728 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3729 (emit_setmem_word_loop): Rename to ...
3730 (emit_setmem_doubleword_loop): ... this.
3731 (nds32_gen_dup_4_byte_to_word_value): New function.
3732 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3733 (nds32_expand_setmem_loop): Refine implementation.
3734 (nds32_expand_setmem_loop_v3m): Ditto.
3735 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3738 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3740 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3742 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3744 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3745 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3746 machine->attr_no_prologue_p.
3747 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3748 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3749 (nds32_expand_epilogue): Consider attr_naked_p.
3750 (nds32_expand_epilogue_v3pop): Likewise.
3751 (nds32_can_use_return_insn): Likewise.
3752 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3753 attr_no_prologue_p fields.
3754 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3756 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3759 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3761 * config/i386/sse.md
3762 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3764 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3766 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3768 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3770 (*<floatsuffix>floatv2div2sf2): Rename to ...
3771 (*float<floatunssuffix>v2div2sf2): ... this.
3772 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3773 (float<floatunssuffix>v2div2sf2_mask): ... this.
3774 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3775 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3776 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3778 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3780 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3782 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3784 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3786 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3788 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3789 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3790 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3791 gen_ufix_truncv8dfv8si2.
3792 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3793 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3794 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3795 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3796 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3797 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3798 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3799 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3801 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3805 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3807 2018-05-25 Jim Wilson <jimw@sifive.com>
3809 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3810 * config/riscv/riscv.c (struct machine_function): Add
3811 interrupt_handler_p and attribute_checked_p fields.
3812 (riscv_attribute_table): Add interrupt.
3813 (riscv_interrupt_type_p): New.
3814 (riscv_save_reg_p): Save extra regs for interrupt handler.
3815 (riscv_use_save_libcall): Return false for interrupt handler.
3816 (riscv_first_stack_step): Add forward declaration.
3817 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3818 for interrupt handler with large frame. Use it for saved reg list.
3819 (riscv_expand_prologue): Move flag_stack_usage_info support to
3820 eliminate duplication.
3821 (riscv_expand_epilogue): Generate mret for interrupt handler.
3822 (riscv_epilogue_uses): New.
3823 (riscv_can_use_return_insn): Return false for interrupt handler.
3824 (riscv_function_ok_for_sibcall): Likewise.
3825 (riscv_set_current_function): Add interrupt handler support.
3826 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3827 * config/riscv/riscv.md (UNSPECV_MRET): New.
3829 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3831 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3833 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3835 PR tree-optimization/85712
3836 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3837 this candidate has already been replaced in-situ by a copy.
3839 2018-05-25 Jason Merrill <jason@redhat.com>
3841 PR c++/80485 - inline function non-zero address.
3842 * symtab.c (nonzero_address): Check DECL_COMDAT.
3844 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3847 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3848 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3849 extension of SImode operation. Use const123_operand predicate.
3850 (*saddsi_1): Remove.
3851 (*saddl_se_1): Ditto.
3853 (*ssubl_se_1): Ditto.
3854 * config/alpha/predicates.md (const123_operand): New predicate.
3855 * config/alpha/constraints.md (P): Use IN_RANGE.
3857 2018-05-25 Richard Biener <rguenther@suse.de>
3859 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3861 (ref_maybe_used_by_stmt_p): Likewise.
3862 (stmt_may_clobber_ref_p): Likewise.
3863 (stmt_may_clobber_ref_p_1): Likewise.
3864 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3866 (ref_maybe_used_by_stmt_p): Likewise.
3867 (stmt_may_clobber_ref_p): Likewise.
3868 (stmt_may_clobber_ref_p_1): Likewise.
3869 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3870 the alias oracle to disambiguate DRs with stmts DR analysis
3872 (vect_analyze_data_refs): Do not give up on not analyzable
3873 DRs for BB vectorization. Remove code truncating the dataref
3876 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3879 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3880 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3881 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3883 2018-05-25 Richard Biener <rguenther@suse.de>
3885 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3886 function, combining stmt data ref gathering and fatal analysis
3888 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3889 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3890 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3891 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3892 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3894 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3896 PR tree-optimization/85720
3897 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3898 SCC if all partitions are builtins.
3899 (version_loop_by_alias_check): New parameter. Generate cancelable
3900 runtime alias check if all partitions are builtins.
3901 (distribute_loop): Update call to above function.
3903 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3905 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3906 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3907 (parm_default_def_partition_arg): Ditto.
3908 (set_parm_default_def_partition): Ditto.
3909 (get_parm_default_def_partitions): Ditto and make it static.
3910 (get_undefined_value_partitions): Ditto and make it static.
3911 (remove_ssa_form): Refactor call to init_var_map here.
3912 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3913 computation for loop region.
3914 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3915 (register_default_def): Delete.
3916 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3917 (parm_default_def_partition_arg): Ditto.
3918 (set_parm_default_def_partition): Ditto.
3919 (get_parm_default_def_partitions): Ditto and make it static.
3920 (get_undefined_value_partitions): Ditto and make it static.
3921 (coalesce_with_default, coalesce_with_default): Update comment.
3922 (create_coalesce_list_for_region): New func factored out from
3923 create_outofssa_var_map.
3924 (populate_coalesce_list_for_outofssa): New func factored out from
3925 create_outofssa_var_map and coalesce_ssa_name.
3926 (create_outofssa_var_map): Delete.
3927 (coalesce_ssa_name): Refactor to support live range computation.
3928 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3929 (get_parm_default_def_partitions): Delete.
3930 (get_undefined_value_partitions): Ditto.
3931 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3932 computation for loop region.
3933 (new_tree_live_info, loe_visit_block): Ditto.
3934 (live_worklist, set_var_live_on_entry): Ditto.
3935 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3936 * tree-ssa-live.h (struct _var_map): New fields.
3937 (init_var_map): Change decl.
3938 (region_contains_p): New.
3940 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3942 * tree-ssa-live.h (live_merge_and_clear): Delete.
3944 2018-05-25 Richard Biener <rguenther@suse.de>
3947 * tree-dump.c (dequeue_and_dump): Remove access to removed
3948 operand 2 of a SWITCH_EXPR.
3950 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3952 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3953 multiplication and division.
3954 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3955 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3956 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3957 (cond_udiv_optab, cond_umod_optab): New optabs.
3958 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3959 (IFN_COND_RDIV): New internal functions.
3960 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3961 TRUNC_MOD_EXPR and RDIV_EXPR.
3962 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3963 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3965 (SVE_INT_BINARY): Include mult.
3966 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3967 (optab, sve_int_op): Handle mult.
3968 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3970 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3971 for SVE_INT_BINARY_SD.
3973 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3975 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3976 (optab, sve_int_op): Handle div and udiv.
3977 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3978 for SVE_INT_BINARY_SD.
3979 (*<optab><mode>3): New insn for the same.
3981 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3983 * tree-vect-patterns.c: Include predict.h.
3984 (vect_recog_divmod_pattern): Restrict check for division support
3985 to when optimizing for size.
3987 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3989 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3990 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3991 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3992 (gimple_match_op::set_op): Likewise.
3993 (gimple_resimplify4): Declare.
3994 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3995 (expr::gen_transform): Likewise.
3996 (decision_tree::gen): Generate a simplification routine for 4 operands.
3997 * gimple-match-head.c (gimple_simplify): Add an overload for
3998 4 operands. In the top-level function, handle up to 4 call
3999 arguments and call gimple_resimplify4.
4000 (gimple_resimplify4): New function.
4001 (build_call_internal): Pass a fourth operand.
4002 (maybe_push_to_seq): Likewise.
4003 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4004 Fold VEC_COND_EXPRs of an operation and a default value into
4005 an IFN_COND_* function if possible.
4006 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4008 (SVE_COND_FP_BINARY): Include them.
4009 (optab, sve_fp_op): Handle them.
4010 (SVE_INT_BINARY_REV): New code iterator.
4011 (SVE_COND_FP_BINARY_REV): New int iterator.
4012 (commutative): New int attribute.
4013 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4015 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4017 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4018 (*cond_<optab><mode>): New patterns for reversed operands.
4020 2018-05-25 Richard Biener <rguenther@suse.de>
4022 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4023 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4024 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4025 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4026 * tree-vect-data-refs.c (everywhere): Adjust users.
4027 * tree-vect-loop.c (everywhere): Likewise.
4028 * tree-vect-slp.c (everywhere): Likewise.
4029 * tree-vect-stmts.c (everywhere): Likewise.
4030 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4032 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4034 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4036 (gcc_cv_as_section_exclude): ... this.
4037 Try Solaris as #exclude syntax.
4038 * configure: Regenerate.
4039 * config.in: Regenerate.
4040 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4042 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4043 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4045 * varasm.c (default_elf_asm_named_section): Don't check if
4046 HAVE_GAS_SECTION_EXCLUDE is defined.
4048 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4050 * doc/md.texi: Update the documentation of the cond_* optabs
4051 to mention the new final operand. Fix GET_MODE_NUNITS call.
4052 Describe the scalar case too.
4053 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4054 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4056 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4057 (get_conditional_internal_fn): Update comment.
4058 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4059 accumulator value as a final argument to conditional functions.
4060 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4061 a define_expand and add an "else" operand. Assert for now that
4062 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4063 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4064 (*cond_<optab><mode>): New patterns.
4065 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4066 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4067 (UNSPEC_COND_EOR): Delete.
4068 (optab): Remove associated mappings.
4069 (SVE_INT_BINARY): New code iterator.
4070 (sve_int_op): Remove int attribute and add "minus" to the code
4072 (SVE_COND_INT_OP): Delete.
4073 (SVE_COND_FP_OP): Rename to...
4074 (SVE_COND_FP_BINARY): ...this.
4076 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4078 * optabs.c (can_reuse_operands_p): New function.
4079 (maybe_legitimize_operands): Try to reuse the results for
4082 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4084 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4085 Add {q} suffix to insn mnemonic.
4087 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4089 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4090 (msp430_warn_func_return): New.
4092 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4094 * fold-const.c (tree_nonzero_bits): New function.
4095 * fold-const.h (tree_nonzero_bits): Likewise.
4096 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4097 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4099 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4103 * varasm.c (assemble_alias): Check ifunc_resolver only on
4106 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4109 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4110 when memory input operand is handled.
4112 2018-05-24 Luis Machado <luis.machado@linaro.org>
4114 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4116 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4118 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4120 * match.pd: Delay FMA folds until after vectorization.
4122 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4125 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4126 address check not strict.
4128 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4130 * gimple-match.h (gimple_match_op): New class.
4131 (mprts_hook): Replace parameters with a gimple_match_op *.
4132 (maybe_build_generic_op): Likewise.
4133 (gimple_simplified_result_is_gimple_val): Replace parameters with
4134 a const gimple_match_op *.
4135 (gimple_simplify): Replace code_helper * and tree * parameters with
4136 a gimple_match_op * parameter.
4137 (gimple_resimplify1): Replace code_helper *, tree and tree *
4138 parameters with a gimple_match_op * parameter.
4139 (gimple_resimplify2): Likewise.
4140 (gimple_resimplify3): Likewise.
4141 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4142 parameters with a gimple_match_op * parameter.
4143 * gimple-match-head.c (gimple_simplify): Change prototypes of
4144 auto-generated functions to take a gimple_match_op * instead of
4145 separate code_helper * and tree * parameters. Make the same
4146 change in the top-level overload and update calls to the
4147 gimple_resimplify routines. Update calls to the auto-generated
4148 functions and to maybe_push_res_to_seq in the publicly-facing
4149 operation-specific gimple_simplify overloads.
4150 (gimple_match_op::MAX_NUM_OPS): Define.
4151 (gimple_resimplify1): Replace rcode and ops with a single res_op
4152 parameter. Update call to gimple_simplify.
4153 (gimple_resimplify2): Likewise.
4154 (gimple_resimplify3): Likewise.
4155 (mprts_hook): Replace parameters with a gimple_match_op *.
4156 (maybe_build_generic_op): Likewise.
4157 (build_call_internal): Replace type, nargs and ops with
4158 a gimple_match_op *.
4159 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4160 with a single gimple_match_op *. Update calls to mprts_hook,
4161 build_call_internal and gimple_simplified_result_is_gimple_val.
4162 Factor out code that is common to the tree_code and combined_fn cases.
4163 * genmatch.c (expr::gen_transform): Replace tem_code and
4164 tem_ops with a gimple_match_op called tem_op. Update calls
4165 to the gimple_resimplify functions and maybe_push_res_to_seq.
4166 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4167 res_ops. Update call to the gimple_resimplify functions.
4168 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4169 (decision_tree::gen): Make the functions take a gimple_match_op *
4170 called res_op instead of separate res_code and res_ops parameters.
4171 Update call accordingly.
4172 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4173 and ops with a single res_op parameter. Update calls to
4174 maybe_build_generic_op and maybe_push_res_to_seq.
4175 (fold_stmt_1): Update calls to gimple_simplify and
4176 replace_stmt_with_simplification.
4177 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4178 and gimple_simplified_result_is_gimple_val.
4179 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4181 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4182 with a gimple_match_op *.
4183 (vn_nary_build_or_lookup): Likewise. Update call to
4184 vn_nary_build_or_lookup_1.
4185 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4186 gimple_match_op *. Update calls to the gimple_resimplify routines
4187 and to gimple_simplified_result_is_gimple_val.
4188 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4189 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4190 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4191 (visit_nary_op): Likewise.
4192 (visit_reference_op_load): Likewise.
4194 2018-05-23 Luis Machado <luis.machado@linaro.org>
4196 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4197 modifier for printing the step amount.
4199 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4202 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4205 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4207 * doc/sourcebuild.texi (Endianness): New subsubsection.
4209 2018-05-23 Luis Machado <luis.machado@linaro.org>
4211 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4212 <prefetch_dynamic_strides>: New const bool field.
4213 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4214 prefetch_dynamic_strides.
4215 (exynosm1_prefetch_tune): Likewise.
4216 (thunderxt88_prefetch_tune): Likewise.
4217 (thunderx_prefetch_tune): Likewise.
4218 (thunderx2t99_prefetch_tune): Likewise.
4219 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4221 (aarch64_override_options_internal): Update to set
4222 PARAM_PREFETCH_DYNAMIC_STRIDES.
4223 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4224 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4225 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4226 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4227 prefetch-dynamic-strides setting.
4229 2018-05-23 Luis Machado <luis.machado@linaro.org>
4231 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4232 <minimum_stride>: New const int field.
4233 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4234 minimum_stride field defaulting to -1.
4235 (exynosm1_prefetch_tune): Likewise.
4236 (thunderxt88_prefetch_tune): Likewise.
4237 (thunderx_prefetch_tune): Likewise.
4238 (thunderx2t99_prefetch_tune): Likewise.
4239 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4240 <default_opt_level>: Set to 3.
4241 (aarch64_override_options_internal): Update to set
4242 PARAM_PREFETCH_MINIMUM_STRIDE.
4243 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4244 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4245 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4246 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4247 stride is constant and is below the minimum stride threshold.
4249 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4251 * config/arm/arm-cpus.in (mode26): Delete.
4252 (armv4): Delete mode26 reference.
4253 * config/arm/arm.c (arm_configure_build_target): Delete use of
4256 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4258 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4260 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4261 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4263 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4264 (floatunsdidf2): Ditto.
4266 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4268 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4269 (fixuns_trunc<mode>si2_avx512f): Ditto.
4270 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4271 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4272 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4274 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4276 PR rtl-optimization/79985
4277 * df-scan.c (df_insn_refs_collect): Remove special case for
4278 global registers and asm statements.
4280 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4282 * extend.texi (Global Register Variables): Rewrite the bullet list.
4283 Note that the register is available for allocation. Note that access
4284 via inline asm must use constraints. Add note about async-signal
4285 handlers. Remove paragraph about automagic register selection.
4287 2018-05-23 Richard Biener <rguenther@suse.de>
4289 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4290 of fixed offset from memset VN.
4292 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4294 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4296 (alloc_cand_and_find_basis): Initialize first_interp field.
4297 (slsr_process_mul): Modify first_interp field.
4298 (slsr_process_add): Likewise.
4299 (slsr_process_cast): Modify first_interp field for each new
4301 (slsr_process_copy): Likewise.
4302 (dump_candidate): Dump first_interp field.
4303 (replace_mult_candidate): Process all interpretations, not just
4305 (replace_rhs_if_not_dup): Likewise.
4306 (replace_one_candidate): Likewise.
4308 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4310 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4312 (aarch64_needs_frame_chain): New function.
4313 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4315 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4318 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4319 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4320 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4321 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4322 as true for strict-align.
4323 (aarch64_can_inline_p): Perform checks even when callee has no
4324 attributes to check for strict alignment.
4325 * doc/extend.texi (AArch64 Function Attributes): Document
4327 * doc/invoke.texi: (AArch64 Options): Likewise.
4329 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4331 PR tree-optimization/85853
4332 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4333 the handling of the root of the node to...
4334 (vect_slp_analyze_node_operations_1): ...this new function,
4335 and run the whole thing with the child nodes' def types
4336 set according to their SLP node's def type.
4338 2018-05-23 Richard Biener <rguenther@suse.de>
4341 * tree-data-ref.c (create_runtime_alias_checks): Defer
4342 and ignore overflow warnings.
4344 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4346 PR tree-optimization/85822
4347 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4350 2018-05-23 Richard Biener <rguenther@suse.de>
4352 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4353 memset constants via native_interpret_expr.
4355 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4358 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4360 (cgraph_node::create_alias): Likewise.
4361 (cgraph_node::get_availability): Check ifunc_resolver instead
4362 of looking up ifunc attribute.
4363 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4364 * varasm.c (do_assemble_alias): Likewise.
4365 (assemble_alias): Likewise.
4366 (default_binds_local_p_3): Likewise.
4367 * cgraph.h (cgraph_node): Add ifunc_resolver.
4368 (cgraph_node::only_called_directly_or_aliased_p): Return false
4370 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4372 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4373 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4374 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4375 instead of looking up ifunc attribute.
4377 2018-05-22 Luis Machado <luis.machado@linaro.org>
4379 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4381 2018-05-22 Martin Sebor <msebor@redhat.com>
4384 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4385 only when expasion succeeds.
4386 (expand_builtin_strcmp): Same.
4387 (expand_builtin_strncmp): Same.
4389 2018-05-22 Martin Sebor <msebor@redhat.com>
4391 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4393 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4394 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4396 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4397 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4399 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4400 Add check for legitimate_address.
4401 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4402 (aarch64_swap_ldrstr_operands): New.
4403 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4406 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4407 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4409 * config/aarch64/aarch64.md: New patterns to generate stp
4411 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4412 single words and double words.
4413 (load_pair_sw, load_pair_dw): Likewise.
4414 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4416 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4418 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4419 for different mode ldpstp and add peephole for merged zero stores.
4421 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4423 (aarch64_gen_store_pair): Rename calls to match new patterns.
4424 (aarch64_gen_load_pair): Rename calls to match new patterns.
4425 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4426 (load_pair<DREG:mode><DREG2:mode>): ... This.
4427 (store_pair<mode>): Rename to...
4428 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4429 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4431 (V_INT_EQUIV): Handle SImode.
4432 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4435 2018-05-22 Martin Sebor <msebor@redhat.com>
4438 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4439 or ajust the presumed bound on an operation to avoid unnecessary
4442 2018-05-22 Martin Sebor <msebor@redhat.com>
4444 PR tree-optimization/85826
4445 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4446 assuming that a DECL necesarily has a constant size.
4448 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4451 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4453 2018-05-22 Richard Biener <rguenther@suse.de>
4455 PR tree-optimization/85834
4456 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4457 non-constant and non-zero memset arguments.
4459 2018-05-22 Martin Liska <mliska@suse.cz>
4462 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4464 2018-05-22 Richard Biener <rguenther@suse.de>
4466 PR tree-optimization/85863
4467 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4468 comparisons when vectype is specified.
4469 (vectorizable_condition): Do not specify vectype for
4470 vect_is_simple_cond when SLP vectorizing.
4472 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4475 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4476 define __ibm128 as long double.
4477 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4478 as a distinct type when IEEE 128-bit support is enabled.
4479 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4480 128-bit types to use the correct functions.
4481 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4482 convert between 128-bit floating point types that have different
4483 modes but the same representation, instead of using gen_lowpart to
4485 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4487 (IFKF_reg): New attributes to give the register constraints for
4489 (extend<mode>tf2_internal): New insns to mark an explicit
4490 conversion between 128-bit floating point types that have a
4491 different mode but share the same representation.
4493 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4495 PR tree-optimization/85814
4496 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4497 a null return from get_strinfo when unsharing the next
4498 strinfo in the chain.
4500 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4503 * varasm.c (weak_finish): Clean up weak_decls.
4505 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4507 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4508 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4509 UNSPEC_UADALP values.
4510 * config/aarch64/iterators.md (ABAL): New int iterator.
4513 (sur): Add mappings for the above.
4514 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4516 (aarch64_<sur>abal<mode>_4): Likewise.
4517 (aarch64_<sur>adalp<mode>_3): Likewise.
4518 (<sur>sadv16qi): New define_expand.
4520 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4522 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4523 (*movdf_internal): Ditto.
4524 (*rcpsf2_sse): Ditto.
4525 (*rsqrtsf2_sse): Ditto.
4526 (*sqrt<mode>2_sse): Ditto.
4528 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4530 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4532 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4533 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4534 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4535 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4537 * config/aarch64/arm_neon.h: Likewise.
4538 * config/aarch64/iterators.md (VQ_I): New.
4540 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4542 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4544 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4547 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4549 * config/nds32/constraints.md (S): New constraint.
4550 * config/nds32/nds32.md (call_internal): Use constraint S.
4551 (call_value_internal): Likewise.
4552 (sibcall_internal): Likewise.
4553 (sibcall_value_internal): Likewise.
4555 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4556 Chung-Ju Wu <jasonwucj@gmail.com>
4558 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4561 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4562 Chung-Ju Wu <jasonwucj@gmail.com>
4564 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4565 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4566 (nds32_rtx_costs_impl): Simplify.
4567 (nds32_address_cost_impl): Simplify.
4568 (nds32_init_rtx_costs): New function.
4569 (nds32_rtx_costs_speed_prefer): Likewise.
4570 (nds32_rtx_costs_size_prefer): Likewise.
4571 (nds32_address_cost_speed_prefer): Likewise.
4572 (nds32_address_cost_speed_fwprop): Likewise.
4573 (nds32_address_cost_size_prefer): Likewise.
4574 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4575 * config/nds32/nds32.c (nds32_option_override): Use
4576 nds32_init_rtx_costs function.
4578 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4580 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4581 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4582 (TARGET_PIPELINE_N8): Likewise.
4583 (TARGET_PIPELINE_N10): Likewise.
4584 (TARGET_PIPELINE_N13): Likewise.
4585 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4587 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4589 * config/nds32/nds32-fpu.md: Update copyright year.
4591 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4593 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4595 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4597 * config/nds32/nds32.c
4598 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4599 * config/nds32/nds32.opt (minline-asm-r15): New option.
4601 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4603 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4605 * config/nds32/nds32.md (abssi2): New pattern.
4607 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4609 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4610 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4611 Merge insn pattern from sse_cvtsi2ss<round_name> and
4612 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4613 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4614 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4615 using SWI48 mode iterator.
4616 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4617 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4618 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4619 pattern from sse_cvttss2si<round_saeonly_name>
4620 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4621 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4622 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4623 using SWI48 mode iterator.
4624 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4625 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4626 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4627 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4628 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4629 using SWI48 mode iterator.
4630 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4631 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4632 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4633 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4634 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4635 SWI48 mode iterator.
4636 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4637 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4638 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4639 pattern from sse_cvttsd2si<round_saeonly_name>
4640 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4642 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4644 * config/nds32/nds32-md-auxiliary.c
4645 (nds32_valid_smw_lwm_base_p): Refine.
4646 (nds32_output_smw_single_word): Refine.
4647 (nds32_output_smw_double_word): New.
4648 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4650 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4652 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4653 (nds32_output_stack_pop): Refine.
4654 (nds32_expand_unaligned_load): Refine.
4655 (nds32_expand_unaligned_store): Refine.
4657 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4658 Chung-Ju Wu <jasonwucj@gmail.com>
4660 * config/nds32/constants.md: Add TP_REGNUM constant.
4661 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4662 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4664 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4665 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4666 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4667 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4668 and PIC code generation.
4669 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4671 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4673 * config/nds32/nds32.md: Support TLS and PIC.
4674 * config/nds32/nds32.c: Support TLS and PIC.
4675 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4676 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4679 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4681 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4682 mode with E_ prefix.
4684 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4685 Chung-Ju Wu <jasonwucj@gmail.com>
4687 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4688 * config/nds32/nds32-md-auxiliary.c
4689 (symbolic_reference_mentioned_p): New.
4690 (nds32_legitimize_ict_address): New.
4691 (nds32_expand_ict_move): New.
4692 (nds32_indirect_call_referenced_p): New.
4693 (nds32_symbol_binds_local_p): Delete.
4694 (nds32_long_call_p): Modify.
4695 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4696 * config/nds32/nds32-protos.h
4697 (symbolic_reference_mentioned_p): Declare.
4698 (nds32_legitimize_ict_address): Declare.
4699 (nds32_expand_ict_move): Declare.
4700 (nds32_indirect_call_referenced_p): Declare.
4701 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4702 (nds32_relax_group): Use nds32_ict_const_p as condition.
4703 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4704 (nds32_asm_file_start): Output ict_model directive in asm code.
4705 (nds32_legitimate_address_p): Consider indirect call.
4706 (nds32_print_operand): Consider indirect call.
4707 (nds32_print_operand_address): Consider indirect call.
4708 (nds32_insert_attributes): Handle "indirect_call" attribute.
4709 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4710 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4711 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4712 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4713 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4714 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4715 (TARGET_ICT_MODEL_SMALL): Define.
4716 (TARGET_ICT_MODEL_LARGE): Define.
4717 * config/nds32/nds32.md (movsi): Consider ict model.
4718 (call, call_value): Consider ict model.
4719 (sibcall, sibcall_value): Consider ict model.
4720 * config/nds32/nds32.opt (mict-model): New option.
4721 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4724 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4725 Monk Chiang <sh.chiang04@gmail.com>
4726 Jim Wilson <jimw@sifive.com>
4728 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4729 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4730 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4731 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4732 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4733 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4734 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4735 compute save_libcall_adjustment properly.
4736 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4737 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4738 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4739 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4740 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4741 (ABI_SPEC): Handle mabi=ilp32e.
4742 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4743 (RVE): Add RVE mask.
4744 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4745 <-march>: Add rv32e as an example.
4747 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4750 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4751 (intra_create_variable_infos): Handle C++ constructors.
4753 2018-05-18 Martin Liska <mliska@suse.cz>
4755 * passes.def: Remove a redundant pass.
4757 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4760 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4762 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4764 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4766 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4768 (armv2, armv2a, armv3, armv3m): Delete architectures.
4769 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4770 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4771 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4773 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4774 (*mulsidi3adddi): Likewise.
4775 (mulsidi3): Likewise.
4776 (*mulsidi3_nov6): Likewise.
4777 (umulsidi3): Likewise.
4778 (umulsidi3_nov6): Likewise.
4779 (umaddsidi4): Likewise.
4780 (*umulsidi3adddi): Likewise.
4781 (smulsi3_highpart): Likewise.
4782 (*smulsi3_highpart_nov6): Likewise.
4783 (umulsi3_highpart): Likewise.
4784 (*umulsi3_highpart_nov6): Likewise.
4785 * config/arm/arm.h (arm_arch3m): Delete.
4786 * config/arm/arm.c (arm_arch3m): Delete.
4787 (arm_option_override_internal): Update armv3-related comment.
4788 (arm_configure_build_target): Delete use of isa_bit_mode32.
4789 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4790 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4791 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4793 (mulusa3): Likewise.
4794 * config/arm/arm-protos.h (arm_arch3m): Delete.
4795 * config/arm/arm-tables.opt: Regenerate.
4796 * config/arm/arm-tune.md: Likewise.
4797 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4798 deleted architectures.
4800 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4802 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4803 (armv5t, armv5te): New features.
4804 (ARMv5, ARMv5e): Delete fgroups.
4805 (ARMv5t, ARMv5te): Adjust for above changes.
4807 (armv5, armv5e): Delete arches.
4808 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4810 (*call_reg_arm): Likewise.
4811 (*call_value_reg_armv5): Likewise.
4812 (*call_value_reg_arm): Likewise.
4813 (*call_symbol): Likewise.
4814 (*call_value_symbol): Likewise.
4815 (*sibcall_insn): Likewise.
4816 (*sibcall_value_insn): Likewise.
4818 (prefetch): Likewise.
4819 (define_split and define_peephole2 dependent on arm_arch5):
4821 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4823 (TARGET_ARM_QBIT): Likewise.
4824 (TARGET_DSP_MULTIPLY): Likewise.
4825 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4826 (arm_arch5, arm_arch5e): Delete.
4827 (arm_arch5t, arm_arch5te): Declare.
4828 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4829 (arm_arch5t): Declare.
4830 (arm_option_reconfigure_globals): Update for the above.
4831 (arm_options_perform_arch_sanity_checks): Update comment, replace
4832 use of arm_arch5 with arm_arch5t.
4833 (use_return_insn): Likewise.
4834 (arm_emit_call_insn): Likewise.
4835 (output_return_instruction): Likewise.
4836 (arm_final_prescan_insn): Likewise.
4837 (arm_coproc_builtin_available): Likewise.
4838 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4839 arm_arch5e with arm_arch5t and arm_arch5te.
4840 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4841 (arm_arch5t, arm_arch5te): Declare.
4842 * config/arm/arm-tables.opt: Regenerate.
4843 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4844 * config/arm/t-multilib: Likewise.
4845 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4846 instead of arm_arch5.
4847 (*call_reg_thumb1): Likewise.
4848 (*call_value_reg_thumb1_v5): Likewise.
4849 (*call_value_reg_thumb1): Likewise.
4850 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4852 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4854 2018-05-18 Martin Liska <mliska@suse.cz>
4856 PR gcov-profile/84846
4857 * doc/gcov.texi: Document -t option of gcov tool.
4859 2018-05-18 Martin Liska <mliska@suse.cz>
4861 PR gcov-profile/84846
4862 * gcov.c (print_usage): Add new -t option.
4863 (process_args): Handle the option.
4864 (generate_results): Use stdout as output when requested by
4867 2018-05-18 Martin Liska <mliska@suse.cz>
4869 PR gcov-profile/84846
4870 * coverage.c (coverage_init): Write PWD to .gcno file.
4871 * doc/gcov.texi: Document how working directory is printed.
4872 * gcov-dump.c (dump_gcov_file): Print PWD.
4873 * gcov.c (output_intermediate_file): Likewise.
4874 (read_graph_file): Read PWD string.
4875 (output_lines): Print PWD.
4877 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4880 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4881 for retval and return false if all args to phi are zero.
4883 2018-05-18 Richard Biener <rguenther@suse.de>
4885 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4887 (evrp_dom_walker::before_dom_children): Call it.
4889 2018-05-18 Richard Biener <rguenther@suse.de>
4891 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4892 results when processing array refs with variable index.
4894 2018-05-18 Toon Moene <toon@moene.org>
4896 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4897 directly after that of -floop-interchange. Indicate that both
4898 options are enabled by default when specifying -O3.
4900 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4902 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4903 iterator. Delete separate integer-mode vec_set<mode> expander.
4904 (aarch64_simd_vec_setv2di): Delete.
4905 (vec_setv2di): Delete.
4906 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4907 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4908 the "w, r" alternative.
4910 2018-05-18 Martin Liska <mliska@suse.cz>
4912 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4913 * tree-pass.h (make_pass_lower_switch_O0): New function.
4914 * tree-switch-conversion.c (node_has_low_bound): Remove.
4915 (node_has_high_bound): Likewise.
4916 (node_is_bounded): Likewise.
4917 (class pass_lower_switch): Make it a template type and create
4919 (pass_lower_switch::execute): Add template argument.
4920 (make_pass_lower_switch): New function.
4921 (make_pass_lower_switch_O0): New function.
4922 (do_jump_if_equal): Remove.
4923 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4924 all the hard work to tree optimization passes.
4926 2018-05-18 Martin Liska <mliska@suse.cz>
4928 * dbgcnt.c (limit_low): Renamed from limit.
4929 (limit_high): New variable.
4930 (dbg_cnt_is_enabled): Check for upper limit.
4931 (dbg_cnt): Adjust dumping.
4932 (dbg_cnt_set_limit_by_index): Add new argument for high
4934 (dbg_cnt_set_limit_by_name): Likewise.
4935 (dbg_cnt_process_single_pair): Parse new format.
4936 (dbg_cnt_process_opt): Use strtok.
4937 (dbg_cnt_list_all_counters): Remove 'value' and add
4939 * doc/invoke.texi: Document changes.
4941 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4943 * doc/sourcebuild.texi (scalar_all_fma): Document.
4944 * tree.def (FMA_EXPR): Delete.
4945 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4946 * internal-fn.c (ternary_direct): New macro.
4947 (expand_ternary_optab_fn): Likewise.
4948 (direct_ternary_optab_supported_p): Likewise.
4949 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4950 * builtins.c (fold_builtin_fma): Delete.
4951 (fold_builtin_3): Don't call it.
4952 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4953 * expr.c (expand_expr_real_2): Likewise.
4954 * fold-const.c (operand_equal_p): Likewise.
4955 (fold_ternary_loc): Likewise.
4956 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4957 * gimple.c (DEFTREECODE): Likewise.
4958 * gimplify.c (gimplify_expr): Likewise.
4959 * optabs-tree.c (optab_for_tree_code): Likewise.
4960 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4961 * tree-eh.c (operation_could_trap_p): Likewise.
4962 (stmt_could_throw_1_p): Likewise.
4963 * tree-inline.c (estimate_operator_cost): Likewise.
4964 * tree-pretty-print.c (dump_generic_node): Likewise.
4965 (op_code_prio): Likewise.
4966 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4967 * tree-ssa-operands.c (get_expr_operands): Likewise.
4968 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4969 * fold-const-call.h (fold_fma): Delete.
4970 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4971 CFN_FNMA and CFN_FNMS.
4973 * genmatch.c (combined_fn): New enum.
4974 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4975 (commutative_op): New function.
4976 (commutate): Use it. Handle more than 2 operands.
4977 (dt_operand::gen_gimple_expr): Use commutative_op.
4978 (parser::parse_expr): Allow :c to be used with non-binary
4979 operators if the commutative operand is known.
4980 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4981 CFN_FMS, CFN_FNMA and CFN_FNMS.
4982 (backprop::process_assign_use): Remove FMA_EXPR handling.
4983 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4984 (gen_hsa_fma): New function.
4985 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4986 IFN_FNMA and IFN_FNMS.
4987 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4988 * gimple-fold.h (follow_all_ssa_edges): Declare.
4989 * gimple-fold.c (follow_all_ssa_edges): New function.
4990 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4991 gimple_build interface and use follow_all_ssa_edges to fold the result.
4992 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4993 instead of checking for optabs directly.
4994 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4995 rather than FMA_EXPRs.
4996 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4997 call to IFN_FMA instead of an FMA_EXPR.
4999 2018-05-17 Jim Wilson <jimw@sifive.com>
5001 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5002 sign extended promoted subreg, and the range does not have the sign bit
5003 set, then do a sign extend.
5005 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5006 test, check for sign extended subreg and/or constant operands, and
5007 do a sign extend in that case.
5009 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5011 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5012 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5014 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5015 Change logics_shift_reg to logics_shift_imm.
5016 (thunderx2t99_fp_loadpair_basic): Delete.
5017 (thunderx2t99_fp_storepair_basic): Delete.
5018 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5019 (thunderx2t99_asimd_polynomial): Delete.
5020 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5021 and neon_fp_mul_d_scalar_q.
5022 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5023 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5024 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5025 (thunderx2t99_asimd_lut): Add missing tbl types.
5026 (thunderx2t99_asimd_ext): Delete.
5027 (thunderx2t99_asimd_load1_1_mult): Delete.
5028 (thunderx2t99_asimd_load1_2_mult): Delete.
5029 (thunderx2t99_asimd_load1_ldp): New.
5030 (thunderx2t99_asimd_load1): New.
5031 (thunderx2t99_asimd_load2): Add missing *load2* types.
5032 (thunderx2t99_asimd_load3): New.
5033 (thunderx2t99_asimd_load4): New.
5034 (thunderx2t99_asimd_store1_1_mult): Delete.
5035 (thunderx2t99_asimd_store1_2_mult): Delete.
5036 (thunderx2t99_asimd_store2_mult): Delete.
5037 (thunderx2t99_asimd_store2_onelane): Delete.
5038 (thunderx2t99_asimd_store_stp): New.
5039 (thunderx2t99_asimd_store1): New.
5040 (thunderx2t99_asimd_store2): New.
5041 (thunderx2t99_asimd_store3): New.
5042 (thunderx2t99_asimd_store4): New.
5044 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5046 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5047 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5049 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5050 Segher Boessenkool <segher@kernel.crashing.org>
5053 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5056 2018-05-17 Richard Biener <rguenther@suse.de>
5058 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5059 for pruning loop and prune defs feeding only already visited PHIs.
5061 2018-05-17 Richard Biener <rguenther@suse.de>
5063 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5065 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5066 Richard Biener <rguenther@suse.de>
5068 PR tree-optimization/85793
5069 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5070 for VMAT_ELEMENTWISE.
5072 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5074 * internal-fn.h (lookup_internal_fn): Declare
5075 * internal-fn.c (lookup_internal_fn): New function.
5076 * gimple.c (gimple_build_call_from_tree): Handle calls to
5078 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5079 internal function names.
5080 * tree-pretty-print.c (dump_generic_node): Likewise.
5081 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5083 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5085 * gimple-fold.h (gimple_build): Make the function forms take
5086 combined_fn rather than built_in_function.
5087 (gimple_simplify): Likewise.
5088 * gimple-match-head.c (gimple_simplify): Likewise.
5089 * gimple-fold.c (gimple_build): Likewise.
5090 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5091 rather than gimple_build_call_internal.
5092 (get_initial_defs_for_reduction): Likewise.
5093 (vect_create_epilog_for_reduction): Likewise.
5094 (vectorizable_live_operation): Likewise.
5096 2018-05-17 Martin Liska <mliska@suse.cz>
5098 * gimple-ssa-sprintf.c (format_directive): Do not use
5099 space in between 'G_' and '('.
5101 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5104 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5105 even if the mask is not all ones.
5108 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5110 (ix86_gimple_fold_builtin): Likewise.
5113 * config/i386/i386.c: Include tree-vector-builder.h.
5114 (ix86_vector_shift_count): New function.
5115 (ix86_fold_builtin): Fold shift builtins by scalar count.
5116 (ix86_gimple_fold_builtin): Likewise.
5118 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5119 _mm512_setzero): New intrinsics.
5121 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5122 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5124 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5125 code generation for cases where splatting a value is not useful.
5126 * simplify-rtx.c (simplify_ternary_operation): Simplify
5127 vec_merge across a vec_duplicate and a paradoxical subreg forming
5128 a vector mode to a vec_concat.
5130 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5132 * config.gcc: Support "goldmont-plus".
5133 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5135 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5136 PROCESSOR_GOLDMONT_PLUS.
5137 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5138 (processor_target_table): Add "goldmont-plus".
5139 (PTA_GOLDMONT_PLUS): Define.
5140 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5141 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5142 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5143 (fold_builtin_cpu): Add "goldmont-plus".
5144 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5145 (ix86_option_override_internal): Add "goldmont-plus".
5146 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5147 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5148 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5149 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5151 2018-05-17 Richard Biener <rguenther@suse.de>
5153 PR tree-optimization/85757
5154 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5155 remove defs that only feed that PHI from further processing.
5157 2018-05-16 Jim Wilson <jimw@sifive.com>
5159 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5161 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5163 2018-05-16 Mark Wielaard <mark@klomp.org>
5165 * dwarf2out.c (count_index_strings): New function.
5166 (output_indirect_strings): Call count_index_strings and generate
5167 header for dwarf_version >= 5.
5169 2018-05-16 Mark Wielaard <mark@klomp.org>
5171 * dwarf2out.c (dwarf_FORM): New function.
5172 (set_indirect_string): Use dwarf_FORM.
5173 (reset_indirect_string): Likewise.
5174 (size_of_die): Likewise.
5175 (value_format): Likewise.
5176 (output_die): Likewise.
5177 (add_skeleton_AT_string): Likewise.
5178 (output_macinfo_op): Likewise.
5179 (index_string): Likewise.
5180 (output_index_string_offset): Likewise.
5181 (output_index_string): Likewise.
5182 (count_index_strings): Likewise.
5184 2018-05-16 Carl Love <cel@us.ibm.com>
5186 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5187 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5189 2018-05-16 Martin Jambor <mjambor@suse.cz>
5191 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5192 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5194 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5196 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5197 (fnma<mode>4): Likewise.
5198 (fms<mode>4): Likewise.
5199 (fnms<mode>4): Likewise.
5200 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5201 (aarch64_fnma<mode>4): Likewise.
5202 (aarch64_fms<mode>4): Likewise.
5203 (aarch64_fnms<mode>4): Likewise.
5204 (aarch64_fnmadd<mode>4): Likewise.
5206 2018-05-16 Jason Merrill <jason@redhat.com>
5208 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5210 2018-05-16 Richard Biener <rguenther@suse.de>
5212 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5213 (dump_stmt_cost): Declare.
5214 (add_stmt_cost): Dump cost we add.
5215 (add_stmt_costs): New function.
5216 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5218 (vect_analyze_stmt): Adjust prototype.
5219 (vectorizable_condition): Likewise.
5220 (vectorizable_live_operation): Likewise.
5221 (vectorizable_reduction): Likewise.
5222 (vectorizable_induction): Likewise.
5223 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5224 cost vector to pass to vectorizable_ and record afterwards.
5225 (vect_model_reduction_cost): Take cost vector argument and adjust.
5226 (vect_model_induction_cost): Likewise.
5227 (vectorizable_reduction): Likewise.
5228 (vectorizable_induction): Likewise.
5229 (vectorizable_live_operation): Likewise.
5230 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5231 SLP_TREE_NUMBER_OF_VEC_STMTS.
5232 (vect_analyze_slp_cost_1): Remove.
5233 (vect_analyze_slp_cost): Likewise.
5234 (vect_slp_analyze_node_operations): Take visited args and
5235 a target cost vector. Avoid processing already visited stmt sets.
5236 (vect_slp_analyze_operations): Use a local cost vector to gather
5237 costs and register those of non-discarded instances.
5238 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5239 (vect_schedule_slp_instance): Remove copying of
5240 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5242 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5243 adding cost. Record cost entry location.
5244 (vect_prologue_cost_for_slp_op): Function to compute cost of
5245 a constant or invariant generated for SLP vect in the prologue,
5246 split out from vect_analyze_slp_cost_1.
5247 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5248 (vect_model_promotion_demotion_cost): Likewise.
5249 (vect_model_store_cost): Likewise, make static.
5250 (vect_model_load_cost): Likewise.
5251 (vectorizable_bswap): Add cost vector arg and adjust.
5252 (vectorizable_call): Likewise.
5253 (vectorizable_simd_clone_call): Likewise.
5254 (vectorizable_conversion): Likewise.
5255 (vectorizable_assignment): Likewise.
5256 (vectorizable_shift): Likewise.
5257 (vectorizable_operation): Likewise.
5258 (vectorizable_store): Likewise.
5259 (vectorizable_load): Likewise.
5260 (vectorizable_condition): Likewise.
5261 (vectorizable_comparison): Likewise.
5262 (can_vectorize_live_stmts): Likewise.
5263 (vect_analyze_stmt): Likewise.
5264 (vect_transform_stmt): Adjust calls to vectorizable_*.
5265 * tree-vectorizer.c: Include gimple-pretty-print.h.
5266 (dump_stmt_cost): New function.
5268 2018-05-16 Richard Biener <rguenther@suse.de>
5270 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5271 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5272 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5273 (check_name): New callback.
5274 (dse_classify_store): Track cycles via a visited bitmap of PHI
5275 defs and simplify handling of in-loop and across loop dead stores
5276 and properly fail for loop-variant refs. Handle byte-tracking with
5277 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5280 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5282 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5283 (vect_get_mask_type_for_stmt): Likewise.
5284 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5286 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5287 to determine the statement's vector type and the vector type that
5288 should be used for calculating nunits. Deal with cases in which
5289 the type has to be deferred.
5290 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5291 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5292 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5293 (vect_determine_vf_for_stmt): New functions, split out from...
5294 (vect_determine_vectorization_factor): ...here.
5295 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5296 (vect_get_mask_type_for_stmt): New functions, split out from
5297 vect_determine_vectorization_factor.
5299 2018-05-16 Richard Biener <rguenther@suse.de>
5301 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5302 verify the [VEC_]COND_EXPR embedded comparison.
5304 2018-05-15 Martin Sebor <msebor@redhat.com>
5306 PR tree-optimization/85753
5307 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5308 RECORD_TYPE in addition to ARRAY_TYPE.
5310 2018-05-15 Martin Sebor <msebor@redhat.com>
5313 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5315 2018-05-15 Richard Biener <rguenther@suse.de>
5317 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5318 add by_clobber_p one. Change algorithm to collect all defs
5319 representing uses we need to walk and try reducing them to
5320 a single one before failing.
5321 (dse_dom_walker::dse_optimize_stmt): Adjust.
5323 2018-05-13 Mark Wielaard <mark@klomp.org>
5325 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5326 (size_of_loc_descr): Likewise.
5327 (output_loc_operands): Likewise.
5328 (output_loc_operands_raw): Likewise.
5329 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5330 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5331 (hash_loc_operands): Likewise.
5332 (compare_loc_operands): Likewise.
5334 2018-05-14 Mark Wielaard <mark@klomp.org>
5336 * dwarf2out.c (count_index_addrs): New function.
5337 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5339 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5341 PR tree-optimization/83648
5342 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5343 return value as malloc candidate.
5345 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5348 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5349 param as true in call to suggest_attribute.
5351 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5353 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5354 -mreadonly-in-sdata.
5356 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5358 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5360 (aarch64_crypto_aesd_fused): Likewise.
5362 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5364 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5365 (movsi_aarch64): Likewise.
5366 (load_pairsi): Likewise.
5367 (load_pairdi): Likewise.
5368 (store_pairsi): Likewise.
5369 (store_pairdi): Likewise.
5370 (load_pairsf): Likewise.
5371 (load_pairdf): Likewise.
5372 (store_pairsf): Likewise.
5373 (store_pairdf): Likewise.
5374 (zero_extend): Likewise.
5375 (trunc): Swap alternatives.
5376 (fcvt_target): Add '?' to prefer w over r.
5378 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5381 * config/i386/i386.md: Disallow non-commutative arithmetics in
5382 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5383 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5384 in the peephole2 before it.
5386 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5388 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5389 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5390 (ix86_handle_option): Handle -mcldemote.
5391 * config.gcc: New header.
5392 * config/i386/cldemoteintrin.h: New file.
5393 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5394 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5396 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5397 OPTION_MASK_ISA_CLDEMOTE.
5398 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5399 (ix86_valid_target_attribute_inner_p): Ditto.
5400 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5401 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5402 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5403 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5404 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5406 * config/i386/i386.opt: Add -mcldemote.
5407 * config/i386/x86intrin.h: New header.
5408 * doc/invoke.texi: Add -mcldemote.
5410 2018-05-14 Richard Biener <rguenther@suse.de>
5412 * doc/match-and-simplify.texi: Adjust :s documentation.
5414 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5416 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5417 intended memcpy size.
5418 (REORDER_45): Likewise.
5420 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5422 * sort.cc: New file.
5423 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5424 * vec.c (qsort_chk): Use gcc_qsort.
5425 * Makefile.in (OBJS-libcommon): Add sort.o.
5426 (build/sort.o): New target. Use it...
5427 (BUILD_RTL): ... here, and...
5428 (build/gencfn-macros): ... here, and...
5429 (build/genmatch): ... here.
5431 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5432 Chung-Ju Wu <jasonwucj@gmail.com>
5434 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5435 * config/nds32/nds32-graywolf.md: New file.
5436 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5437 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5439 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5440 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5441 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5442 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5443 * config/nds32/pipelines.md: Include n15 settings.
5445 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5446 Chung-Ju Wu <jasonwucj@gmail.com>
5448 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5449 * config/nds32/nds32-n13.md: New file.
5450 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5451 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5453 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5454 * config/nds32/nds32.md (pipeline_model): Add n13.
5455 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5456 * config/nds32/pipelines.md: Include n13 settings.
5458 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5459 Chung-Ju Wu <jasonwucj@gmail.com>
5461 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5462 * config/nds32/nds32-n10.md: New file.
5463 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5464 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5466 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5467 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5468 * config/nds32/nds32.md (pipeline_model): Add n10.
5469 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5470 * config/nds32/pipelines.md: Include n10 settings.
5472 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5473 Kito Cheng <kito.cheng@gmail.com>
5474 Chung-Ju Wu <jasonwucj@gmail.com>
5476 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5477 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5478 Add enum values for DSP extension instructions.
5479 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5481 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5482 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5484 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5485 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5486 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5487 * config/nds32/nds32-intrinsic.md: Likewise.
5488 * config/nds32/nds32_intrinsic.h: Likewise.
5489 * config/nds32/nds32-md-auxiliary.c: Likewise.
5490 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5491 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5492 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5493 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5494 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5495 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5496 TYPE_DMAC in switch statement.
5497 * config/nds32/nds32.c: New checking and implementation for DSP
5498 extension instructions.
5499 * config/nds32/nds32.h: Likewise.
5500 * config/nds32/nds32.md: Likewise.
5501 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5502 * config/nds32/predicates.md: Implement new predicates for DSP
5505 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5507 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5508 Reformat alternatives and attributes so it is easier to identify
5509 which constraints/attributes go with which instruction.
5510 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5511 (mov<mode>_softfloat32, FMOVE64): Likewise.
5512 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5513 (mov<mode>_softfloat64, FMOVE64): Likewise.
5515 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5517 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5519 (Basic PowerPC Built-in Functions): The new name of the
5520 subsection previously known as "PowerPC Built-in Functions".
5521 (Basic PowerPC Built-in Functions Available on all Configurations):
5523 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5524 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5525 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5526 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5528 2018-05-11 Martin Jambor <mjambor@suse.cz>
5531 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5534 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5537 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5539 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5541 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5542 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5543 (ix86_handle_option): Handle -mwaitpkg.
5544 * config.gcc: New header.
5545 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5546 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5547 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5549 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5550 OPTION_MASK_ISA_WAITPKG.
5551 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5552 (ix86_option_override_internal): Add PTA_WAITPKG.
5553 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5554 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5555 IX86_BUILTIN_TPAUSE.
5556 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5557 __builtin_ia32_umwait and __builtin_ia32_tpause.
5558 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5559 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5560 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5561 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5562 UNSPECV_TPAUSE): New.
5563 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5564 * config/i386/i386.opt: Add -mwaitpkg.
5565 * config/i386/waitpkgintrin.h: New file.
5566 * config/i386/x86intrin.h: New header.
5567 * doc/invoke.texi: Add -mwaitpkg.
5569 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5572 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5574 (cortex-m0): Use armv6s-m isa.
5575 (cortex-m0plus): Likewise.
5576 (cortex-m1): Likewise.
5577 (cortex-m0.small-multiply): Likewise.
5578 (cortex-m0plus.small-multiply): Likewise.
5579 (cortex-m1.small-multiply): Likewise.
5581 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5582 Jakub Jelinek <jakub@redhat.com>
5584 PR tree-optimization/85692
5585 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5586 source permute as well.
5588 2018-05-11 Martin Liska <mliska@suse.cz>
5591 * doc/extend.texi: Document LLVM style format for no_sanitize
5594 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5596 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5597 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5598 (mode_supports_vsx_dform_quad): Likewise.
5599 (mode_supports_vmx_dform): Move these functions to be next to the
5600 other mode_supports functions.
5601 (mode_supports_dq_form): Likewise.
5602 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5603 mode_supports_dq_form.
5604 (reg_offset_addressing_ok_p): Likewise.
5605 (offsettable_ok_by_alignment): Likewise.
5606 (rs6000_legitimate_offset_address_p): Likewise.
5607 (legitimate_lo_sum_address_p): Likewise.
5608 (rs6000_legitimize_address): Likewise.
5609 (rs6000_legitimize_reload_address): Likewise.
5610 (rs6000_secondary_reload_inner): Likewise.
5611 (rs6000_preferred_reload_class): Likewise.
5612 (rs6000_output_move_128bit): Likewise.
5614 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5616 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5617 Generate SImode target register for null target.
5618 <case IX86_BUILTIN_XGETBV>: Ditto.
5619 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5620 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5622 2018-05-10 Carl Love <cel@us.ibm.com>
5624 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5625 dcbtt and dcbtstt if operands[2] is 0.
5627 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5630 * config/i386/sse.md (usadv64qi): New expander.
5632 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5634 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5635 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5636 -maltivec=be support.
5637 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5638 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5639 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5640 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5641 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5642 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5643 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5644 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5645 altivec_vsumsws): Adjust.
5646 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5647 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5648 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5650 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5651 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5652 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5653 (altivec_lve<VI_char>x): Delete expand.
5654 (*altivec_lve<VI_char>x_internal): Rename to...
5655 (altivec_lve<VI_char>x): ... this.
5656 (altivec_lvxl_<mode>): Delete expand.
5657 (*altivec_lvxl_<mode>_internal): Rename to ...
5658 (altivec_lvxl_<mode>): ... this.
5659 (altivec_stvxl_<mode>): Delete expand.
5660 (*altivec_stvxl_<mode>_internal): Rename to ...
5661 (altivec_stvxl_<mode>): ... this.
5662 (altivec_stve<VI_char>x): Delete expand.
5663 (*altivec_stve<VI_char>x_internal): Rename to ...
5664 (altivec_stve<VI_char>x): ... this.
5665 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5666 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5667 reduc_plus_scal_<mode>): Adjust.
5668 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5670 (rs6000_cpu_cpp_builtins): Adjust.
5671 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5672 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5673 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5674 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5675 -maltivec=be support.
5676 (rs6000_split_vec_extract_var): Adjust.
5677 (rs6000_split_v4si_init): Adjust.
5678 (swap_selector_for_mode): Delete.
5679 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5680 altivec_expand_stvex_be): Delete.
5681 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5682 -maltivec=be support.
5683 (rs6000_gimple_fold_builtin): Ditto.
5684 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5686 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5687 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5688 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5689 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5690 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5691 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5692 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5693 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5694 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5695 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5696 anonymous split): Adjust.
5697 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5698 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5700 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5702 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5703 when --with-gxx-include-dir is also specified.
5704 * configure: Regenerate.
5706 2018-05-09 Jim Wilson <jimw@sifive.com>
5709 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5710 * config/riscv/t-withmultilib: New.
5711 * config/riscv/withmultilib.h: New.
5712 * doc/install.texi: Document RISC-V --with-multilib-list support.
5714 2018-05-09 Richard Biener <rguenther@suse.de>
5716 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5718 (vect_bb_vectorization_profitable_p): Adjust. Compute
5719 actual scalar cost using the cost vector and the add_stmt_cost
5722 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5724 PR rtl-optimization/85645
5725 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5726 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5728 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5730 PR rtl-optimization/85645
5731 * shrink-wrap.c (spread_components): Return a boolean saying if
5732 anything was changed.
5733 (try_shrink_wrapping_separate): Iterate spread_components until
5734 nothing changes anymore.
5736 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5738 PR rtl-optimization/85645
5739 * regrename.c (build_def_use): Also kill the chains that include the
5740 destination of a REG_CFA_REGISTER note.
5742 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5744 PR rtl-optimization/85645
5745 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5746 insn that has a REG_CFA_REGISTER note.
5748 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5750 * cfgexpand.c (expand_clobber): New function.
5751 (expand_gimple_stmt_1): Use it.
5752 * tree-vect-stmts.c (vect_clobber_variable): New function,
5754 (vectorizable_simd_clone_call): ...here.
5755 (vectorizable_store): Emit a clobber either side of an
5756 IFN_STORE_LANES sequence.
5757 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5759 2018-05-09 Tom de Vries <tom@codesourcery.com>
5762 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5763 (define_insn "trap_if_false"): Add exit after trap.
5765 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5767 PR rtl-optimization/85638
5768 * bb-reorder.c: Include common/common-target.h.
5769 (create_forwarder_block): New function extracted from...
5770 (fix_up_crossing_landing_pad): ...here. Rename into...
5771 (dw2_fix_up_crossing_landing_pad): ...this.
5772 (sjlj_fix_up_crossing_landing_pad): New function.
5773 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5774 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5775 from both partitions and exit the loop after one iteration.
5777 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5780 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5782 (Basic PowerPC Built-in Functions): The new name of the
5783 subsection previously known as "PowerPC Built-in Functions".
5784 (Basic PowerPC Built-in Functions Available on all Configurations):
5786 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5788 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5790 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5792 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5795 2018-05-08 Jim Wilson <jimw@sifive.com>
5797 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5798 (LD_EMUL_SUFFIX): New.
5799 (LINK_SPEC): Use it.
5801 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5803 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5805 (Basic PowerPC Built-in Functions): The new name of the
5806 subsection previously known as "PowerPC Built-in Functions".
5807 (Basic PowerPC Built-in Functions Available on all Configurations):
5809 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5811 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5813 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5815 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5818 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5821 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5822 after cmpelim optimization.
5824 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5826 * config.gcc: Support "goldmont".
5827 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5828 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5830 * config/i386/i386.c (m_GOLDMONT): Define.
5831 (processor_target_table): Add "goldmont".
5832 (PTA_GOLDMONT): Define.
5833 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5834 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5835 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5836 (fold_builtin_cpu): Add "goldmont".
5837 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5838 (ix86_option_override_internal): Add "goldmont".
5839 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5840 (processor_type): Add PROCESSOR_GOLDMONT.
5841 * config/i386/i386.md: Add CPU "glm".
5842 * config/i386/glm.md: New file.
5843 * config/i386/x86-tune.def: Add m_GOLDMONT.
5844 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5846 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5849 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5851 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5852 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5853 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5856 * config/i386/i386.c (ix86_fold_builtin): Handle
5857 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5860 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5861 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5863 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5866 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5867 (check_arch): Likewise.
5868 (check_fpu): Return the result rather than printing it.
5869 (end arch): Fix operator precedence.
5870 (end cpu): Likewise.
5871 (END): Print the result from check_fpu.
5873 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5874 Alan Hayward <alan.hayward@arm.com>
5875 David Sherwood <david.sherwood@arm.com>
5877 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5878 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5879 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5880 (*fcmuo<mode>_and): New patterns.
5882 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5884 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5885 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5886 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5887 (cmp_op, sve_imm_con): New code attributes.
5888 (SVE_COND_INT_CMP, imm_con): Delete.
5889 (cmp_op): Remove above unspecs from int attribute.
5890 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5892 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5893 comparison-specific unspecs.
5894 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5895 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5896 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5897 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5898 (*vec_fcm<cmp_op><mode>): Rename to...
5899 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5900 (*vec_fcmuo<mode>): Rename to...
5901 (*fcmuo<mode>): ...this and adjust likewise.
5902 (*pred_fcm<cmp_op><mode>): New pattern.
5903 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5904 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5906 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5908 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5909 (aarch64_emit_sve_predicated_cond): New function.
5910 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5911 (aarch64_emit_unspec_cond_or): Replace with...
5912 (aarch64_emit_sve_or_conds): ...this new function. Use
5913 aarch64_emit_sve_ptrue_op for the individual comparisons and
5914 aarch64_emit_binop to OR them together.
5915 (aarch64_emit_inverted_unspec_cond): Replace with...
5916 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5917 aarch64_emit_sve_ptrue_op for the comparison and
5918 aarch64_emit_unop to invert the result.
5919 (aarch64_expand_sve_vec_cmp_float): Update after the above
5920 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5922 2018-05-07 Nathan Sidwell <nathan@acm.org>
5924 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5925 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5926 (Backwards Compatibility): Likewise.
5928 2018-05-07 Luis Machado <luis.machado@linaro.org>
5932 2018-05-07 Luis Machado <luis.machado@linaro.org>
5934 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5935 <prefetch_dynamic_strides>: New const bool field.
5936 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5937 prefetch_dynamic_strides.
5938 (exynosm1_prefetch_tune): Likewise.
5939 (thunderxt88_prefetch_tune): Likewise.
5940 (thunderx_prefetch_tune): Likewise.
5941 (thunderx2t99_prefetch_tune): Likewise.
5942 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5944 (aarch64_override_options_internal): Update to set
5945 PARAM_PREFETCH_DYNAMIC_STRIDES.
5946 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5947 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5948 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5949 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5950 prefetch-dynamic-strides setting.
5952 2018-05-07 Luis Machado <luis.machado@linaro.org>
5954 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5955 <minimum_stride>: New const int field.
5956 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5957 minimum_stride field.
5958 (exynosm1_prefetch_tune): Likewise.
5959 (thunderxt88_prefetch_tune): Likewise.
5960 (thunderx_prefetch_tune): Likewise.
5961 (thunderx2t99_prefetch_tune): Likewise.
5962 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5963 (aarch64_override_options_internal): Update to set
5964 PARAM_PREFETCH_MINIMUM_STRIDE.
5965 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5966 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5967 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5968 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5969 stride is constant and is below the minimum stride threshold.
5971 2018-05-07 Luis Machado <luis.machado@linaro.org>
5973 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5976 2018-05-07 Luis Machado <luis.machado@linaro.org>
5978 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5979 <prefetch_dynamic_strides>: New const bool field.
5980 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5981 prefetch_dynamic_strides.
5982 (exynosm1_prefetch_tune): Likewise.
5983 (thunderxt88_prefetch_tune): Likewise.
5984 (thunderx_prefetch_tune): Likewise.
5985 (thunderx2t99_prefetch_tune): Likewise.
5986 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5988 (aarch64_override_options_internal): Update to set
5989 PARAM_PREFETCH_DYNAMIC_STRIDES.
5990 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5991 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5992 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5993 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5994 prefetch-dynamic-strides setting.
5996 2018-05-07 Luis Machado <luis.machado@linaro.org>
5998 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5999 <minimum_stride>: New const int field.
6000 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6001 minimum_stride field.
6002 (exynosm1_prefetch_tune): Likewise.
6003 (thunderxt88_prefetch_tune): Likewise.
6004 (thunderx_prefetch_tune): Likewise.
6005 (thunderx2t99_prefetch_tune): Likewise.
6006 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6007 (aarch64_override_options_internal): Update to set
6008 PARAM_PREFETCH_MINIMUM_STRIDE.
6009 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6010 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6011 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6012 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6013 stride is constant and is below the minimum stride threshold.
6015 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6018 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6019 the type is addressable. Don't force op into register if it has
6022 2018-05-05 Roland McGrath <mcgrathr@google.com>
6025 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6026 any section for which we don't know a specific type it should have,
6027 regardless of name. Previously this was done only for the exact
6028 names ".init_array", ".fini_array", and ".preinit_array".
6029 (default_elf_asm_named_section): Add comment about
6030 relationship with default_section_type_flags and SECTION_NOTYPE.
6031 (get_section): Don't consider it a type conflict if one side has
6032 SECTION_NOTYPE and the other doesn't, as long as neither has the
6033 SECTION_BSS et al used in the default_section_type_flags logic.
6035 2018-05-05 Tom de Vries <tom@codesourcery.com>
6038 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6039 (workaround_barsyncs): New function.
6040 (nvptx_reorg): Use workaround_barsyncs.
6041 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6042 (define_expand "nvptx_membar_cta"): New define_expand.
6043 (define_insn "*nvptx_membar_cta"): New insn.
6045 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6047 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6048 To improve optimization opportunities.
6049 * builtin-types.def: The new needed builtin types for the above.
6051 2018-05-04 Richard Biener <rguenther@suse.de>
6053 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6054 * gimple-ssa-store-merging.c
6055 (imm_store_chain_info::output_merged_store): Remove redundant create,
6056 release split_store vector contents on failure.
6057 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6058 scalar stmt vector on cache hit.
6060 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6062 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6064 * config.gcc (powerpc-xilinx-eabi*): Remove.
6065 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6067 (fusion_addis_mem_combo_load): Ditto.
6068 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6070 (rs6000_cpu_cpp_builtins): Ditto.
6071 * config/rs6000/rs6000-linux.c
6072 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6073 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6074 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6076 (rs6000_setup_reg_addr_masks): Ditto.
6077 (rs6000_init_hard_regno_mode_ok): Ditto.
6078 (rs6000_option_override_internal): Ditto.
6079 (legitimate_lo_sum_address_p): Ditto.
6080 (rs6000_legitimize_address): Ditto.
6081 (rs6000_legitimize_reload_address): Ditto.
6082 (rs6000_legitimate_address_p): Ditto.
6083 (abi_v4_pass_in_fpr): Ditto.
6084 (setup_incoming_varargs): Ditto.
6085 (rs6000_gimplify_va_arg): Ditto.
6086 (rs6000_split_multireg_move): Ditto.
6087 (rs6000_savres_strategy): Ditto.
6088 (rs6000_emit_prologue_components): Ditto.
6089 (rs6000_emit_epilogue_components): Ditto.
6090 (rs6000_emit_prologue): Ditto.
6091 (rs6000_emit_epilogue): Ditto.
6092 (rs6000_elf_file_end): Ditto.
6093 (rs6000_function_value): Ditto.
6094 (rs6000_libcall_value): Ditto.
6095 * config/rs6000/rs6000.h: Ditto.
6096 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6097 (TARGET_MINMAX): ... this. New.
6098 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6099 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6100 (*movsi_internal1_single): Delete.
6101 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6102 mfpu=, mxilinx-fpu): Delete.
6103 * config/rs6000/singlefp.h: Delete.
6104 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6105 * config/rs6000/t-rs6000: Ditto.
6106 * config/rs6000/t-xilinx: Delete.
6107 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6108 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6109 (VStype_simple): Delete.
6110 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6111 * config/rs6000/xfpu.h: Delete.
6112 * config/rs6000/xfpu.md: Delete.
6113 * config/rs6000/xilinx.h: Delete.
6114 * config/rs6000/xilinx.opt: Delete.
6115 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6116 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6118 2018-05-04 Tom de Vries <tom@codesourcery.com>
6121 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6124 2018-05-04 Richard Biener <rguenther@suse.de>
6127 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6128 (expand_complex_div_wide): Likewise.
6129 (expand_complex_operations_1): Likewise.
6130 (expand_complex_libcall): Preserve EH info of the original stmt.
6131 (tree_lower_complex): Handle removed blocks.
6132 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6133 on complex multiplication and division libcall builtins.
6135 2018-05-04 Richard Biener <rguenther@suse.de>
6138 * fold-const.c (negate_expr_p): Restrict negation of operand
6139 zero of a division to when we know that can happen without
6141 (fold_negate_expr_1): Likewise.
6143 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6146 * real.h (real_nextafter): Declare.
6147 * real.c (real_nextafter): New function.
6148 * fold-const-call.c (fold_const_nextafter): New function.
6149 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6150 CASE_CFN_NEXTTOWARD.
6151 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6152 even when arg1_mode is different from arg0_mode.
6154 2018-05-03 Nathan Sidwell <nathan@acm.org>
6156 * doc/extend.texi (Deprecated Features): Remove
6158 (Backwards Compatibility): Likewise.
6159 * doc/invoke.texi (C++ Language Options): Likewise.
6160 (C++ Dialect Options): Likewise.
6162 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6165 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6166 _mm512_mask_mullox_epi64): New intrinsics.
6168 2018-05-03 Tom de Vries <tom@codesourcery.com>
6171 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6172 dump files): Add offload-tree.
6174 2018-05-03 Richard Biener <rguenther@suse.de>
6176 PR tree-optimization/85615
6177 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6178 to loops not nested in BBs loop father to avoid creating multi-entry
6181 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6183 PR tree-optimization/70291
6184 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6185 arguments. Change return type to tree. Emit libcall as a new
6186 statement rather than replacing existing one when inplace_p is true.
6187 (expand_complex_multiplication_components): New function.
6188 (expand_complex_multiplication): Expand floating-point complex
6189 multiplication using the above.
6190 (expand_complex_division): Rename inner_type parameter to type.
6191 Update expand_complex_libcall call-site.
6192 (expand_complex_operations_1): Update expand_complex_multiplication
6193 and expand_complex_division call-sites.
6195 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6198 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6199 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6200 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6201 the highest significant bit of the shift count mask is clear. In
6202 check whether and[sq]i3 is needed verify that all significant bits
6203 of the shift count other than the highest are set.
6205 2018-05-02 Tom de Vries <tom@codesourcery.com>
6208 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6209 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6210 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6211 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6212 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6213 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6214 __builtin_goacc_parlevel_size.
6216 2018-05-02 Richard Biener <rguenther@suse.de>
6218 PR tree-optimization/85597
6219 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6220 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6223 2018-05-02 Tom de Vries <tom@codesourcery.com>
6226 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6227 dump files): Add ltrans-tree.
6229 2018-05-02 Tom de Vries <tom@codesourcery.com>
6232 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6233 dump files): Add wpa-ipa.
6235 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6237 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6238 powerpc*-*-linux*paired* target.
6239 * config/rs6000/750cl.h: Delete.
6240 * config/rs6000/paired.h: Delete.
6241 * config/rs6000/paired.md: Delete.
6242 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6244 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6245 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6246 comment. Remove paired float support.
6247 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6248 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6250 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6251 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6253 * config/rs6000/rs6000.c: Remove paired float support.
6254 (paired_expand_vector_init, paired_expand_vector_move,
6255 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6256 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6257 paired_expand_builtin, paired_expand_predicate_builtin,
6258 paired_init_builtins): Delete.
6259 * config/rs6000/rs6000.h: Remove paired float support.
6260 * config/rs6000/rs6000.md: Remove paired float support.
6261 (move_from_CR_ov_bit): Delete.
6262 * config/rs6000/rs6000.opt (mpaired): Delete.
6263 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6264 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6266 2018-05-02 Richard Biener <rguenther@suse.de>
6269 * gimplify.c (gimplify_save_expr): When in SSA form allow
6270 SAVE_EXPRs to compute to SSA vars.
6272 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6275 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6276 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6277 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6278 clobber operands[2], instead use a new pseudo. Formatting fixes.
6280 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6282 PR tree-optimization/85586
6283 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6284 exit early for statements in the same group if the accesses are
6287 2018-05-02 Tom de Vries <tom@codesourcery.com>
6290 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6293 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6295 PR tree-optimization/85143
6296 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6298 2018-05-01 Tom de Vries <tom@codesourcery.com>
6301 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6302 not found" error message.
6304 2018-05-01 Tom de Vries <tom@codesourcery.com>
6307 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6308 * vec.c (test_ordered_remove_if): New function.
6309 (vec_c_tests): Call test_ordered_remove_if.
6310 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6311 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6312 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6313 VEC_ORDERED_REMOVE_IF.
6315 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6317 PR tree-optimization/82665
6318 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6319 pointer subtraction where arguments come from a memchr call.
6321 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6323 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6324 --push-state --as-needed and --pop-state instead of --as-needed and
6325 --no-as-needed if ld supports it.
6326 * configure: Regenerated.
6329 * doc/install.texi2html: Replace _002d with - and _002a with * in
6330 generated html files using sed.
6332 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6335 * gcc-rich-location.c (blank_line_before_p): New function.
6336 (use_new_line): New function.
6337 (gcc_rich_location::add_fixit_insert_formatted): New function.
6338 * gcc-rich-location.h
6339 (gcc_rich_location::add_fixit_insert_formatted): New function.
6341 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6343 * selftest.c (assert_streq): Rename "expected" and "actual" to
6344 "val1" and "val2". Extend NULL-handling to cover both inputs
6345 symmetrically, while still requiring both to be non-NULL for a pass.
6346 * selftest.h (assert_streq): Rename "expected" and "actual" to
6348 (ASSERT_EQ): Likewise.
6349 (ASSERT_EQ_AT): Likewise.
6350 (ASSERT_KNOWN_EQ): Likewise.
6351 (ASSERT_KNOWN_EQ_AT): Likewise.
6352 (ASSERT_NE): Likewise.
6353 (ASSERT_MAYBE_NE): Likewise.
6354 (ASSERT_MAYBE_NE_AT): Likewise.
6355 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6356 the assertion to pass.
6357 (ASSERT_STREQ_AT): Likewise.
6359 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6361 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6362 interaction with -pie.
6364 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6366 * selftest.h: Fix alphabetization of per-source-file selftest
6369 2018-04-30 Jason Merrill <jason@redhat.com>
6371 PR c++/61982 - dead stores to destroyed objects.
6372 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6375 2018-04-30 Jason Merrill <jason@redhat.com>
6377 * tree.c (build_clobber): New.
6378 * tree.h: Declare it.
6379 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6381 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6383 * diagnostic-show-locus.c (layout::layout): Update for
6384 location_get_source_line returning a char_span.
6385 (struct char_span): Move to input.h.
6386 (struct correction): Update for fields in char_span becoming
6388 (struct source_line): Update for location_get_source_line
6389 returning a char_span.
6390 (layout::print_line): Likewise.
6391 * edit-context.c (edited_file::print_content): Likewise.
6392 (edited_file::print_diff_hunk): Likewise.
6393 (edited_file::print_run_of_changed_lines): Likewise.
6394 (edited_file::get_num_lines): Likewise.
6395 (edited_line::edited_line): Likewise.
6396 * final.c (asm_show_source): Likewise.
6397 * input.c (location_get_source_line): Convert return type
6398 from const char * to char_span, losing the final "line_len"
6400 (dump_location_info): Update for the above.
6401 (get_substring_ranges_for_loc): Likewise. Use a char_span
6402 when handling the literal within the line.
6403 (test_reading_source_line): Update for location_get_source_line
6404 returning a char_span.
6405 * input.h (class char_span): Move here from
6406 diagnostic-show-locus.c, converting from a struct to a class.
6407 Make data members private.
6408 (char_span::operator bool): New.
6409 (char_span::length): New.
6410 (char_span::get_buffer): New.
6411 (char_span::operator[]): New.
6412 (char_span::subspan): Make const.
6413 (char_span::xstrdup): New.
6414 (location_get_source_line): Convert return type from const char *
6415 to char_span, losing the final "line_size" param.
6417 2018-04-30 Jan Hubicka <jh@suse.cz>
6419 * lto-wrapper.c (ltrans_priorities): New static var.
6420 (cmp_priority): New.
6421 (run_gcc): Read priorities and if doing parallel build order
6422 the Makefile by them.
6424 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6426 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6428 2018-04-30 Richard Biener <rguenther@suse.de>
6430 * tree-cfg.c (verify_address): Remove base argument, add
6431 flag whether to check TREE_ADDRESSABLE and do that.
6432 (verify_expr): Remove.
6433 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6434 (verify_gimple_assign_single): Likewise.
6435 (verify_gimple_switch): Likewise.
6436 (verify_expr_location_1): Dereference tp once. Add (disabled)
6437 piece from verify_expr.
6438 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6440 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6442 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6444 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6446 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6447 (small_data_pattern): Likewise.
6448 (arc_rewrite_small_data): Likewise.
6449 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6450 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6451 (get_symbol_alignment): New function.
6452 (legitimate_small_data_address_p): Likewise.
6453 (legitimate_scaled_address): Update, call
6454 legitimate_small_data_address_p.
6455 (output_sdata): New static variable.
6456 (arc_print_operand): Update how we handle small data operands.
6457 (arc_print_operand_address): Likewise.
6458 (arc_legitimate_address_p): Update, use
6459 legitimate_small_data_address_p.
6460 (arc_rewrite_small_data_p): Remove.
6461 (arc_rewrite_small_data_1): Likewise.
6462 (arc_rewrite_small_data): Likewise.
6463 (small_data_pattern): Likewise.
6464 (compact_sda_memory_operand): Update to use
6465 legitimate_small_data_address_p and get_symbol_alignment.
6466 (prepare_move_operands): Don't rewite sdata pattern.
6467 (prepare_extend_operands): Remove.
6468 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6470 (zero_extendqisi2): Likewise.
6471 (zero_extendhisi2): Likewise.
6472 (extendqihi2): Likewise.
6473 (extendqisi2): Likewise.
6474 (extendhisi2): Likewise.
6478 * config/arc/constraints.md (Usd): Change it to memory constraint.
6480 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6482 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6483 as source of std instructions.
6484 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6485 allow 6-bit constants as source for store instructions.
6486 (movdi_insn): Update instruction pattern to allow 6-bit constants
6487 as source for store instructions.
6489 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6491 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6493 2018-04-30 Nathan Sidwell <nathan@acm.org>
6494 Sandra Loosemore <sandra@codesourcery.com>
6496 * dumpfile.c (dump_open): Allow '-' for stdout.
6497 * doc/invoke.texi (Developer Options): Document dump filename
6498 determination early. Document stdin/stdout selection.
6500 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6502 Microblaze Target: PIC data text relative
6504 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6505 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6507 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6508 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6509 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6510 New addressing mode for data-text relative position indepenedent code.
6511 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6512 'ADDRESS_SYMBOLIC_TXT_REL'.
6513 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6514 (microblaze_legitimate_pic_operand): Exclude function calls from
6515 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6516 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6518 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6519 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6520 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6521 for 'address + offset'.
6522 (microblaze_expand_prologue): Add new function prologue call for
6524 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6525 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6526 table in case of TARGET_PIC_DATA_TEXT_REL.
6527 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6528 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6529 Add new macros 'UNSPEC_TEXT',
6530 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6531 + exclude function calls from 'UNSPEC_PLT' in case of data text
6533 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6534 new target hook for generating address diff vector tables in case of
6536 * doc/tm.texi : Regenerate.
6537 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6538 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6539 of addr diff vector generation.
6540 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6541 target hook definition.
6542 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6543 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6544 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6545 option for data text relative.
6547 2018-04-30 Richard Biener <rguenther@suse.de>
6549 * tree-chrec.h (evolution_function_is_constant_p): Remove
6551 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6553 2018-04-30 Richard Biener <rguenther@suse.de>
6556 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6558 2018-04-30 Richard Biener <rguenther@suse.de>
6560 PR tree-optimization/28364
6561 PR tree-optimization/85275
6562 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6563 copying first exit test.
6565 2018-04-28 Mark Wielaard <mark@klomp.org>
6567 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6569 (dwarf_AT): Handle DW_AT_addr_base.
6570 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6572 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6575 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6576 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6577 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6578 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6580 2018-04-28 Richard Biener <rguenther@suse.de>
6582 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6583 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6584 to reflect use. Only add interesting stmts.
6586 2018-04-27 Martin Jambor <mjambor@suse.cz>
6589 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6590 the jump function allows for passing through aggregate values.
6592 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6594 * input.h (in_system_header_at): Convert from macro to inline
6596 (from_macro_expansion_at): Likewise.
6597 (from_macro_definition_at): Likewise.
6599 2018-04-27 Jeff Law <law@redhat.com>
6601 * config.gcc: Mark tile* targets as deprecated/obsolete.
6603 2018-04-27 Richard Biener <rguenther@suse.de>
6605 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6608 2018-04-27 Richard Biener <rguenther@suse.de>
6610 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6612 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6614 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6615 with Yd constraint. Set "preferred_for_speed" attribute from
6616 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6618 (*movdi_internal): Ditto.
6619 (movti_interunit splitters): Remove
6620 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6621 (movdi_interunit splitters): Ditto.
6622 * config/i386/constraints.md (Ye): Remove.
6623 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6625 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6628 * config/aarch64/constraints.md (Usg): Limit to 31.
6631 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6633 PR tree-optimization/85529
6634 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6635 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6636 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6637 zero extension or masking of the MSB bit.
6638 (optimize_range_tests): Add FIRST_BB argument, pass it through
6639 to optimize_range_tests_var_bound.
6640 (maybe_optimize_range_tests, reassociate_bb): Adjust
6641 optimize_range_tests callers.
6643 2018-04-26 Richard Biener <rguenther@suse.de>
6644 Jakub Jelinek <jakub@redhat.com>
6646 * cgraph.h (symbol_table): Just declare debug method here.
6647 * symtab.c (symbol_table::debug): Define.
6649 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6651 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6653 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6655 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6656 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6657 (*movdi_internal): Substitute Yi and Yj constraint with x
6658 and Ym and Yn constraint with y constraint. Update "isa"
6659 attribute and set "preferred_for_speed" attribute from
6660 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6661 (*movsi_internal): Ditto.
6662 (*movdf_internal): Ditto.
6663 (*movsf_internal): Ditto.
6664 (*zero_extendsidi2): Ditto.
6665 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6666 (sse2_loadld): Ditto.
6667 (*vec_extract<ssevecmodelower>_0): Ditto.
6668 (*vec_extractv4si_0_zext_sse4): Ditto.
6669 (vec_concatv2di): Ditto.
6670 (*vec_dup<mode>): Ditto.
6671 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6672 * config/i386/constraints.md (Yi): Remove.
6677 2018-04-26 Nathan Sidwell <nathan@acm.org>
6679 * dumpfile.c (dump_open): New.
6680 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6681 (dump_finish): Detect stdio/stderr by value not name.
6683 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6685 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6687 2018-04-26 Tom de Vries <tom@codesourcery.com>
6690 * config/nvptx/nvptx.c (verify_neutering_jumps)
6691 (verify_neutering_labels): New function
6692 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6694 2018-04-26 Tom de Vries <tom@codesourcery.com>
6697 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6698 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6701 2018-04-26 Richard Biener <rguenther@suse.de>
6702 Tom de Vries <tom@codesourcery.com>
6705 * lto-streamer-out.c (output_function): Fixup loops if required to match
6706 discovery done in the reader.
6708 2018-04-26 Richard Biener <rguenther@suse.de>
6710 PR tree-optimization/85116
6711 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6712 have a loop exit from the single latch predecessor. Remove
6713 case of header with just condition.
6714 (ch_base::copy_headers): Exclude infinite loops from any
6716 (pass_ch::execute): Record exits.
6718 2018-04-26 Richard Biener <rguenther@suse.de>
6720 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6721 prologue cost vector and pass it to vect_get_load_cost.
6722 (vect_get_peeling_costs_all_drs): Likewise.
6723 (vect_peeling_hash_get_lowest_cost): Likewise.
6724 (vect_enhance_data_refs_alignment): Likewise.
6726 2018-04-26 Richard Biener <rguenther@suse.de>
6729 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6730 checking of integer<->pointer conversions.
6731 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6732 sign-/zero-extending pointer types.
6733 (expand_omp_for_static_chunk): Likewise.
6735 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6736 Jean Lee <xiaoyur347@gmail.com>
6738 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6739 (TARGET_ASAN_SHADOW_OFFSET): Define.
6740 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6741 true for -fsanitize=address.
6743 2018-04-25 Mark Wielaard <mark@klomp.org>
6745 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6748 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6750 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6751 than "alu", remove explicit "memory" and "imm_disp" attributes.
6752 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6755 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6756 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6757 gen_lowpart_no_emit.
6759 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6762 * config/i386/i386.c (ix86_expand_builtin): Change memory
6763 operand to XI, extend p0 to Pmode.
6764 * config/i386/i386.md: Change unspec volatile and operand
6765 1 mode to XI, change operand 0 mode to P.
6767 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6769 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6770 GET_MODE_MASK before any checking.
6771 (nds32_can_use_bset_p): Likewise.
6772 (nds32_can_use_btgl_p): Likewise.
6774 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6776 * config/nds32/nds32-doubleword.md: New define_split pattern for
6777 illegal register number.
6779 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6781 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6783 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6785 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6787 2018-04-25 Richard Biener <rguenther@suse.de>
6789 * lto-streamer.h (LTO_major_version): Bump to 8.
6791 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6793 * BASE-VER: Set to 9.0.0.
6795 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6797 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6798 in __abskf2 and __powikf2.
6800 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6803 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6804 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6805 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6806 Use the above on operand 2. Reindent.
6807 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6809 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6812 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6814 * config/i386/i386.opt (mcet): Removed.
6815 * doc/install.texi: Remove -mcet documentation.
6816 * doc/invoke.texi: Likewise.
6818 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6821 * doc/install.texi: Remove -mcet from bootstrap-cet.
6823 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6826 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6827 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6831 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6832 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6833 containing a CONST_VECTOR.
6835 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6837 * doc/install.texi: Update newlib dependency for nvptx.
6839 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6842 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6843 instead of INTVAL when shifting x left.
6845 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6847 PR tree-optimization/85478
6848 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6849 vect_grouped_store_supported for single element vectors.
6851 2018-04-24 Richard Biener <rguenther@suse.de>
6854 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6855 load cost increase to the case of non-constant step.
6857 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6860 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6861 destination if any_malformed_asm.
6863 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6866 * expr.c (store_field): In the bitfield case, if the value comes from
6867 a function call and is returned in registers by means of a PARALLEL,
6868 do not change the mode of the temporary unless BLKmode and VOIDmode.
6870 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6872 PR rtl-optimization/85423
6873 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6874 dependencies to debug insns when the previous insn is non-debug.
6876 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6878 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6879 enums into a single definition.
6880 (fls): Fix predicates and printing.
6883 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6885 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6886 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6887 and short u6 immediate.
6888 (check_if_valid_sleep_operand): Remove.
6889 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6891 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6893 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6894 flag_always_save_lp condition.
6895 * config/nds32/nds32.opt (malways-save-lp): New option.
6897 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6899 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6900 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6901 * config/nds32/nds32.h
6902 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6903 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6905 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6907 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6908 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6910 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6911 Chung-Ju Wu <jasonwucj@gmail.com>
6913 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6915 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6916 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6918 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6920 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6922 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6924 * config/nds32/nds32-protos.h (nds32_data_alignment,
6925 nds32_local_alignment): Declare.
6926 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6927 nds32_local_alignment): New functions.
6928 (TARGET_CONSTANT_ALIGNMENT): Define.
6929 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6931 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6933 * config/nds32/nds32.c
6934 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6935 (TARGET_MODES_TIEABLE_P): Likewise.
6937 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6939 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6942 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6943 Chung-Ju Wu <jasonwucj@gmail.com>
6945 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6946 for unaligned access.
6947 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6949 * config/nds32/nds32-intrinsic.md: Likewise.
6950 * config/nds32/nds32_intrinsic.h: Likewise.
6951 * config/nds32/nds32.h (nds32_builtins): Likewise.
6952 * config/nds32/nds32.opt (munaligned-access): New option.
6953 * config/nds32/nds32.c (nds32_asm_file_start): Display
6954 flag_unaligned_access status.
6956 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6958 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6959 -mno-relax is present.
6960 * config/riscv/linux.h (LINK_SPEC): Ditto.
6962 2018-04-20 Martin Sebor <msebor@redhat.com>
6965 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6967 (gimple_fold_builtin_stxcpy_chk): Same.
6968 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6970 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6973 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6974 __powikf2 when long double is IEEE 128-bit.
6976 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6978 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6979 step to make sure stack always aligned.
6981 2018-04-20 Carl Love <cel@us.ibm.com>
6984 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6985 size check for arg0.
6987 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6988 Tom de Vries <tom@codesourcery.com>
6991 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6992 Emit insns for calls too.
6993 (nvptx_find_par): Always look for worker-level predecessor insn.
6994 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6996 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6997 (nvptx_process_pars): Propagate frames for calls.
6999 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7002 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7004 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7005 (ix86_handle_option): Don't handle OPT_mibt.
7006 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7008 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7010 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7011 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7012 (ix86_target_macros): Define __CET__ with flag_cf_protection
7013 for -fcf-protection.
7014 * config/i386/i386.c (isa2_opts): Remove -mibt.
7015 * config/i386/i386.h (TARGET_IBT): Removed.
7016 (TARGET_IBT_P): Likewise.
7017 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7018 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7019 * config/i386/i386.opt (mcet): Update help message.
7022 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7023 -mcet as an alias for -mshstk.
7025 2018-04-20 Richard Biener <rguenther@suse.de>
7028 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7029 complexity by forcing a single use of the multiply operand.
7031 2018-04-20 Martin Jambor <mjambor@suse.cz>
7034 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7035 recursion dependency to only apply to non-clones.
7037 2018-04-20 Martin Jambor <mjambor@suse.cz>
7040 * ipa-cp.c (create_specialized_node): Check that clones of
7041 self-recursive edges exist during IPA-CP.
7043 2018-04-19 Toon Moene <toon@moene.org>
7045 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7048 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7050 PR tree-optimization/85467
7051 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7052 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7053 VECTOR_CST element to type.
7055 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7058 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7059 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7060 (builtin_longjmp): Likewise.
7061 (save_stack_nonlocal): New pattern.
7062 (restore_stack_nonlocal): Likewise.
7064 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7067 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7068 Replace ASM_OUTPUT_LABEL with fprintf.
7070 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7073 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7074 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7075 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7076 define __IBT__ and __SHSTK__ for -fcf-protection.
7077 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7079 (ix86_trampoline_init): Likewise.
7080 (x86_output_mi_thunk): Likewise.
7081 (ix86_notrack_prefixed_insn_p): Likewise.
7082 (ix86_option_override_internal): Don't disallow -fcf-protection.
7083 * config/i386/i386.md (rdssp<mode>): Also enable for
7085 (incssp<mode>): Likewise.
7086 (nop_endbr): Likewise.
7087 * config/i386/i386.opt (mcet): Change help message to built-in
7091 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7092 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7093 enable CET built-in functions.
7095 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7097 * common/config/i386/i386-common.c
7098 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7099 OPTION_MASK_ISA_MOVDIRI_UNSET,
7100 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7101 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7102 * config.gcc (movdirintrin.h): New header.
7103 * config/i386/cpuid.h (bit_MOVDIRI,
7104 bit_MOVDIR64B): New bits.
7105 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7107 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7108 (VOID, PVOID, PCVOID)): New function types.
7109 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7110 __builtin_ia32_directstoreu_u64,
7111 __builtin_ia32_movdir64b): New builtins.
7112 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7113 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7115 (ix86_valid_target_attribute_inner_p): Ditto.
7116 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7117 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7118 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7119 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7120 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7121 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7122 (movdiri<mode>, movdir64b_<mode>): New.
7123 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7124 * config/i386/immintrin.h: Include movdirintrin.h.
7125 * config/i386/movdirintrin.h: New file.
7126 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7128 2018-04-19 Richard Biener <rguenther@suse.de>
7131 * cfg.c (clear_bb_flags): When loop state says we have
7132 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7134 2018-04-19 Richard Biener <rguenther@suse.de>
7136 PR tree-optimization/84737
7137 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7138 copying restrict info.
7139 (vect_setup_realignment): Use it.
7140 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7141 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7142 the first DR to all generated stores.
7143 (vectorizable_load): Likewise for loads.
7145 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7147 PR tree-optimization/85446
7148 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7149 the integral and pointer types to have the same precision.
7151 * doc/install.texi: Document --disable-cet being the default and
7154 2018-04-18 Martin Liska <mliska@suse.cz>
7156 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7159 2018-04-18 Martin Liska <mliska@suse.cz>
7162 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7165 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7166 arguments if they are comparable.
7168 2018-04-18 Martin Liska <mliska@suse.cz>
7171 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7174 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7177 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7180 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7181 ENDBR after calling __morestack.
7183 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7186 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7187 by using gcc_base_ver to generate a gcc_driver_version, and use
7188 it when generating GCC_DRIVER_NAME.
7189 * configure: Regenerate.
7191 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7194 * config.gcc: Obsolete powerpc*-*-*spe*.
7196 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7199 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7200 (stabstr_D): Change type of unum from unsigned int to
7201 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7204 2018-04-17 Jim Wilson <jimw@sifive.com>
7207 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7208 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7209 Set arg_pointer_offset after using pretend_args_size.
7211 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7213 PR rtl-optimization/85431
7214 * dse.c (record_store): Ignore zero width stores.
7217 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7218 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7219 __builtin_stack_restore rather than after it.
7220 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7221 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7222 argument instead of virtual_dynamic_stack_rtx.
7224 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7226 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7228 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7229 Add note to error message to explain internal mapping of overloaded
7230 built-in function name to non-overloaded built-in function name.
7231 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7234 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7237 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7238 where the inputs overlap with the output.
7240 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7242 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7243 (=v, v) alternative and explicit "memory" attribute.
7244 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7245 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7247 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7248 "sselog1" type instead of "sselog".
7249 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7250 "sselog". Remove explicit "memory" attribute.
7251 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7252 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7254 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7255 "sselog1" type instead of "sselog". Remove explicit "memory"
7257 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7258 use "sselog1" type instead of "sselog". Remove explicit "memory"
7260 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7261 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7263 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7264 "sselog1" type instead of "sselog". Remove explicit "memory"
7266 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7267 use "sselog1" type instead of "sselog". Remove explicit "memory"
7271 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7274 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7277 2018-04-17 Martin Jambor <mjambor@suse.cz>
7280 * ipa-cp.c (create_specialized_node): Call
7281 expand_all_artificial_thunks if necessary.
7283 2018-04-17 Martin Liska <mliska@suse.cz>
7286 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7287 in message, remote space in between '_G' and '('.
7289 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7292 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7293 avx512f_vmcmp<mode>3<round_saeonly_name>,
7294 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7295 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7296 avx512f_rndscale<mode><round_saeonly_name>,
7297 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7298 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7299 Use %<iptr>2 instead of %2 for -masm=intel.
7300 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7301 avx512f_vcvttss2usi<round_saeonly_name>,
7302 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7304 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7305 avx512f_vcvttsd2usi<round_saeonly_name>,
7306 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7307 Use %q1 instead of %1 for -masm=intel.
7308 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7309 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7310 of %3 for -masm=intel.
7311 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7313 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7315 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7317 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7318 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7320 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7322 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7323 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7324 %g1 and one with %0 and %1.
7325 (avx512er_vmrcp28<mode><round_saeonly_name>,
7326 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7328 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7329 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7330 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7331 of %0 and %{%4%} for -masm=intel.
7332 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7333 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7334 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7335 order of %0 and %{%5%}%{z%} for -masm=intel.
7337 2018-04-17 Jan Hubicka <jh@suse.cz>
7340 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7342 2018-04-17 Martin Liska <mliska@suse.cz>
7345 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7346 for target_clone error message. Make default implementation
7347 clone to be a local declaration.
7348 (separate_attrs): Add new argument and check for an empty
7350 (expand_target_clones): Handle it.
7351 (ipa_target_clone): Make redirection just for target_clones
7354 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7355 Tom de Vries <tom@codesourcery.com>
7358 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7359 tiled basic blocks without omp continue statements.
7361 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7364 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7365 vec_extract expression as having side effects to make sure it gets
7368 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7371 * config/i386/i386.c (get_builtin_code_for_version): Check
7374 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7377 * gcc/config.gcc: Support "skylake".
7378 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7380 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7381 (processor_target_table): Add "skylake".
7382 (ix86_option_override_internal): Add "skylake".
7383 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7384 PROCESSOR_CANNONLAKE.
7385 (get_builtin_code_for_version): Fix priority for
7386 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7387 PROCESSOR_SKYLAKE-AVX512.
7388 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7389 (processor_type): Add PROCESSOR_SKYLAKE.
7391 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7392 Jason Merrill <jason@redhat.com>
7395 * convert.c (convert_to_integer_1): Use direct recursion for
7396 enumeral types and types with a precision less than the number
7397 of bits in their mode.
7399 2018-04-16 Julia Koval <julia.koval@intel.com>
7402 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7403 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7405 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7408 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7409 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7410 and -mno-direct-move.
7412 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7415 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7416 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7417 Ensure negative shifts result in {0}.
7419 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7421 PR rtl-optimization/79916
7422 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7423 regs (if any) to define how to gnerate SD moves when LRA is in
7426 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7428 PR rtl-optimization/85393
7429 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7430 * except.c (expand_dw2_landing_pad_for_region): Make static.
7431 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7432 a label and unconditional jump to old_bb, rather than
7433 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7436 PR rtl-optimization/85376
7437 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7438 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7439 instead of a specific value.
7441 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7442 Bin Cheng <bin.cheng@arm.com>
7444 PR tree-optimization/82965
7445 PR tree-optimization/83991
7446 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7447 by_profile_only parameter.
7448 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7449 information if the loop was predicted to iterate too many times.
7450 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7452 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7455 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7458 2018-04-13 Martin Liska <mliska@suse.cz>
7459 Jakub Jelinek <jakub@redhat.com>
7462 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7463 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7464 * builtins.c (expand_builtin_memory_copy_args): Use
7465 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7466 handle dest_addr == pc_rtx.
7468 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7471 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7472 asked to not generate direct moves.
7473 (fix_trunc<mode>si2_stfiwx): Similar.
7474 (fix_trunc<mode>si2_internal): Similar.
7476 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7479 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7480 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7481 lookup if dest in some wider mode is known to be const0_rtx and
7482 if so, record permanent equivalence for it to be ZERO_EXTEND of
7483 the narrower mode destination.
7485 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7487 * lto-streamer-out.c (output_function): Revert 259346.
7488 * omp-expand.c (expand_oacc_for): Likewise.
7490 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7492 PR rtl-optimization/85354
7493 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7494 * sel-sched.c (sel_global_init): ... here.
7496 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7499 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7500 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7501 mode for PE-COFF targets.
7502 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7503 (i386_pe_asm_lto_end): Likewise.
7504 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7505 (TARGET_ASM_LTO_END): Likewise.
7506 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7507 (i386_pe_asm_lto_start): New function.
7508 (i386_pe_asm_lto_end): Likewise.
7510 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7511 Richard Biener <rguenther@suse.de>
7514 * lto-streamer-out.c (output_function): Fix CFG loop state before
7516 * omp-expand.c (expand_oacc_for): Handle calls to internal
7517 functions like regular functions.
7519 2018-04-12 Richard Biener <rguenther@suse.de>
7522 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7523 for the early LTO debug to properly generate references to it
7524 during DIE emission. Do not re-use that for the skeleton for
7526 (dwarf2out_early_finish): Likewise.
7528 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7531 * config/i386/sse.md
7532 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7533 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7534 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7535 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7536 and output is a reg, avoid creating invalid lowpart subreg, but
7537 instead split into a 512-bit move. Don't split if not AVX512VL,
7538 input is xmm16+ reg and output is a mem.
7539 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7540 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7541 xmm16+ reg and output is a mem.
7543 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7545 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7546 also for flag_dwarf2_cfi_asm.
7548 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7550 PR rtl-optimization/85342
7551 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7552 a bool scalar var inside of the loop instead. Don't try to update
7553 recog_data.operand after failed apply_change_group.
7555 2018-04-12 Tom de Vries <tom@codesourcery.com>
7558 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7559 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7560 array with flexible array member as array without given dimension.
7561 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7562 argument for undefined param to true.
7564 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7567 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7568 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7569 from PowerPC section.
7570 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7571 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7572 help text that is too long.
7573 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7574 help text that is too long.
7575 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7576 help text that is too long.
7578 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7580 * config/alpha/alpha.md (stack_probe_internal): Rename
7581 from "probe_stack". Update all callers.
7583 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7585 PR rtl-optimization/84566
7586 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7587 sched_macro_fuse_insns.
7589 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7592 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7593 (compute_block_dependences): ... from here.
7595 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7597 PR tree-optimization/85331
7598 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7599 from int to HOST_WIDE_INT.
7601 2018-04-11 Martin Jambor <mjambor@suse.cz>
7604 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7605 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7606 not the same as the source val.
7607 (cgraph_edge_brings_value_p): New parameter.
7608 (gather_edges_for_value): Pass destination value to
7609 cgraph_edge_brings_value_p.
7610 (perhaps_add_new_callers): Likewise.
7611 (get_info_about_necessary_edges): Likewise and exclude values brought
7612 only by self-recursive edges.
7613 (create_specialized_node): Redirect only clones of self-calling edges.
7614 (+self_recursive_pass_through_p): New function.
7615 (find_more_scalar_values_for_callers_subset): Use it.
7616 (find_aggregate_values_for_callers_subset): Likewise.
7617 (known_aggs_to_agg_replacement_list): Removed.
7618 (decide_whether_version_node): Re-calculate known constants for all
7619 remaining context clones.
7621 2018-04-11 Richard Biener <rguenther@suse.de>
7624 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7625 from early DWARF output.
7626 (dwarf2out_early_finish): Output line info unconditionally into
7627 early DWARF and add reference to it.
7629 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7632 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7633 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7634 other than V2DFmode using iptr mode attribute.
7635 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7637 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7639 PR rtl-optimization/84659
7640 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7642 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7645 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7647 (output_loc_list): Pass address of a dummy size variable even in the
7648 locview handling loop.
7649 (index_location_lists): Add comment on why skip_loc_list_entry can't
7652 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7655 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7658 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7661 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7662 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7663 and -mstring-compare-inline-limit.
7665 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7668 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7669 for stack clash protection in a register whenever we need it to be in
7672 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7674 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7675 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7677 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7680 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7682 (mlong-double-): Ditto.
7683 * config/rs6000/sysv4.opt (msdata=): Ditto.
7684 (mtls-size=): Ditto.
7686 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7688 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7689 erroneous entries for
7690 "vector int vec_ldl (int, long int *)", and
7691 "vector unsigned int vec_ldl (int, unsigned long int *)".
7692 Add comments and entries for
7693 "vector bool char vec_ldl (int, bool char *)",
7694 "vector bool short vec_ldl (int, bool short *)",
7695 "vector bool int vec_ldl (int, bool int *)",
7696 "vector bool long long vec_ldl (int, bool long long *)",
7697 "vector pixel vec_ldl (int, pixel *)",
7698 "vector long long vec_ldl (int, long long *)",
7699 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7700 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7701 type tree bool_long_long_type_node and correct definition of
7702 bool_V2DI_type_node to make reference to this new type tree.
7703 (rs6000_mangle_type): Replace erroneous reference to
7704 bool_long_type_node with bool_long_long_type_node.
7705 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7706 comments to emphasize sign distinctions for char and int types and
7707 replace RS6000_BTI_bool_long constant with
7708 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7709 use of RS6000_BTI_pixel.
7710 (bool_long_type_node): Remove this macro definition.
7711 (bool_long_long_type_node): New macro definition
7713 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7715 PR rtl-optimization/85300
7716 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7717 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7718 simplify_unary_operation fails.
7720 2018-04-10 Martin Liska <mliska@suse.cz>
7722 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7723 cgraph_edge and ipa_ref.
7725 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7729 * config/i386/sse.md
7730 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7731 computation of the VEC_MERGE selector from mask.
7732 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7733 Fix decoding of the VEC_MERGE selector into mask.
7735 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7737 PR tree-optimization/85286
7738 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7740 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7742 * final.c (final_1): Set insn_last_address as well as
7743 insn_current_address.
7745 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7748 * explow.c (emit_stack_probe): Call validize_mem on memory location
7749 before passing it to gen_probe_stack. Create address operand and
7750 legitimize it for the probe_stack_address case.
7752 2018-04-09 Jan Hubicka <jh@suse.cz>
7755 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7756 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7757 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7759 (free_lang_data): Rebuild type inheritance graph.
7761 2018-04-09 Martin Sebor <msebor@redhat.com>
7763 * invoke.texi (-finline-small-functions): Mention other optimization
7765 (-findirect-inlining, -fpartial-inlining): Same.
7766 (-finline-functions-called-once): Same.
7767 (-freorder-blocks-and-partition): Same.
7769 2018-04-09 Jan Hubicka <jh@suse.cz>
7772 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7773 jumps; choose last target that matches the criteria (i.e.
7774 no partition changes for non-crossing jumps).
7775 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7776 support for redirecting crossing jumps to non-crossing.
7778 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7780 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7781 also for naked functions.
7783 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7785 * config/arc/arc.md (add_shift): New pattern.
7786 (add_shift2): Likewise.
7787 (sub_shift): Likewise.
7788 (sub_shift_cmp0_noout): Likewise.
7789 (compare_si_ashiftsi): Likewise.
7790 (xbfu_cmp0_noout): New combine pattern.
7791 (xbfu_cmp0"): Likewise.
7792 (movsi_set_cc_insn): Place the predicable variant first.
7793 (commutative_binary_cmp0_noout): Remove clobber.
7794 (commutative_binary_cmp0): New pattern.
7795 (noncommutative_binary_cmp0): Likewise.
7796 (noncommutative_binary_cmp0_noout): Likewise.
7797 (noncommutative_binary_comparison_result_used): Removed.
7798 (rsub_cmp0): New pattern.
7799 (rsub_cmp0_noout): Likewise.
7800 (extzvsi): Changed, keep only meaningful variants.
7801 (SQH, SEZ): New iterators.
7802 (SQH_postfix): New mode attribute.
7803 (SEZ_prefix): New code attribute.
7804 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7805 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7806 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7808 (noncommutative_operator): Check the availability of barrel
7811 2018-04-09 Richard Biener <rguenther@suse.de>
7813 PR tree-optimization/85284
7814 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7815 Only use the niter constraining form of simple_iv when the exit
7818 2018-04-09 Tom de Vries <tom@codesourcery.com>
7821 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7822 (define_expand "*memory_barrier"): New define_expand.
7823 (define_insn "memory_barrier"): New insn.
7825 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7827 PR rtl-optimization/80463
7828 PR rtl-optimization/83972
7829 PR rtl-optimization/83480
7831 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7832 correct producer for the insn.
7833 (tidy_control_flow): Fixup seqnos in case of debug insns.
7835 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7837 PR rtl-optimization/83913
7839 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7840 different sched-times when merging exprs.
7842 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7844 PR rtl-optimization/83962
7846 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7847 tidy_fallthru_edge and tidy_control_flow.
7849 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7851 PR rtl-optimization/83530
7853 * sel-sched.c (force_next_insn): New global variable.
7854 (remove_insn_for_debug): When force_next_insn is true, also leave only
7855 next insn in the ready list.
7856 (sel_sched_region): When the region wasn't scheduled, make another pass
7857 over it with force_next_insn set to 1.
7859 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7861 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7863 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7864 for interrupt control.
7865 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7866 functions for interrupt control.
7867 * config/nds32/nds32-intrinsic.md: Likewise.
7868 * config/nds32/nds32_intrinsic.h: Likewise.
7869 * config/nds32/nds32.h (nds32_builtins): Likewise.
7871 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7873 * config/nds32/nds32.c (nds32_init_machine_status,
7874 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7875 strict_aligned_p field.
7876 (nds32_expand_to_rtl_hook): New function.
7877 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7878 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7880 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7881 Chung-Ju Wu <jasonwucj@gmail.com>
7883 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7884 * config/nds32/nds32-n7.md: New file.
7885 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7886 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7888 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7889 * config/nds32/nds32.md (pipeline_model): Add n7.
7890 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7891 * config/nds32/pipelines.md: Include n7 settings.
7893 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7894 Chung-Ju Wu <jasonwucj@gmail.com>
7896 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7897 * config/nds32/nds32-e8.md: New file.
7898 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7899 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7901 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7902 * config/nds32/nds32.md (pipeline_model): Add e8.
7903 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7904 * config/nds32/pipelines.md: Include e8 settings.
7906 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7907 Chung-Ju Wu <jasonwucj@gmail.com>
7909 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7910 * config/nds32/nds32-n8.md: New file.
7911 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7912 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7914 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7915 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7916 * config/nds32/nds32.md (pipeline_model): Add n8.
7917 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7918 * config/nds32/pipelines.md: Include n8 settings.
7920 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7921 Chung-Ju Wu <jasonwucj@gmail.com>
7923 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7924 * config/nds32/nds32-n9-2r1w.md: New file.
7925 * config/nds32/nds32-n9-3r2w.md: New file.
7926 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7927 nds32_register_ports): New or modify for cpu n9.
7928 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7930 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7931 * config/nds32/nds32-utils.c: New file.
7932 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7933 TARGET_MUL_SLOW): Define.
7934 * config/nds32/nds32.md (pipeline_model): New attribute.
7935 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7936 New options that support cpu n9.
7937 * config/nds32/pipelines.md: Include n9 settings.
7938 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7940 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7942 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7943 information if necessary.
7944 (output_cond_branch_compare_zero): Likewise.
7945 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7946 (nds32_target_alignment): Refine for alignment.
7947 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7948 (FUNCTION_BOUNDARY): Modify.
7949 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7951 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7953 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7955 * config/nds32/constants.md (unspec_volatile_element): Add values for
7956 TLB operation and data prefetch.
7957 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7958 functions for TLB operation and data prefetch.
7959 * config/nds32/nds32-intrinsic.md: Likewise.
7960 * config/nds32/nds32_intrinsic.h: Likewise.
7961 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7962 (nds32_print_operand): Likewise.
7963 * config/nds32/nds32.h (nds32_builtins): Likewise.
7965 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7966 Andrew Pinski <pinsika@gcc.gnu.org>
7969 * match.pd: Use constant_boolean_node of correct type instead of
7970 boolean_true_node or boolean_false_node for simplifying
7971 pointer comparisons to zero.
7973 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7975 PR tree-optimization/80021
7976 * tree.c (verify_type_variant): Make error call in verify_variant_match
7977 translatable and remove final full stop.
7979 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7981 * config/nds32/constants.md (unspec_volatile_element): Add
7982 UNSPEC_VOLATILE_EH_RETURN.
7983 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7984 nds32_output_stack_pop): Support dwarf exception handling process.
7985 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7986 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7987 exception handling process.
7988 (nds32_compute_stack_frame): Likewise.
7989 (nds32_return_addr_rtx): Likewise.
7990 (nds32_initial_elimination_offset): Likewise.
7991 (nds32_expand_prologue): Likewise.
7992 (nds32_expand_epilogue): Likewise.
7993 (nds32_dynamic_chain_address): New function.
7994 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7996 (DYNAMIC_CHAIN_ADDRESS): Define.
7997 (EH_RETURN_DATA_REGNO): Define.
7998 (EH_RETURN_STACKADJ_RTX): Define.
7999 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8000 patterns for dwarf exception handling.
8002 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8004 * config/nds32/nds32.h: Clean up obsolete macros.
8006 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8008 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8009 Add enum values for particular instructions.
8010 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8011 particular intrinsic functions.
8012 * config/nds32/nds32-intrinsic.md: Likewise.
8013 * config/nds32/nds32_intrinsic.h: Likewise.
8014 * config/nds32/nds32.h (nds32_builtins): Likewise.
8015 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8016 (btst, ave): New patterns for particular instructions.
8018 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8020 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8021 Add enum values for atomic load/store and memory sync.
8022 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8024 * config/nds32/nds32-intrinsic.md: Likewise.
8025 * config/nds32/nds32_intrinsic.h: Likewise.
8026 * config/nds32/nds32.h (nds32_builtins): Likewise.
8028 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8030 PR tree-optimization/85257
8031 * fold-const.c (native_encode_vector): If not all elts could fit
8032 and off is -1, return 0 rather than offset.
8033 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8034 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8035 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8036 adjust buffer in native_interpret_expr call.
8038 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8040 * config/nds32/constants.md (unspec_volatile_element): Add cache
8041 control enum values.
8042 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8043 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8044 * config/nds32/nds32.c (nds32_cctl_names): New.
8045 (nds32_print_operand): Handle cache control register names.
8046 * config/nds32/nds32.h (nds32_builtins): New enum values.
8047 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8049 * config/nds32/nds32.md (type): Add mmu.
8050 * config/nds32/pipelines.md (simple_insn): Add mmu.
8052 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8054 * config/nds32/nds32.md (type): Remove call.
8055 * config/nds32/pipelines.md (simple_insn): Likewise.
8057 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8059 * config/nds32/constants.md (unspec_volatile_element): Add
8060 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8061 UNSPEC_VOLATILE_FMFCFG.
8062 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8063 description for fmfcfg and fmfcsr.
8064 (bdesc_1arg): Add fmtcsr.
8065 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8066 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8067 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8068 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8069 unspec_fmfcfg): New patterns.
8070 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8071 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8072 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8073 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8074 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8075 __nds32__fmfcfg): Define.
8077 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8079 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8080 intrinsic register names.
8081 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8082 intrinsic register enum values and macros.
8084 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8086 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8087 for load/store addressing form.
8088 (nds32_print_operand_address): Likewise.
8090 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8093 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8094 based on LABEL_REF. Remove useless assertion.
8095 (pic_address_needs_scratch): Fix formatting.
8096 (sparc_legitimize_pic_address): Minor tweaks.
8097 (sparc_delegitimize_address): Adjust assertion accordingly.
8098 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8099 into symbolic_operand.
8100 (movsi_high_pic_label_ref): Likewise.
8101 (movsi_lo_sum_pic_label_ref): Likewise.
8102 (movdi_pic_label_ref): Likewise.
8103 (movdi_high_pic_label_ref): Likewise.
8104 (movdi_lo_sum_pic_label_ref): Likewise.
8106 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8108 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8109 custom LIB_SPEC setup.
8111 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8112 Kito Cheng <kito.cheng@gmail.com>
8114 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8115 * config/riscv/freebsd.h: New.
8117 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8119 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8120 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8123 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8124 Kito Cheng <kito.cheng@gmail.com>
8126 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8127 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8128 * config/nds32/nds32-protos.h (nds32_output_call,
8129 nds32_output_return): Declare.
8130 * config/nds32/nds32.md: Refine all the call and return patterns.
8132 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8135 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8136 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8138 PR rtl-optimization/84872
8139 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8140 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8143 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8145 * expr.c (copy_blkmode_to_reg): Revert 254862.
8146 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8148 2018-04-06 Richard Biener <rguenther@suse.de>
8151 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8152 after seeing a component reference with an adjacent field. Treat
8153 refs to arrays at struct end of external decls similar to
8154 refs to unconstrained commons.
8156 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8159 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8160 look through SAVE_EXPRs with non-side-effects argument. Adjust
8162 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8165 2018-04-06 Richard Biener <rguenther@suse.de>
8168 * alias.c (find_base_term): New wrapper around find_base_term
8169 unwinding CSELIB_VAL_PTR changes.
8170 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8173 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8175 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8177 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8178 constant definitions.
8179 ("nop"): lr 0,0 -> nopr r0
8180 ("nop_lr0", "nop_lr1"): New insn definitions.
8182 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8184 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8185 NDS32_V3PUSH_AVAILABLE_P macro.
8187 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8188 Chung-Ju Wu <jasonwucj@gmail.com>
8190 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8191 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8192 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8193 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8194 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8195 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8196 * config/nds32/constraints.md: New constraints and checking for hard
8197 float configuration.
8198 * config/nds32/iterators.md: New mode iterator and attribute for hard
8199 float configuration.
8200 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8202 * config/nds32/nds32-fpu.md: New file.
8203 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8204 deal with hard float code generation.
8205 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8207 (abi_type, float_reg_number): New enum type.
8208 * config/nds32/nds32-predicates.c: New predicates for hard float.
8209 * config/nds32/nds32-protos.h: Declare functions for hard float.
8210 * config/nds32/nds32.c: Implementation for hard float configuration.
8211 * config/nds32/nds32.h: Definitions for hard float configuration.
8212 * config/nds32/nds32.md: Include hard float machine description and
8213 modify patterns for hard float configuration.
8214 * config/nds32/nds32.opt: New options for hard float configuration.
8215 * config/nds32/predicates.md: New predicates for hard float
8218 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8220 * common/config/nds32/nds32-common.c
8221 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8223 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8226 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8227 CONSTRUCTOR_ELT (ctor, ...)->value.
8229 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8232 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8234 2018-04-05 Tom de Vries <tom@codesourcery.com>
8237 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8240 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8241 Kito Cheng <kito.cheng@gmail.com>
8243 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8244 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8245 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8246 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8248 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8249 Kito Cheng <kito.cheng@gmail.com>
8251 * config/nds32/constraints.md (Ufe): New memory constraint.
8252 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8253 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8254 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8256 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8257 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8259 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8261 * config/nds32/nds32.md: Use optimize_size in the condition for
8262 alu-shift instructions.
8264 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8266 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8268 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8270 * config/nds32/nds32.md (negsi2): Refine pattern.
8272 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8273 Chung-Ju Wu <jasonwucj@gmail.com>
8275 * config/nds32/iterators.md (shift_rotate): New code iterator.
8276 (shift): New code attribute.
8277 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8278 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8279 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8280 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8281 bit-wise operations.
8282 (andsi3, *andsi3): Ditto.
8283 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8284 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8285 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8286 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8287 nds32_ior_operand, nds32_xor_operand): New predicates.
8289 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8291 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8292 (addsi3, subsi3): ... this.
8294 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8296 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8298 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8300 * config/nds32/nds32.md: Adjust indention.
8302 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8304 * config/nds32/nds32.md (feature): New attribute.
8306 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8308 * config/nds32/nds32.md (subtype): New attribute.
8310 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8313 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8314 expansion to perform a bitwise AND of the argument followed by a
8315 boolean negation of the result.
8317 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8319 PR rtl-optimization/84878
8320 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8321 the basic block. Assert the use reference is not artificial and that
8322 it has an associated insn.
8324 2018-04-04 Michael Matz <matz@suse.de>
8326 * builtins.c (compute_objsize): Pass correct operand
8327 to array_at_struct_end_p.
8329 2018-04-04 Richard Biener <rguenther@suse.de>
8332 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8333 from contexts for DINFO_LEVEL_TERSE and below.
8335 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8337 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8338 resiter_operand condition.
8339 * config/nds32/nds32.md (*move<mode>): Ditto.
8341 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8342 Monk Chiang <sh.chiang04@gmail.com>
8344 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8346 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8348 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8350 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8351 Kito Cheng <kito.cheng@gmail.com>
8353 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8354 nds32_cond_code_str, output_cond_branch,
8355 output_cond_branch_compare_zero, nds32_expand_cbranch,
8356 nds32_expand_cstore, nds32_expand_movcc,
8357 nds32_output_cbranchsi4_equality_zero,
8358 nds32_output_cbranchsi4_equality_reg,
8359 nds32_output_cbranchsi4_equality_reg_or_const_int,
8360 nds32_output_cbranchsi4_greater_less_zero: New functions.
8361 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8362 nds32_expand_cstore, nds32_expand_movcc,
8363 nds32_output_cbranchsi4_equality_zero,
8364 nds32_output_cbranchsi4_equality_reg,
8365 nds32_output_cbranchsi4_equality_reg_or_const_int,
8366 nds32_output_cbranchsi4_greater_less_zero): Declare.
8367 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8368 nds32_rimm11s_operand): New predicates.
8369 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8370 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8373 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8375 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8376 * config/nds32/nds32.md: Ditto.
8377 * config/nds32/pipelines.md: Ditto.
8379 2018-04-04 Richard Biener <rguenther@suse.de>
8381 PR tree-optimization/85168
8382 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8383 propagating abnormals.
8385 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8387 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8389 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8390 Kito Cheng <kito.cheng@gmail.com>
8392 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8393 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8394 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8395 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8396 * config/nds32/nds32.md (sibcall_internal): New.
8397 (sibcall_register): Remove.
8398 (sibcall_immediate): Remove.
8399 (sibcall_value_internal): New.
8400 (sibcall_value_register): Remove.
8401 (sibcall_value_immediate): Remove.
8402 * config/nds32/predicates.md (nds32_general_register_operand): New.
8403 (nds32_call_address_operand): New.
8405 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8407 PR rtl-optimization/85167
8408 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8409 bb_defs if *split_p, instead preinitialize it to NULL.
8411 PR tree-optimization/85156
8412 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8413 evaluating the argument multiple times.
8415 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8417 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8419 (_mm_cvtpd_ps): Likewise.
8420 (_mm_cvttpd_epi32): Likewise.
8421 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8422 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8423 vector, pixel, and bool following altivec.h include.
8425 2018-04-03 Martin Sebor <msebor@redhat.com>
8427 * doc/extend.texi (Common Function Attributes): Clarify.
8428 (const attribute): Likewise.
8429 (pure attribute): Likewise.
8431 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8434 * config/i386/i386.c (ix86_expand_vector_set): Use
8435 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8437 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8439 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8440 instructions when changing rounding bits to preserve precision bits
8441 in the x87 control word.
8443 2018-04-03 Martin Liska <mliska@suse.cz>
8445 PR tree-optimization/82491
8446 * rtl.h (strip_offset_and_add): Replace += suboffset with
8447 poly_uint64 () + suboffset.
8449 2018-03-29 Martin Liska <mliska@suse.cz>
8450 Martin Jambor <mjambor@suse.cz>
8453 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8454 param_type is not an integral or pointer type.
8456 2018-04-03 Richard Biener <rguenther@suse.de>
8458 * sese.h (recompute_all_dominators): Remove.
8460 2018-04-02 Martin Sebor <msebor@redhat.com>
8462 * doc/invoke.texi (-Wrestrict): Fix typos.
8464 2018-04-02 Jim Wilson <jimw@sifive.com>
8466 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8467 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8468 (<optab>di3, <optab>si3_extend): Likewise.
8469 (<optab>si3_mask, <optab>si3_mask_1): New.
8470 (<optab>di3_mask, <optab>di3_mask_1): New.
8471 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8472 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8473 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8475 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8477 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8480 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8482 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8483 (nds32_canonicalize_comparison): New function.
8485 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8486 Kito Cheng <kito.cheng@gmail.com>
8487 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8489 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8490 * config/nds32/constants.md (unspec_volatile_element): Add
8491 UNSPEC_VOLATILE_RELAX_GROUP.
8492 * config/nds32/nds32-relax-opt.c: New file.
8493 * config/nds32/nds32-predicates.c
8494 (nds32_symbol_load_store_p): New function.
8495 * config/nds32/nds32-protos.h
8496 (nds32_symbol_load_store_p): Declare function.
8497 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8498 * config/nds32/nds32.c
8499 (nds32_register_pass): New function to register pass.
8500 (nds32_register_passes): New function to register passes.
8501 * config/nds32/nds32.md (relax_group): New pattern.
8502 * config/nds32/nds32.opt (mrelax-hint): New option.
8503 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8505 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8507 * config/nds32/t-nds32: Modify files dependency.
8509 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8511 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8512 (PROFILE_HOOK): Define its implementation.
8514 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8516 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8517 type and 32-bit size.
8519 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8522 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8523 (V_128_256): New mode iterator.
8524 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8525 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8526 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8528 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8529 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8531 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8534 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8535 NaN inputs correctly.
8537 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8540 * config/rs6000/vsx.md (??r): New mode attribute.
8541 (*vsx_mov<mode>_64bit): Use it.
8542 (*vsx_mov<mode>_32bit): Likewise.
8544 2018-03-30 Martin Sebor <msebor@redhat.com>
8546 PR tree-optimization/84818
8547 * builtins.c (check_access): Use warning_n.
8549 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8552 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8554 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8557 2018-03-30 Julia Koval <julia.koval@intel.com>
8560 * x86-tune.def (movx, partial_reg_dependency): Enable for
8563 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8566 * lra-constraints.c (process_alt_operands): Move setting
8567 this_alternative_matches below.
8569 2018-03-29 Martin Liska <mliska@suse.cz>
8572 * doc/invoke.texi: Document how LTO works with debug info.
8573 Describe auto-load support of binutils. Mention 'x86-64'
8574 as valid option value of -march option.
8576 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8578 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8581 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8582 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8583 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8586 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8589 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8590 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8591 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8592 for RS6000_BTM_POWERPC64.
8593 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8594 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8595 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8600 2018-03-28 Carl Love <cel@us.ibm.com>
8603 2017-09-27 Carl Love <cel@us.ibm.com>
8605 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8606 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8607 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8610 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8612 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8613 instead of __vector bool.
8614 (_mm_max_pu8): Likewise.
8615 (_mm_min_pi16): Likewise.
8617 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8620 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8621 (DIVWEUO): Likewise.
8623 (DIVDEUO): Likewise.
8624 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8625 DIVWEUO and DIVDEUO.
8626 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8627 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8628 (div_extend): Likewise.
8629 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8631 (__builtin_divweuo): Likewise.
8632 (__builtin_divdeo): Likewise.
8633 (__builtin_divdeuo): Likewise.
8635 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8638 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8639 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8641 PR tree-optimization/82004
8642 * gimple-match-head.c (optimize_pow_to_exp): New function.
8643 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8644 Don't fold to exp if optimize_pow_to_exp is false.
8646 2018-03-28 Martin Liska <mliska@suse.cz>
8649 * calls.c (initialize_argument_information): Fix trailing space.
8650 * common.opt: Fix typo and provide better explanation for
8651 -fsanitize-coverage option.
8652 * config/i386/i386.opt: Fix typo.
8654 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8655 Martin Liska <mliska@suse.cz>
8658 * gimplify.c (asan_poison_variable): Don't do the check for
8659 gimplify_omp_ctxp here.
8660 (gimplify_decl_expr): Do it here.
8661 (gimplify_target_expr): Likewise.
8663 2018-03-28 Martin Liska <mliska@suse.cz>
8666 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8667 chkp_type_bounds_count if MPX is not enabled.
8669 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8671 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8673 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8676 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8677 function to create the function decl for complex long double
8678 multiply and divide for -mabi=ieeelongdouble.
8679 (init_float128_ieee): Call it.
8681 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8684 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8685 -fcf-protection=branch -mibt.
8686 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8688 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8691 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8692 arm_word_relocations.
8694 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8697 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8698 extern array declarations.
8700 2018-03-27 Richard Biener <rguenther@suse.de>
8703 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8704 explicit single_use checks.
8706 2018-03-27 Richard Biener <rguenther@suse.de>
8708 PR tree-optimization/85082
8709 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8712 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8714 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8715 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8716 Turn on fasynchronous-unwind-tables and funwind-tables.
8718 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8721 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8722 (*bmi_blsr_<mode>_ccz): Ditto.
8724 2018-03-26 Tom de Vries <tom@codesourcery.com>
8726 PR tree-optimization/85063
8727 * omp-general.c (offloading_function_p): New function. Factor out
8729 * omp-offload.c (pass_omp_target_link::gate): ... here.
8730 * omp-general.h (offloading_function_p): Declare.
8731 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8732 with attribute omp declare target for offloading functions.
8734 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8736 PR tree-optimization/84005
8737 * tree-data-ref.h (get_base_for_alignment): Declare.
8738 * tree-data-ref.c (get_base_for_alignment_1): New function.
8739 (get_base_for_alignment): Likewise.
8740 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8741 get_base_for_alignment to find a suitable base object, instead
8742 of always using drb->base_address.
8744 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8747 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8748 known size by default.
8750 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8753 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8754 and non BLKmode operands.
8756 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8759 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8760 Clean up attributes.
8762 2018-03-23 Richard Biener <rguenther@suse.de>
8765 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8766 we are going to emit early debug for LTO.
8768 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8771 * function.c (match_asm_constraints_1): Don't optimize if input
8772 doesn't satisfy general_operand predicate for output's mode.
8775 * alias.c (write_dependence_p): Don't require for x_canonicalized
8776 non-VOIDmode if x has VOIDmode.
8779 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8780 just don't try to optimize it rather than assert it never happens.
8782 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8784 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8785 macro expansions for definition of ST_INTERNAL_<mode> and
8786 LD_INTERNAL_<mode> builtins.
8787 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8789 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8791 (altivec_expand_st_builtin): Likewise.
8792 (altivec_expand_builtin): Remove calls to deleted functions.
8793 (rs6000_address_for_altivec): Delete this function.
8794 * config/rs6000/vector.md: Remove expands for
8795 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8797 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8800 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8801 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8802 re-computing once computed.
8803 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8804 (arm_init_machine_status): Initialize
8805 machine->static_chain_stack_bytes.
8807 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8810 * doc/extend.texi: Add four new prototypes for vec_ld.
8811 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8812 definitions for more logical presentation.
8813 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8814 entries for V1TI variants of __builtin_altivec_ld builtin.
8815 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8816 handling of V1TI variant of LVX icode pattern.
8817 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8818 (rs6000_gimple_fold_builtin): Likewise.
8819 (altivec_init_builtins): Add code to define
8820 __builtin_altivec_lvx_v1ti function.
8822 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8825 * function.c (match_asm_constraints_1): Don't do the optimization
8826 if input isn't a REG, SUBREG, MEM or constant.
8828 2018-03-22 Tom de Vries <tom@codesourcery.com>
8830 PR tree-optimization/84956
8831 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8832 bb_has_abnormal_pred.
8834 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8837 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8838 DECL_INITIAL (decl) to decl at the end.
8839 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8842 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8844 * doc/extend.texi (__builtin_tgmath): Document when complex
8845 integer types are treated as _Complex _Float64.
8847 2018-03-21 Tom de Vries <tom@codesourcery.com>
8849 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8851 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8853 PR tree-optimization/84960
8854 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8855 if it is ENTRY block, move them into single succ of ENTRY in that case.
8857 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8859 PR tree-optimization/84811
8860 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8861 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8862 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8863 (known_subrange_p): Update accordingly. Cast each value involved
8864 in the size comparison, rather than casting the result of the
8867 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8869 PR tree-optimization/84982
8870 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8871 by flipping the least significant bit rather than all bits from
8872 bitpos to bitpos + bitsize - 1.
8874 2018-03-21 Nathan Sidwell <nathan@acm.org>
8876 * doc/extend.texi (Deprecated Features): Remove mention of
8877 long-deleted deprecations.
8879 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8882 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8883 * configure: Regenerate.
8885 2018-03-21 Tom de Vries <tom@codesourcery.com>
8887 PR tree-optimization/83126
8888 * tree-parloops.c (num_phis): New function.
8889 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8891 2018-03-21 Nathan Sidwell <nathan@acm.org>
8893 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8894 mention anon-struct/union members and trailing attributes.
8896 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8898 PR tree-optimization/84969
8899 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8900 builtin memset partitions if they set different rhs values.
8902 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8904 PR rtl-optimization/84989
8905 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8906 VEC_DUPLICATE with scalar result mode.
8908 2018-03-21 Martin Liska <mliska@suse.cz>
8911 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8912 not intended return statement.
8914 2018-03-21 Martin Liska <mliska@suse.cz>
8917 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8918 (chkp_find_bound_slots_1): Limit number of iterations.
8920 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8923 * Minor grammar fixes for x86 options.
8925 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8928 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8929 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8931 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8934 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8935 (altivec_lvx_<mode>_1op): Likewise.
8936 (altivec_stvx_<mode>_2op): Likewise.
8937 (altivec_stvx_<mode>_1op): Likewise.
8938 (altivec_lvx_<VM2:mode>): New define_expand.
8939 (altivec_stvx_<VM2:mode>): Likewise.
8940 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8941 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8942 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8943 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8944 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8945 (rs6000_gen_lvx): Likewise.
8946 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8947 (altivec_expand_stv_builtin): Likewise.
8948 (altivec_expand_builtin): Likewise.
8949 * config/rs6000/vector.md: Likewise.
8951 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8954 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8957 2018-03-20 Richard Biener <rguenther@suse.de>
8960 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8961 sign-conversions as zero, fall back to standard scalar_stmt
8964 2018-03-20 Martin Liska <mliska@suse.cz>
8967 * predict.c (rebuild_frequencies): Handle case when we have
8968 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8970 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8973 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8974 flag_section_anchors.
8975 * varasm.c (use_blocks_for_decl_p): Remove hack for
8976 dw2_force_const_mem.
8979 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8981 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8982 be created, use lowpart_subreg of operands[0] rather than operands[0]
8984 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8985 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8986 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8987 and n constraint instead of aarch64_shift_imm_di and Usd.
8988 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8989 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8991 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8994 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8995 to favor GPR over NEON registers.
8996 (<shift>di3_neon): Likewise.
8998 2018-03-20 Tom de Vries <tom@codesourcery.com>
9001 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9002 (nvptx_process_pars): Emit bar.sync asap and alap.
9004 2018-03-20 Tom de Vries <tom@codesourcery.com>
9007 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9008 seen_label if seen_label is already set.
9010 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9013 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9014 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9015 Use 1U instead of 1. Formatting fixes.
9018 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9019 instead of TREE_TYPE (s1) for the return value.
9021 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9023 PR tree-optimization/84946
9024 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9025 bitsize + bitsize in poly_uint64 rather than poly_int64.
9028 * dwarf2asm.c: Include fold-const.c.
9029 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9030 of decl rather than decl itself.
9032 PR rtl-optimization/84643
9033 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9035 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9038 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9039 calling assemble_variable.
9041 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9044 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9045 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9047 2018-03-19 Jim Wilson <jimw@sifive.com>
9050 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9051 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9052 (riscv_first_stack_step): Likewise.
9053 (riscv_option_override): Use STACK_BOUNDARY instead of
9055 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9057 (BIGGEST_ALIGNMENT): Set to 128.
9058 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9059 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9062 2018-03-19 Richard Biener <rguenther@suse.de>
9064 PR tree-optimization/84933
9065 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9066 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9068 2018-03-19 Richard Biener <rguenther@suse.de>
9070 PR tree-optimization/84859
9071 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9072 (cond_if_else_store_replacement): Perform sinking operation on
9073 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9074 Generalize what a BB with a single eligible store is.
9076 2018-03-19 Richard Biener <rguenther@suse.de>
9078 PR tree-optimization/84929
9079 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9080 chrec_is_positive against non-chrec arg.
9082 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9085 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9087 2018-03-18 Martin Liska <mliska@suse.cz>
9089 PR rtl-optimization/84635
9090 * regrename.c (build_def_use): Use matches_mode only when
9093 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9095 PR tree-optimization/84913
9096 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9097 vectorize chains of COND_EXPRs.
9099 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9101 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9103 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9105 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9107 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9109 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9111 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9112 Kito Cheng <kito.cheng@gmail.com>
9114 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9115 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9116 (nds32_adjust_reg_alloc_order): New function.
9117 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9119 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9121 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9122 nds32_print_operand, nds32_print_operand_address): Use
9123 HOST_WIDE_INT_PRINT_DEC instead.
9125 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9127 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9129 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9132 * config/i386/i386.c (initial_ix86_tune_features,
9133 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9135 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9136 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9137 rather than 1u << ix86_tune. Formatting fix.
9138 (ix86_option_override_internal): Change ix86_arch_mask from
9139 unsigned int to unsigned HOST_WIDE_INT, initialize to
9140 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9141 (ix86_function_specific_restore): Likewise.
9143 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9146 * postreload.c (reload_combine_recognize_pattern): Perform
9147 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9148 truncate_int_for_mode the result for the destination's mode.
9151 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9152 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9154 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9157 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9160 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9161 regno_allocno_class_array and sorted_pseudos.
9162 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9163 insns where regno is used.
9165 2018-03-16 Martin Liska <mliska@suse.cz>
9168 * multiple_target.c (create_dispatcher_calls): Redirect
9169 reference in the symbol table.
9171 2018-03-16 Martin Liska <mliska@suse.cz>
9174 * multiple_target.c (create_dispatcher_calls): Redirect also
9177 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9181 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9182 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9183 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9185 2018-03-16 Julia Koval <julia.koval@intel.com>
9187 * doc/invoke.texi (Skylake Server): Add CLWB.
9188 Cannonlake): Remove CLWB.
9190 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9192 PR tree-optimization/84841
9193 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9195 (FLOAT_ONE_CONST_TYPE): Define.
9196 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9197 (sort_by_operand_rank): Put entries with higher constant_type last
9198 rather than first to match comments.
9200 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9202 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9205 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9208 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9209 check and use address of target always.
9211 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9214 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9215 (indirect_thunk_bnd_needed): Likewise.
9216 (indirect_thunks_used): Likewise.
9217 (indirect_thunks_bnd_used): Likewise.
9218 (indirect_return_needed): New.
9219 (indirect_return_bnd_needed): Likewise.
9220 (output_indirect_thunk_function): Add a bool argument for
9222 (output_indirect_thunk_function): Don't generate alias for
9223 function return thunk.
9224 (ix86_code_end): Call output_indirect_thunk_function to generate
9225 function return thunks.
9226 (ix86_output_function_return): Set indirect_return_bnd_needed
9227 and indirect_return_needed instead of indirect_thunk_bnd_needed
9228 and indirect_thunk_needed.
9230 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9232 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9233 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9234 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9236 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9237 Paul Hua <paul.hua.gm@gmail.com>
9240 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9242 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9244 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9245 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9248 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9251 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9252 instead of GET_MODE_SIZE when comparing Units.
9254 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9257 * varasm.c (hash_section): Return an unchangeble hash value
9258 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9259 Return !aarch64_can_use_per_function_literal_pools_p ().
9261 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9264 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9265 rather than address of cmode as last argument to prepare_cmp_insn.
9267 2018-03-15 Julia Koval <julia.koval@intel.com>
9269 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9270 F_AVX512VNNI, F_AVX512BITALG): New.
9272 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9275 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9276 insn for floating-point loads and stores.
9278 2018-03-14 Carl Love <cel@us.ibm.com>
9280 * config/rs6000/rs6000-c.c: Add macro definitions for
9281 ALTIVEC_BUILTIN_VEC_PERMXOR.
9282 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9283 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9284 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9285 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9287 * config/doc/extend.texi: Add prototypes for vec_permxor.
9289 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9292 * diagnostic-show-locus.c (class layout_point): Convert m_line
9293 from int to linenum_type.
9294 (line_span::comparator): Use linenum "compare" function when
9295 comparing line numbers.
9296 (test_line_span): New function.
9297 (layout_range::contains_point): Convert param "row" from int to
9299 (layout_range::intersects_line_p): Likewise.
9300 (layout::will_show_line_p): Likewise.
9301 (layout::print_source_line): Likewise.
9302 (layout::should_print_annotation_line_p): Likewise.
9303 (layout::print_annotation_line): Likewise.
9304 (layout::print_leading_fixits): Likewise.
9305 (layout::annotation_line_showed_range_p): Likewise.
9306 (struct line_corrections): Likewise for field m_row.
9307 (line_corrections::line_corrections): Likewise for param "row".
9308 (layout::print_trailing_fixits): Likewise.
9309 (layout::get_state_at_point): Likewise.
9310 (layout::get_x_bound_for_row): Likewise.
9311 (layout::print_line): Likewise.
9312 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9313 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9314 * input.c (selftest::test_linenum_comparisons): New function.
9315 (selftest::input_c_tests): Call it.
9316 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9317 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9318 * selftest.h (ASSERT_GT): New macro.
9319 (ASSERT_GT_AT): New macro.
9320 (ASSERT_LT): New macro.
9321 (ASSERT_LT_AT): New macro.
9323 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9325 PR rtl-optimization/84780
9326 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9328 2018-03-14 Martin Liska <mliska@suse.cz>
9330 * tree.c (record_node_allocation_statistics): Use
9331 get_stats_node_kind.
9332 (get_stats_node_kind): New function extracted from
9333 record_node_allocation_statistics.
9334 (free_node): Use get_stats_node_kind.
9336 2018-03-14 Richard Biener <rguenther@suse.de>
9338 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9339 that the value-set of ANTIC_IN doesn't grow.
9342 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9344 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9345 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9347 2018-03-14 Julia Koval <julia.koval@intel.com>
9349 * config.gcc (icelake-client, icelake-server): New.
9351 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9352 (initial_ix86_arch_features): Ditto.
9353 (PTA_SKYLAKE): Add SGX.
9354 (PTA_ICELAKE): Remove.
9355 (PTA_ICELAKE_CLIENT): New.
9356 (PTA_ICELAKE_SERVER): New.
9357 (ix86_option_override_internal): Split up icelake on icelake client and
9359 (get_builtin_code_for_version): Ditto.
9360 (fold_builtin_cpu): Ditto.
9361 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9362 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9363 * config/i386/i386.h (processor_type): Ditto.
9364 * doc/invoke.texi: Ditto.
9366 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9369 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9370 INTEGER_CST offset, add it together with bitpos / 8 and
9371 sign extend based on POINTER_SIZE.
9375 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9378 * config/i386/constraints.md (Yc): New register constraint.
9379 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9380 Use Yc constraint for alternative 2 of operand 0. Remove
9381 preferred_for_speed attribute.
9383 2018-03-14 Richard Biener <rguenther@suse.de>
9385 PR tree-optimization/84830
9386 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9387 with the old one to avoid oscillations.
9389 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9392 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9394 (assign_by_spills): Return a flag of reload assignment failure.
9395 Do not process the reload assignment failures. Do not spill other
9396 reload pseudos if they has the same reg class. Update n if
9398 (lra_assign): Add a return arg. Set up from the result of
9399 assign_by_spills call.
9400 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9401 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9402 usage_insns if it is not NULL.
9403 (spill_hard_reg_in_range): New function.
9404 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9405 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9406 function prototypes.
9407 (lra_assign): Change prototype.
9408 * lra.c (lra): Add code to deal with fails by splitting hard reg
9411 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9413 * config/riscv/riscv.opt (mrelax): New option.
9414 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9415 "norelax" when riscv_mrelax is disabled.
9416 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9418 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9421 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9422 reassociation for int modes.
9424 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9426 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9427 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9429 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9430 * config/aarch64/aarch64-sve.md
9431 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9432 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9433 (*extend<mode><Vwide>2): Rename to...
9434 (aarch64_sve_extend<mode><Vwide>2): ...this.
9435 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9436 renaming the old pattern to...
9437 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9439 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9440 define_expand, renaming the old pattern to...
9441 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9442 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9443 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9444 account when deciding which SVE instruction the optab should use.
9445 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9447 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9449 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9450 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9451 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9452 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9453 (tlsdesc_small_<mode>): Turn a define_expand and use
9454 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9455 (tlsdesc_small_advsimd_<mode>): ...this.
9456 (tlsdesc_small_sve_<mode>): New pattern.
9458 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9460 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9461 (UNSPEC_UMUL_HIGHPART): New constants.
9462 (MUL_HIGHPART): New int iteraor.
9463 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9464 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9466 (*<su>mul<mode>3_highpart): New define_insn.
9468 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9471 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9474 2018-03-13 Martin Liska <mliska@suse.cz>
9477 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9479 (sem_item_optimizer::~sem_item_optimizer): Release it.
9480 (sem_item_optimizer::merge_classes): Register variable aliases.
9481 (sem_item_optimizer::fixup_pt_set): New function.
9482 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9483 * ipa-icf.h: Declare new variables and functions.
9485 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9488 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9489 integer_pow2p@2 and test integer_pow2p in condition.
9490 (A < 0 ? C : 0): Similarly for @1.
9493 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9494 characters starting at p contain '\0' character, don't look beyond
9498 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9499 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9502 * reg-stack.c (change_stack): Change update_end var from int to
9503 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9504 also call set_block_for_insn on the newly added insns and rescan.
9507 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9508 on the last operand.
9511 * tree.c (stabilize_reference_1): Return save_expr (e) for
9512 STATEMENT_LIST even if it doesn't have side-effects.
9514 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9516 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9518 2018-03-12 Renlin Li <renlin.li@arm.com>
9520 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9521 aarch64_output_scalar_simd_mov_immediate.
9523 2018-03-12 Martin Sebor <msebor@redhat.com>
9525 PR tree-optimization/83456
9526 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9527 for perfectly overlapping calls to memcpy.
9528 (gimple_fold_builtin_memory_chk): Same.
9529 (gimple_fold_builtin_strcpy): Handle no-warning.
9530 (gimple_fold_builtin_stxcpy_chk): Same.
9531 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9533 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9535 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9536 parameter. Use it for SFmode.
9537 (rs6000_function_arg_advance_1): Adjust.
9538 (rs6000_function_arg): Adjust.
9539 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9541 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9543 PR rtl-optimization/84169
9544 PR rtl-optimization/84780
9545 * combine.c (can_combine_p): Check for a 2-insn combination whether
9546 the destination register is used between the two insns, too.
9548 2018-03-12 Richard Biener <rguenther@suse.de>
9550 PR tree-optimization/84803
9551 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9552 for refs DR analysis didn't process.
9554 2018-03-12 Richard Biener <rguenther@suse.de>
9556 PR tree-optimization/84777
9557 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9558 force-vectorize loops ignore whether we are optimizing for size.
9560 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9562 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9563 (TARGET_MD_ASM_ADJUST): Define.
9565 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9566 Kito Cheng <kito.cheng@gmail.com>
9567 Chung-Ju Wu <jasonwucj@gmail.com>
9569 * config/nds32/nds32.c (nds32_compute_stack_frame,
9570 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9571 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9572 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9573 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9574 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9575 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9576 * config/nds32/nds32.md (prologue, epilogue): Use macro
9577 NDS32_V3PUSH_AVAILABLE_P to do checking.
9579 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9582 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9583 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9584 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9585 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9586 addition of most attributes on !orig_type_die or the attribute not
9587 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9589 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9590 Chung-Ju Wu <jasonwucj@gmail.com>
9592 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9594 * config/nds32/nds32.opt (mvh): New option.
9596 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9597 Chung-Ju Wu <jasonwucj@gmail.com>
9599 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9601 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9602 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9605 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9606 Chung-Ju Wu <jasonwucj@gmail.com>
9608 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9610 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9611 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9613 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9614 Kito Cheng <kito.cheng@gmail.com>
9615 Chung-Ju Wu <jasonwucj@gmail.com>
9617 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9618 UNSPEC_FFMISM and UNSPEC_FLMISM.
9619 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9620 for ffb, ffmism and flmism.
9621 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9622 (unspec_ffmism): Ditto.
9623 (unspec_flmism): Ditto.
9624 (nds32_expand_builtin_impl): Check if string extension is available.
9625 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9626 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9628 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9631 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9634 * lra-assigns.c (assign_by_spills): Return a flag of reload
9635 assignment failure. Do not process the reload assignment
9636 failures. Do not spill other reload pseudos if they has the same
9638 (lra_assign): Add a return arg. Set up from the result of
9639 assign_by_spills call.
9640 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9641 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9642 usage_insns if it is not NULL.
9643 (spill_hard_reg_in_range): New function.
9644 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9645 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9646 function prototypes.
9647 (lra_assign): Change prototype.
9648 * lra.c (lra): Add code to deal with fails by splitting hard reg
9651 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9654 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9656 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9659 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9660 (dw_val_node): Add val_symbolic_view.
9661 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9662 (symview_upper_bound): New.
9663 (new_line_info_table): Initialize symviews_since_reset.
9664 (dwarf2out_source_line): Count symviews_since_reset and set
9665 symview_upper_bound.
9666 (dw_val_equal_p): Handle symview.
9667 (add_AT_symview): New.
9668 (print_dw_val): Handle symview.
9669 (attr_checksum, attr_checksum_ordered): Likewise.
9670 (same_dw_val_p, size_of_die): Likewise.
9671 (value_format, output_die): Likewise.
9672 (add_high_low_attributes): Use add_AT_symview for entry_view.
9673 (dwarf2out_finish): Reset symview_upper_bound, clear
9676 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9679 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9680 Add strict argument and use it.
9681 (rs6000_split_multireg_move): Update for new strict argument.
9682 (mem_operand_gpr): Disallow all non-offsettable addresses.
9683 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9685 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9688 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9689 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9690 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9693 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9694 decl, use remap_type if we want to use the type.
9696 2018-03-09 Martin Sebor <msebor@redhat.com>
9698 PR tree-optimization/84526
9699 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9701 (builtin_access::generic_overlap): Be prepared to handle non-array
9704 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9706 PR rtl-optimization/84682
9707 * lra-constraints.c (process_address_1): Check is_address flag
9708 for address constraints.
9709 (process_alt_operands): Likewise.
9710 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9711 preprocess_constraints.
9712 * recog.h (preprocess_constraints): Add oploc parameter.
9714 * recog.c (preprocess_constraints): Test address_operand for
9715 CT_ADDRESS constraints.
9717 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9720 * lra-assigns.c (assign_by_spills): Return a flag of reload
9721 assignment failure. Do not process the reload assignment
9722 failures. Do not spill other reload pseudos if they has the same
9724 (lra_assign): Add a return arg. Set up from the result of
9725 assign_by_spills call.
9726 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9727 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9728 usage_insns if it is not NULL.
9729 (spill_hard_reg_in_range): New function.
9730 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9731 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9732 function prototypes.
9733 (lra_assign): Change prototype.
9734 * lra.c (lra): Add code to deal with fails by splitting hard reg
9737 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9740 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9741 Accept complain bool parameter. Only emit errors if it is true.
9742 (arm_parse_cpu_option_name): Likewise.
9743 (arm_target_thumb_only): Adjust callers of the above.
9744 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9745 prototype to take a default true bool parameter.
9746 (arm_parse_arch_option_name): Likewise.
9748 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9749 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9753 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9754 * configure: Regenerate.
9755 * configure.ac ("linker --version-script option"): New.
9756 ("linker soname option"): New.
9758 2018-03-09 Richard Biener <rguenther@suse.de>
9760 PR tree-optimization/84775
9761 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9762 immediate uses of predicate stmts and mark them modified.
9765 PR tree-optimization/84178
9766 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9768 (version_loop_for_if_conversion): Delay update_ssa call.
9769 (tree_if_conversion): Delay update_ssa until after predicate
9772 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9775 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9776 when the function accesses prior frames.
9778 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9781 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9782 gen_llsym, otherwise call maybe_gen_llsym.
9785 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9786 has ',' character inside of it.
9788 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9791 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9792 as clobbering CC_REGNUM.
9794 2018-03-08 Richard Biener <rguenther@suse.de>
9797 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9798 (follow_copies_to_constant): Do not follow SSA names registered
9801 2018-03-08 Richard Biener <rguenther@suse.de>
9803 PR tree-optimization/84178
9804 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9806 (version_loop_for_if_conversion): Delay update_ssa call.
9807 (tree_if_conversion): Delay update_ssa until after predicate
9810 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9812 PR tree-optimization/84178
9813 * tree-if-conv.c (release_bb_predicate): Remove the
9814 the assertion that the stmts have NULL use_ops.
9815 Discard the statements, asserting that they haven't
9816 yet been added to a BB.
9818 2018-03-08 Richard Biener <rguenther@suse.de>
9820 PR tree-optimization/84746
9821 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9822 (phi_translate): Pass in destination ANTIC_OUT set.
9823 (phi_translate_1): Likewise. For a simplified result lookup
9824 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9825 (phi_translate_set): Adjust.
9826 (do_pre_regular_insertion): Likewise.
9827 (do_pre_partial_partial_insertion): Likewise.
9829 2018-03-08 Martin Liska <mliska@suse.cz>
9831 PR gcov-profile/84735
9832 * doc/gcov.texi: Document usage of profile files.
9833 * gcov-io.h: Document changes in the format.
9835 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9839 * dwarf2out.c (struct dw_line_info_table): Update comments for
9841 (FORCE_RESET_NEXT_VIEW): New.
9842 (FORCE_RESETTING_VIEW_P): New.
9843 (RESETTING_VIEW_P): Check for -1 too.
9844 (ZERO_VIEW_P): Likewise.
9845 (new_line_info_table): Force-reset next view.
9846 (dwarf2out_begin_function): Likewise.
9847 (dwarf2out_source_line): Simplify zero_view_p initialization.
9848 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9849 view directly. Omit view when omitting .loc at line 0.
9851 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9853 PR tree-optimization/84740
9854 * tree-switch-conversion.c (process_switch): Call build_constructors
9855 only if info.phi_count is non-zero.
9857 PR tree-optimization/84739
9858 * tree-tailcall.c (find_tail_calls): Check call arguments against
9859 DECL_ARGUMENTS (current_function_decl) rather than
9860 DECL_ARGUMENTS (func) when checking for tail recursion.
9862 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9864 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9865 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9866 Volker Reichelt's entry and add entries for people that perform
9867 GCC fuzzy testing and report numerous bugs.
9869 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9872 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9873 readonly data in sdata, if that is disabled.
9874 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9875 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9876 -mreadonly-in-sdata option.
9878 2018-03-07 Martin Sebor <msebor@redhat.com>
9880 PR tree-optimization/84468
9881 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9882 basic block when looking for nul assignment.
9884 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9887 * except.h (output_function_exception_table): Adjust prototype.
9888 * except.c (output_function_exception_table): Remove FNNAME parameter
9889 and add SECTION parameter. Ouput one part of the table at a time.
9890 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9891 the first part of the exception table and emit unwind directives.
9892 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9893 (i386_pe_seh_cold_init): Likewise.
9894 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9895 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9896 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9897 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9898 * config/i386/winnt.c: Include except.h.
9899 (struct seh_frame_state): Add reg_offset, after_prologue and
9900 in_cold_section fields.
9901 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9902 (i386_pe_seh_cold_init): New function.
9903 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9904 to seh->in_cold_section.
9905 (seh_emit_push): Record the offset of the push.
9906 (seh_emit_save): Record the offet of the save.
9907 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9908 Test seh->after_prologue to disregard the epilogue.
9909 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9910 (i386_pe_end_cold_function): New function.
9912 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9915 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9916 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9919 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9921 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9922 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9925 * multiple_target.c: Include tree-inline.h and intl.h.
9926 (expand_target_clones): Diagnose and fail if node->definition and
9927 !tree_versionable_function_p (node->decl).
9929 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9931 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9933 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9934 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9935 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9937 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9940 * combine.c (try_combine): Use reg_or_subregno instead of handling
9941 just paradoxical SUBREGs and REGs.
9943 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9945 * config/arc/arc.c (arc_finalize_pic): Remove function.
9946 (arc_must_save_register): We use single base PIC register, remove
9947 checks to save/restore the PIC register.
9948 (arc_expand_prologue): Likewise.
9949 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9951 (arc_verify_short): Likewise.
9952 (arc_attr_type): Likewise.
9953 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9954 (walk_stores): Likewise.
9955 (arc_address_cost): Make it static.
9956 (arc_verify_short): Likewise.
9957 (branch_dest): Likewise.
9958 (arc_attr_type): Likewise.
9959 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9960 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9961 (arc_final_prescan_insn): Remove inserting the nops due to
9962 hardware hazards. It is done in reorg step.
9963 (insn_length_variant_t): Remove.
9964 (insn_length_parameters_t): Likewise.
9965 (arc_insn_length_parameters): Likewise.
9966 (arc_get_insn_variants): Likewise.
9967 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9969 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9972 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9975 PR tree-optimization/84687
9976 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9978 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9980 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9982 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9983 Rename to ppc_speculation_barrier.
9984 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9985 __builtin_ppc_speculation_barrier.
9987 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9990 * combine.c (combine_simplify_rtx): Don't try to simplify if
9991 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9994 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9996 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9997 to 32 bytes when compiling for POWER9.
9999 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10002 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10003 regparm >= 3 with no arg reg available also for calls with
10004 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10007 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10009 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10011 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10014 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10016 2018-03-05 Richard Biener <rguenther@suse.de>
10018 PR tree-optimization/84486
10019 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10020 When inserting a __builtin_assume_aligned call set the LHS
10021 SSA name alignment info accordingly.
10023 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10025 PR tree-optimization/84114
10026 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10027 Avoid reassociation of FLOAT_MODE addition.
10029 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10031 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10032 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10033 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10034 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10035 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10036 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10037 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10039 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10040 __builtin_ia32_wbinvd): New builtins.
10041 (SPECIAL_ARGS2): New.
10042 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10043 (SPECIAL_ARGS2): New.
10044 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10045 (ix86_valid_target_attribute_inner_p): Ditto.
10046 (ix86_init_mmx_sse_builtins): Add special_args2.
10047 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10048 TARGET_WBNOINVD_P): New.
10049 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10050 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10051 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10052 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10053 * config/i386/pconfigintrin.h: New file.
10054 * config/i386/wbnoinvdintrin.h: Ditto.
10055 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10057 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10059 2018-03-05 Richard Biener <rguenther@suse.de>
10061 PR tree-optimization/84670
10062 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10064 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10065 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10066 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10067 doesn't grow if all successors (recursively) were visited at least
10070 2018-03-05 Richard Biener <rguenther@suse.de>
10072 PR tree-optimization/84650
10073 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10074 if executed in the loop pipeline.
10076 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10078 * doc/configfiles.texi (Configuration Files): Move info about
10079 conditionalizing $target-protos.h to...
10080 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10081 differs from $target-protos.h.
10083 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10084 Chung-Ju Wu <jasonwucj@gmail.com>
10086 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10087 * config/nds32/nds32-multiple.md (setmemsi): Define.
10088 * config/nds32/nds32-memory-manipulation.c
10089 (nds32_gen_dup_4_byte_to_word_value): New.
10090 (emit_setmem_word_loop): New.
10091 (emit_setmem_byte_loop): New.
10092 (nds32_expand_setmem_loop): New.
10093 (nds32_expand_setmem_loop_v3m): New.
10094 (nds32_expand_setmem_unroll): New.
10095 (nds32_expand_setmem): New.
10097 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10098 Chung-Ju Wu <jasonwucj@gmail.com>
10100 * config/nds32/nds32-memory-manipulation.c
10101 (nds32_emit_load_store): New.
10102 (nds32_emit_post_inc_load_store): New.
10103 (nds32_emit_mem_move): New.
10104 (nds32_emit_mem_move_block): New.
10105 (nds32_expand_movmemsi_loop_unknown_size): New.
10106 (nds32_expand_movmemsi_loop_known_size): New.
10107 (nds32_expand_movmemsi_loop): New.
10108 (nds32_expand_movmemsi_unroll): New.
10109 (nds32_expand_movmemqi): Rename ...
10110 (nds32_expand_movmemsi): ... to this.
10111 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10112 (movmemsi): ... to this.
10113 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10114 (nds32_expand_movmemsi): ... to this.
10116 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10117 Monk Chiang <sh.chiang04@gmail.com>
10118 Chung-Ju Wu <jasonwucj@gmail.com>
10120 * config/nds32/nds32-protos.h
10121 (nds32_expand_load_multiple): New arguments.
10122 (nds32_expand_store_multiple): Ditto.
10123 (nds32_valid_multiple_load_store): Rename ...
10124 (nds32_valid_multiple_load_store_p): ... to this.
10125 * config/nds32/nds32-memory-manipulation.c
10126 (nds32_expand_load_multiple): Refine implementation.
10127 (nds32_expand_store_multiple): Ditto.
10128 * config/nds32/nds32-multiple.md
10129 (load_multiple): Update nds32_expand_load_multiple interface.
10130 (store_multiple): Update nds32_expand_store_multiple interface.
10131 * config/nds32/nds32-predicates.c
10132 (nds32_valid_multiple_load_store): Rename ...
10133 (nds32_valid_multiple_load_store_p): ... to this and refine
10135 * config/nds32/predicates.md
10136 (nds32_load_multiple_and_update_address_operation): New predicate.
10137 (nds32_store_multiple_and_update_address_operation): New predicate.
10139 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10140 Chung-Ju Wu <jasonwucj@gmail.com>
10142 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10143 (combo): New attribute.
10144 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10146 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10148 * config/nds32/nds32.opt: Change -mcmodel= default value.
10150 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10151 Monk Chiang <sh.chiang04@gmail.com>
10152 Chung-Ju Wu <jasonwucj@gmail.com>
10154 * config/nds32/constants.md (unspec_element): New enum.
10155 * config/nds32/constraints.md (Umw): New constraint.
10156 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10157 * config/nds32/nds32-intrinsic.md: Likewise.
10158 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10159 (nds32_valid_smw_lwm_base_p): New.
10160 (nds32_output_smw_single_word): New.
10161 (nds32_output_lmw_single_word): New.
10162 (nds32_expand_unaligned_load): New.
10163 (nds32_expand_unaligned_store): New.
10164 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10165 (nds32_output_smw_single_word): Declare.
10166 (nds32_output_lmw_single_word): Declare.
10167 (nds32_expand_unaligned_load): Declare.
10168 (nds32_expand_unaligned_store): Declare.
10169 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10170 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10171 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10172 NDS32_BUILTIN_UASTORE_DW.
10173 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10176 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10177 Kito Cheng <kito.cheng@gmail.com>
10178 Chung-Ju Wu <jasonwucj@gmail.com>
10180 * config/nds32/nds32-intrinsic.c
10181 (nds32_expand_builtin_null_ftype_reg): Delete.
10182 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10183 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10184 (nds32_read_argument): New.
10185 (nds32_legitimize_target): Ditto.
10186 (nds32_legitimize_argument): Ditto.
10187 (nds32_check_constant_argument): Ditto.
10188 (nds32_expand_unop_builtin): Ditto.
10189 (nds32_expand_unopimm_builtin): Ditto.
10190 (nds32_expand_binop_builtin): Ditto.
10191 (nds32_builtin_decl_impl): Ditto.
10192 (builtin_description): Ditto.
10193 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10194 (nds32_init_builtins_impl): Ditto.
10195 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10196 (nds32_builtin_decl): New.
10197 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10198 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10200 2018-03-02 Jeff Law <law@redhat.com>
10202 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10203 (redundant_insn, fill_simple_delay_slots): Likewise.
10204 (fill_slots_from_thread): Likewise.
10205 * resource.c (mark_referenced_resources): Likewise.
10206 (mark_set_resources, find_dead_or_set_registers): Likewise.
10208 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10210 * substring-locations.h (format_warning_va): Formatting fix for
10211 ATTRIBUTE_GCC_DIAG.
10212 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10214 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10215 * substring-locations.c: Include intl.h.
10216 (format_warning_va): Turned into small wrapper around
10217 format_warning_n_va, renamed to ...
10218 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10219 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10221 (format_warning_at_substring_n): New function.
10222 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10223 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10224 format_warning_at_substring with just a shorter name instead of
10225 const function pointer.
10226 (fmtwarn_n): New function.
10227 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10228 appropriate, get rid of all the fmtstr temporaries, move conditionals
10229 with G_() wrapped string literals directly into fmtwarn arguments,
10230 cast dir.len to (int), formatting fixes.
10232 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10234 * doc/invoke.texi: Remove "Cilk Plus" references.
10236 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10237 Richard Biener <rguenther@suse.de>
10240 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10241 for error or warning attributes if CALL_FROM_THUNK_P is set.
10244 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10247 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10248 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10251 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10252 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10254 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10255 instead of -1U in last predictors element's probability member.
10257 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10260 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10261 arguments if they are comparable.
10263 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10265 PR tree-optimization/84634
10266 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10267 masks and masked_loop_p with a single loop_masks, making sure it's
10268 null for bb vectorization.
10270 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10272 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10273 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10274 loop->force_vectorize to check whether there is no alias.
10276 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10279 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10281 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10283 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10284 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10285 instead of a loop around prev_real_insn.
10286 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10289 PR inline-asm/84625
10290 * config/i386/i386.c (ix86_print_operand): Use conditional
10291 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10294 2018-03-02 Richard Biener <rguenther@suse.de>
10296 PR tree-optimization/84427
10297 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10298 (bitmap_set_subtract_values): Rewrite to handle multiple
10301 (prune_clobbered_mems): Likewise.
10302 (phi_translate): Take edge instead of pred/phiblock.
10303 (phi_translate_1): Likewise.
10304 (phi_translate_set): Likewise. Insert all translated
10305 exprs for a value into the set, keeping possibly multiple
10306 expressions per value.
10307 (compute_antic_aux): Adjust for phi_translate changes.
10308 When intersecting union the expressions and prune those
10309 not in the final value set, keeping possibly multiple
10310 expressions per value. Do not use value-insertion
10311 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10312 all expressions. Add verification that the value-sets
10313 only shrink during iteration.
10314 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10315 (do_pre_regular_insertion): Likewise.
10316 (do_pre_partial_partial_insertion): Likewise.
10318 2018-03-02 Richard Biener <rguenther@suse.de>
10321 * config/darwin.c (saved_debug_info_level): New static global.
10322 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10323 (darwin_asm_lto_end): Restore debug info generation settings.
10325 2018-03-01 Martin Liska <mliska@suse.cz>
10328 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10329 volatile arguments.
10331 2018-03-01 Richard Biener <rguenther@suse.de>
10334 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10335 type annotation with LTO when debug was disabled at compile-time.
10337 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10339 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10341 (mips_final_postscan_insn): Likewise.
10343 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10345 PR rtl-optimization/84528
10346 * alias.c (init_alias_target): Add commentary.
10347 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10348 a unique base value if the frame pointer is not eliminated
10349 to the stack pointer.
10351 2018-03-01 Tom de Vries <tom@codesourcery.com>
10353 PR rtl-optimization/83327
10354 * lra-int.h (hard_regs_spilled_into): Declare.
10355 * lra.c (hard_regs_spilled_into): Define.
10356 (init_reg_info): Init hard_regs_spilled_into.
10357 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10358 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10359 (process_bb_lives): Handle hard_regs_spilled_into.
10360 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10361 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10363 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10365 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10366 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10367 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10368 * config/rs6000/aix72.h: New file.
10370 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10372 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10373 instead of warning_at with conditional singular and plural messages
10377 * stor-layout.c (update_alignment_for_field): For
10378 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10379 && !DECL_PACKED (field), do the alignment update, just use
10380 only desired_align instead of MAX (type_align, desired_align)
10382 (place_field): Don't do known_align < desired_align handling
10383 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10384 is non-NULL, instead do it after rli->prev_field handling and
10385 only if not within a bitfield word. For DECL_PACKED (field)
10386 use type_align of BITS_PER_UNIT.
10388 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10390 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10391 superfluous parentheses and trailing spaces.
10393 2018-02-28 Richard Biener <rguenther@suse.de>
10395 PR tree-optimization/84584
10396 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10397 SCoPs with fake exit edge.
10399 2018-02-28 Martin Liska <mliska@suse.cz>
10402 * timevar.c (timer::print): Fix format to properly print 100%
10405 2018-02-28 Richard Biener <rguenther@suse.de>
10407 PR middle-end/84607
10408 * genmatch.c (capture_info::walk_match): Do not mark
10409 captured expressions without operands as expr_p given
10410 they act more like predicates and should be subject to
10411 "lost tail" side-effect preserving.
10413 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10415 PR rtl-optimization/81611
10416 * auto-inc-dec.c (attempt_change): Move dead note from
10417 mem_insn if it's the next use of regno
10418 (find_address): Take address use of reg holding
10419 non-incremented value. Add parm to limit search to the named
10421 (merge_in_block): Attempt to use a mem insn that is the next
10422 use of the original regno.
10424 2018-02-27 Martin Sebor <msebor@redhat.com>
10427 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10428 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10430 2018-02-27 Martin Sebor <msebor@redhat.com>
10432 PR translation/84207
10433 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10434 n argument to unsigned HOST_WIDE_INT.
10435 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10436 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10437 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10438 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10440 2018-02-27 Richard Biener <rguenther@suse.de>
10442 PR tree-optimization/84512
10443 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10444 Do not use the estimate returned from record_stmt_cost for
10445 the scalar iteration cost but sum properly using add_stmt_cost.
10447 2018-02-27 Richard Biener <rguenther@suse.de>
10449 PR tree-optimization/84466
10450 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10451 Adjust last change to less strictly validate use operands.
10453 2018-02-27 Martin Liska <mliska@suse.cz>
10455 PR gcov-profile/84548
10456 * gcov.c (process_file): Allow partial overlap and consider it
10457 also as group functions.
10458 (output_lines): Properly calculate range of lines for a group.
10460 2018-02-27 Martin Liska <mliska@suse.cz>
10462 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10463 'ggc' suffixes. Change first column width.
10464 (timer::print): Fix formatting of the column.
10466 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10468 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10469 preserve inline entry blocks for the sake of debug inline
10470 entry point markers alone.
10471 (remove_unused_locals): Suggest in comments a better place to
10472 force the preservation of inline entry blocks that are
10473 otherwise unused, but do not preserve them.
10475 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10477 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10479 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10482 * config/i386/constraints.md (Bs): Replace
10483 ix86_indirect_branch_register with
10484 TARGET_INDIRECT_BRANCH_REGISTER.
10486 * config/i386/i386.md (indirect_jump): Likewise.
10487 (tablejump): Likewise.
10488 (*sibcall_memory): Likewise.
10489 (*sibcall_value_memory): Likewise.
10490 Peepholes of indirect call and jump via memory: Likewise.
10491 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10492 (*sibcall_value_GOT_32): Likewise.
10493 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10494 (GOT_memory_operand): Likewise.
10495 (call_insn_operand): Likewise.
10496 (sibcall_insn_operand): Likewise.
10497 (GOT32_symbol_operand): Likewise.
10498 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10500 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10502 PR rtl-optimization/83496
10503 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10504 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10505 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10506 redundant insn, if any.
10507 (relax_delay_slots): Likewise.
10508 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10510 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10512 PR tree-optimization/83965
10513 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10514 that grouped statements are part of a reduction chain. Return
10515 true if the statement is not marked as a reduction itself but
10516 is part of a group.
10517 (vect_recog_dot_prod_pattern): Don't check whether the statement
10518 is part of a group here.
10519 (vect_recog_sad_pattern): Likewise.
10520 (vect_recog_widen_sum_pattern): Likewise.
10522 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10525 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10527 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10530 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10532 (ix86_output_indirect_function_return): New prototype.
10533 (ix86_split_simple_return_pop_internal): Likewise.
10534 * config/i386/i386.c (indirect_return_via_cx): New.
10535 (indirect_return_via_cx_bnd): Likewise.
10536 (indirect_thunk_name): Handle return va CX_REG.
10537 (output_indirect_thunk_function): Create alias for
10538 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10539 (ix86_output_indirect_jmp): Remove the bool argument.
10540 (ix86_output_indirect_function_return): New function.
10541 (ix86_split_simple_return_pop_internal): Likewise.
10542 * config/i386/i386.md (*indirect_jump): Don't pass false
10543 to ix86_output_indirect_jmp.
10544 (*tablejump_1): Likewise.
10545 (simple_return_pop_internal): Change it to define_insn_and_split.
10546 Call ix86_split_simple_return_pop_internal to split it for
10547 -mfunction-return=.
10548 (simple_return_indirect_internal): Call
10549 ix86_output_indirect_function_return instead of
10550 ix86_output_indirect_jmp.
10552 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10555 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10556 memset and value initialization afterwards.
10558 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10560 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10562 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10565 * common/config/aarch64/aarch64-common.c
10566 (aarch_option_optimization_table[]): Switch
10567 off fomit-frame-pointer
10569 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10570 Chung-Ju Wu <jasonwucj@gmail.com>
10572 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10574 (store_multiple): Ditto.
10576 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10578 * config.gcc: Add --with-cpu support for nds32 target.
10579 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10580 * config/nds32/nds32.opt: Add -mcpu= option.
10582 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10584 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10585 isel=yes): Warn for these deprecated options.
10587 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10589 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10590 ISA_2_5_MASKS_EMBEDDED.
10592 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10594 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10595 p->max as pointers rather than using iterative_hash_expr.
10597 2018-02-23 Carl Love <cel@us.ibm.com>
10599 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10600 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10602 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10603 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10604 P8V_BUILTIN_VEC_VUNSIGNED2.
10606 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10609 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10610 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10612 * lra-constraints.c (curr_insn_transform): Set up
10613 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10615 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10616 * lra-eliminations.c (spill_pseudos): Ditto.
10617 (process_insn_for_elimination): Ditto.
10618 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10619 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10620 LRA_NON_CLOBBERED_ALT.
10622 2018-02-22 Martin Sebor <msebor@redhat.com>
10624 PR tree-optimization/84480
10625 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10626 to maybe_diag_stxncpy_trunc. Call it.
10627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10628 from gimple_fold_builtin_strcpy. Print inlining stack.
10629 (handle_builtin_stxncpy): Print inlining stack.
10630 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10632 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10635 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10636 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10637 and -fcheck-pointer-bounds are used together.
10638 (indirect_thunk_prefix): New enum.
10639 (indirect_thunk_need_prefix): New function.
10640 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10641 "_nt" instead of "_bnd" for NOTRACK prefix.
10642 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10643 (output_indirect_thunk_function): Likewise.
10645 (ix86_code_end): Update output_indirect_thunk_function calls.
10646 (ix86_output_indirect_branch_via_reg): Replace
10647 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10648 (ix86_output_indirect_branch_via_push): Likewise.
10649 (ix86_output_function_return): Likewise.
10650 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10651 incompatible with -fcf-protection=branch and
10652 -fcheck-pointer-bounds.
10654 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10657 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10658 Change gcc_assert call to output_operand_lossage.
10660 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10662 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10664 2018-02-22 DJ Delorie <dj@redhat.com>
10665 Sebastian Perta <sebastian.perta@renesas.com>
10666 Oleg Endo <olegendo@gcc.gnu.org>
10668 * config/rx/rx.c (rx_rtx_costs): New function.
10669 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10671 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10673 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10675 2018-02-22 Martin Liska <mliska@suse.cz>
10678 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10679 Add "native" as a possible value.
10681 2018-02-22 Martin Liska <mliska@suse.cz>
10684 * config/i386/i386.c (ix86_option_override_internal):
10685 Add "native" as a possible value for -march and -mtune.
10687 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10690 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10691 to all type variants.
10693 PR tree-optimization/84503
10694 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10695 width as info->bitpos + info->bitsize - start.
10696 (merged_store_group::merge_overlapping): Simplify width computation.
10697 (check_no_overlap): New function.
10698 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10699 start + width and last_order of the group, fail if check_no_overlap
10701 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10702 to group if check_no_overlap fails.
10704 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10706 * config/rs6000/altivec.md: Delete contraint arguments to
10707 define_expand, define_split, and define_peephole2, and in
10708 define_insn_and_split if always unused.
10709 * config/rs6000/darwin.md: Ditto.
10710 * config/rs6000/dfp.md: Ditto.
10711 * config/rs6000/rs6000.md: Ditto.
10712 * config/rs6000/sync.md: Ditto.
10713 * config/rs6000/vector.md: Ditto.
10714 * config/rs6000/vsx.md: Ditto.
10716 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10718 * config/rs6000/altivec.md: Write output control strings as braced
10719 blocks instead of double-quoted strings.
10720 * config/rs6000/darwin.md: Ditto.
10721 * config/rs6000/rs6000.md: Ditto.
10722 * config/rs6000/vector.md: Ditto.
10723 * config/rs6000/vsx.md: Ditto.
10725 2018-02-21 Jason Merrill <jason@redhat.com>
10727 PR c++/84314 - ICE with templates and fastcall attribute.
10728 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10730 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10732 * ipa-cp.c (determine_versionability): Fix comment typos.
10734 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10737 * ipa-cp.c (determine_versionability): Do not version functions caling
10740 2018-02-21 Martin Liska <mliska@suse.cz>
10743 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10744 Add "native" as a possible value.
10745 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10746 the macro when native cpu detection is available.
10748 2018-02-21 Martin Liska <mliska@suse.cz>
10751 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10752 Add "native" as a possible value.
10753 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10754 when native cpu detection is available.
10756 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10757 Martin Sebor <msebor@redhat.com>
10759 PR tree-optimization/84478
10760 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10762 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10763 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10764 support which is conservatively correct, for 2 only stay conservative
10765 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10766 argument to the 2 argument get_range_strlen, adjust 6 arg
10767 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10769 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10770 (gimple_fold_builtin_strlen): Pass true as last argument to
10773 2018-02-20 Martin Sebor <msebor@redhat.com>
10775 PR middle-end/84095
10776 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10777 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10778 (builtin_memref::builtin_memref): Factor out parts into
10779 set_base_and_offset and call it.
10781 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10783 PR middle-end/84406
10784 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10785 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10786 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10787 search at the associated MODE_INT.
10789 2018-02-20 Jeff Law <law@redhat.com>
10791 PR middle-end/82123
10792 PR tree-optimization/81592
10793 PR middle-end/79257
10794 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10795 for range data rather than using global data.
10796 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10797 range data rather than using global data.
10798 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10799 pass it to children as needed.
10800 (struct directive::fmtresult): Similarly.
10801 (struct directive::set_width): Similarly.
10802 (struct directive::set_precision): Similarly.
10803 (format_integer, format_directive, parse_directive): Similarly.
10804 (format_none): Accept unnamed vr_values parameter.
10805 (format_percent, format_floating, format_character): Similarly.
10806 (format_string, format_plain): Similarly.
10807 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10808 the EVRP range analyzer for range data rather than using global data.
10809 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10810 gimple-ssa-evrp-analyze.h
10811 (class sprintf_dom_walker): Add after_dom_children member function.
10812 Add evrp_range_analyzer member.
10813 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10814 range analyzer as needed.
10815 (sprintf_dom_walker::after_dom_children): New member function.
10816 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10818 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10819 (evrp_range_analyzer::pop_to_marker): Likewise.
10821 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10823 PR tree-optimization/84419
10824 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10825 with the required type if its current type is compatible but
10828 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10830 PR middle-end/82004
10831 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10832 after vectorization.
10834 2018-02-20 Martin Liska <mliska@suse.cz>
10837 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10838 possible values if we don't have a hint.
10840 2018-02-20 Martin Liska <mliska@suse.cz>
10844 * final.c (shorten_branches): Build align_tab array with one
10846 * opts.c (finish_options): Add alignment option limit check.
10847 (MAX_CODE_ALIGN): Likewise.
10848 (MAX_CODE_ALIGN_VALUE): Likewise.
10849 * doc/invoke.texi: Document maximum allowed option value for
10850 all -falign-* options.
10852 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10855 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10856 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10857 * var-tracking.c (emit_note_insn_var_location): Remove all references
10858 to NOTE_INSN_CALL_ARG_LOCATION.
10859 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10860 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10861 Use copy_rtx_if_shared.
10862 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10863 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10864 (dwarf2out_var_location): Remove handling of
10865 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10867 * final.c (final_scan_insn): Remove all references to
10868 NOTE_INSN_CALL_ARG_LOCATION.
10869 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10870 before dumping final insns.
10871 * except.c (emit_note_eh_region_end): Remove all references to
10872 NOTE_INSN_CALL_ARG_LOCATION.
10873 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10874 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10875 * config/arc/arc.c (hwloop_optimize): Likewise.
10876 * config/arm/arm.c (create_fix_barrier): Likewise.
10877 * config/s390/s390.c (s390_chunkify_start): Likewise.
10878 * config/sh/sh.c (find_barrier): Likewise.
10879 * config/i386/i386.c (rest_of_insert_endbranch,
10880 ix86_seh_fixup_eh_fallthru): Likewise.
10881 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10882 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10883 * config/frv/frv.c (frv_function_prologue): Likewise.
10884 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10886 (note_outside_basic_block_p): Remove all references to
10887 NOTE_INSN_CALL_ARG_LOCATION.
10888 * gengtype.c (adjust_field_rtx_def): Likewise.
10889 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10891 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10892 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10895 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10898 PR tree-optimization/84452
10899 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10900 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10903 2018-02-19 Martin Liska <mliska@suse.cz>
10906 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10908 2018-02-19 Martin Liska <mliska@suse.cz>
10909 Richard Sandiford <richard.sandiford@linaro.org>
10911 PR tree-optimization/82491
10912 * gimple-fold.c (get_base_constructor): Make earlier bail out
10915 2018-02-19 Carl Love <cel@us.ibm.com>
10917 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10918 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10920 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10921 P8V_BUILTIN_VEC_NEG.
10923 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10925 * config/rl78/rl78.md (movdf): New define expand.
10927 2018-02-19 Martin Liska <mliska@suse.cz>
10930 * doc/invoke.texi: Fix typo.
10931 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10933 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10935 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10936 handle rs6000_single_float and rs6000_double_float specially for
10939 2018-02-16 Jeff Law <law@redhat.com>
10941 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10942 (pop_regs): Use it.
10944 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10947 * ipa-inline.c (inline_small_functions): Fix a typo.
10949 2018-02-16 Nathan Sidwell <nathan@acm.org>
10951 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10953 2018-02-16 Carl Love <cel@us.ibm.com>
10955 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10956 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10957 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10958 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10959 expansion to P8V_BUILTIN_VEC_FLOAT2.
10961 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10963 PR rtl-optimization/70023
10964 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10965 src_regno into account.
10967 2018-02-16 Carl Love <cel@us.ibm.com>
10969 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10970 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10971 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10972 * config/rs6000/rs6000.c: Remove case statements for
10973 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10974 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10975 and P9V_BUILTIN_VEC_VINSERT4B.
10976 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10977 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10978 * config/rs6000/vsx.md:
10979 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10982 2018-02-16 Carl Love <cel@us.ibm.com>
10984 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10986 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10988 * config/rs6000/rs6000-c.c: Add the definitions for
10989 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10990 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10991 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10992 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10993 definition for insert4b and define insn *insert3b_internal.
10994 * doc/extend.texi: Add documentation for vec_extract4b.
10996 2018-02-16 Nathan Sidwell <nathan@acm.org>
10998 * doc/extend.texi (Backwards Compatibility): Mention friend
10999 injection. Note for-scope is deprecated.
11000 * doc/invoke.texi (-ffriend-injection): Deprecate.
11002 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11004 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11005 that moved to I2, also allow destinations that are a paradoxical
11006 subreg (instead of a normal reg).
11008 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11011 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11014 2018-02-16 Richard Biener <rguenther@suse.de>
11016 PR tree-optimization/84037
11017 PR tree-optimization/84016
11019 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11020 Adjust vec_construct for the fact we need additional higher latency
11021 128bit inserts for AVX256 and AVX512 vector builds.
11022 (ix86_add_stmt_cost): Scale vector construction cost for
11025 2018-02-16 Richard Biener <rguenther@suse.de>
11027 PR tree-optimization/84417
11028 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11029 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11030 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11032 2018-02-16 Martin Liska <mliska@suse.cz>
11035 * internal-fn.def (ASAN_CHECK): Set proper flags.
11036 (ASAN_MARK): Likewise.
11038 2018-02-16 Julia Koval <julia.koval@intel.com>
11040 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11041 from PTA_CANNONLAKE.
11043 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11046 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11047 Use ++iter rather than iter++ for std::list iterators.
11048 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11049 defer deleting them until all nodes in the forest are processed. Do
11050 free even leaf nodes. Change to_process into auto_vec.
11053 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11054 * vec.h (vec_default_construct): Use memset instead of placement new
11055 if BROKEN_VALUE_INITIALIZATION is defined.
11056 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11057 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11060 PR rtl-optimization/83723
11061 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11062 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11063 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11065 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11067 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11069 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11071 PR rtl-optimization/81443
11072 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11073 from inner REGs to paradoxical SUBREGs.
11075 2018-02-16 Richard Biener <rguenther@suse.de>
11077 PR tree-optimization/84399
11078 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11079 For operands we can analyze at their definition make sure we can
11080 analyze them at each use as well.
11082 2018-02-16 Richard Biener <rguenther@suse.de>
11084 PR tree-optimization/84190
11085 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11086 volatile accesses if the decl isn't volatile.
11088 2018-02-15 Jason Merrill <jason@redhat.com>
11090 PR c++/84314 - ICE with templates and fastcall attribute.
11091 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11092 TYPE_CANONICAL on an existing type.
11094 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11096 PR tree-optimization/84383
11097 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11098 dstoff nor call operand_equal_p if dstbase is NULL.
11100 PR tree-optimization/84334
11101 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11102 also a CONSTANT_CLASS_P, punt.
11104 2018-02-14 Jim Wilson <jimw@sifive.com>
11106 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11107 first SMALL_OPERAND check. New local min_second_step. Move assert
11108 to where locals are set. Add TARGET_RVC support.
11109 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11111 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11113 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11115 2018-02-14 Martin Sebor <msebor@redhat.com>
11117 PR tree-optimization/83698
11118 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11119 arrays constrain the offset range to their bounds.
11120 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11121 (builtin_access::overlap): Avoid setting the size of overlap if it's
11123 (maybe_diag_overlap): Also consider arrays when deciding what values
11124 of offsets to include in diagnostics.
11126 2018-02-14 Martin Sebor <msebor@redhat.com>
11129 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11130 that correspond to the kind of a declaration.
11132 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11135 * config/pa/pa.md: Load address of PIC label using the linkage table
11136 if the label is nonlocal.
11138 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11140 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11141 warning message if user requests -maltivec=be.
11142 * doc/invoke.texi: Document deprecation of -maltivec=be.
11144 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11147 * config/rs6000/rs6000-c.c: Update definitions for
11148 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11149 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11151 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11154 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11155 add _get_ssp intrinsics. Remove argument from
11156 __builtin_ia32_rdssp[d|q].
11157 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11158 * config/i386/i386-builtin.def: Remove argument from
11159 __builtin_ia32_rdssp[d|q].
11160 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11161 ix86_expand_special_args_builtin for _rdssp[d|q].
11162 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11163 Clear register before usage.
11164 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11165 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11167 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11169 PR tree-optimization/84357
11170 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11171 operand 1 of an ARRAY_REF too.
11173 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11176 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11177 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11179 (set_of_reg): New struct.
11180 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11181 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11182 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11184 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11185 Split into bitclr, bitset, bitinvert patterns if appropriate.
11186 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11187 use rx_fuse_in_memory_bitop.
11188 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11189 to named insn, correct maximum insn length.
11191 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11194 * machmode.def: Define a complex mode for PARTIAL_INT.
11195 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11197 * doc/rtl.texi: Document CSPImode.
11198 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11200 (msp430_hard_regno_nregs_with_padding): Likewise.
11202 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11205 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11207 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11209 PR rtl-optimization/84169
11210 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11211 we generated a parallel as new i3 and we split that to new i2 and i3
11212 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11213 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11214 those to i2, not i1. Partially rewrite this scan code.
11216 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11219 * stor-layout.c (place_field): For variable length fields, adjust
11220 offset_align afterwards not just based on the field's alignment,
11221 but also on the size.
11223 PR middle-end/84309
11224 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11225 of exps and logs in the use_exp2 case.
11227 2018-02-13 Jeff Law <law@redhat.com>
11229 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11230 entry for "vector".
11232 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11235 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11239 * common.opt (gas-loc-support, gas-locview-support): New.
11240 (ginline-points, ginternal-reset-location-views): New.
11241 * doc/invoke.texi: Document them. Use @itemx where intended.
11242 (gvariable-location-views): Adjust.
11243 * target.def (reset_location_view): New.
11244 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11245 (TARGET_RESET_LOCATION_VIEW): New.
11246 * doc/tm.texi: Rebuilt.
11247 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11248 (dwarf2out_default_as_locview_support): New.
11249 (output_asm_line_debug_info): Use option variables.
11250 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11251 (output_loc_list): Likewise.
11252 (add_high_low_attributes): Check option variables.
11253 Don't output entry view attribute in strict mode.
11254 (gen_inlined_subroutine_die): Check option variables.
11255 (dwarf2out_inline_entry): Likewise.
11256 (init_sections_and_labels): Likewise.
11257 (dwarf2out_early_finish): Likewise.
11258 (maybe_reset_location_view): New, from...
11259 (dwarf2out_var_location): ... here. Call it.
11260 * debug.h (dwarf2out_default_as_loc_support): Declare.
11261 (dwarf2out_default_as_locview_support): Declare.
11262 * hooks.c (hook_int_rtx_insn_0): New.
11263 * hooks.h (hook_int_rtx_insn_0): Declare.
11264 * toplev.c (process_options): Take -gas-loc-support and
11265 -gas-locview-support from dwarf2out. Enable
11266 -gvariable-location-views by default only with locview
11267 assembler support. Enable -ginternal-reset-location-views by
11268 default only if the target defines the corresponding hook.
11269 Enable -ginline-points by default if location views are
11270 enabled; force it disabled if statement frontiers are
11272 * tree-inline.c (expand_call_inline): Check option variables.
11273 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11275 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11277 PR tree-optimization/84321
11278 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11279 handling. Also check whether the anti-range contains any values
11280 that satisfy the mask; switch to a VR_RANGE if not.
11282 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11285 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11287 2018-02-13 Martin Jambor <mjambor@suse.cz>
11290 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11291 of call statements, also set location of a load to a temporary.
11293 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11295 * config/rl78/rl78.c (add_vector_labels): New function.
11296 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11297 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11298 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11299 which checks that no arguments are passed.
11300 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11301 * doc/extend.texi: Documentation for the new attribute.
11303 2018-02-13 Andreas Schwab <schwab@suse.de>
11305 * config/riscv/linux.h (CPP_SPEC): Define.
11307 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11310 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11311 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11312 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11313 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11314 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11315 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11316 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11317 temporarily for AES and PCLMUL builtins.
11319 PR tree-optimization/84339
11320 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11321 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11324 PR middle-end/84309
11325 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11326 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11327 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11329 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11331 * omp-simd-clone.h: New file.
11332 * omp-simd-clone.c: Include omp-simd-clone.h.
11333 (expand_simd_clones): No longer static.
11334 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11335 cgraph.h and omp-simd-clone.h.
11336 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11337 (vect_recog_widen_shift_pattern): Formatting fix.
11338 (vect_pattern_recog_1): Don't check optab for calls.
11341 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11342 operands[2] into a REG before using gen_lowpart on it.
11344 2018-02-12 Jeff Law <law@redhat.com>
11347 * config/sh/sh.c (find_barrier): Consider a sibling call
11350 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11351 successfully back substituting a reg.
11353 2018-02-12 Richard Biener <rguenther@suse.de>
11355 PR tree-optimization/84037
11356 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11357 parameter, move visited init to caller.
11358 (vect_slp_analyze_operations): Separate cost from validity
11359 check, initialize visited once for all instances.
11360 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11362 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11364 (vect_model_promotion_demotion_cost): Likewise.
11365 (vectorizable_bswap): Guard cost modeling with !slp_node
11366 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11368 (vectorizable_call): Likewise.
11369 (vectorizable_conversion): Likewise.
11370 (vectorizable_assignment): Likewise.
11371 (vectorizable_shift): Likewise.
11372 (vectorizable_operation): Likewise.
11373 (vectorizable_store): Likewise.
11374 (vectorizable_load): Likewise.
11375 (vectorizable_condition): Likewise.
11376 (vectorizable_comparison): Likewise.
11378 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11381 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11382 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11383 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11385 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11387 PR middle-end/83665
11388 * params.def (inline-min-speedup): Increase from 8 to 15.
11389 (max-inline-insns-auto): Decrease from 40 to 30.
11390 * ipa-split.c (consider_split): Add some buffer for function to
11391 be considered inlining candidate.
11392 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11395 2018-02-12 Richard Biener <rguenther@suse.de>
11397 PR tree-optimization/84037
11398 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11399 matched stmts if we cannot swap the non-matched ones.
11401 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11403 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11404 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11405 _mm_maskz_scalef_round_ss): New intrinsics.
11406 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11407 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11408 __builtin_ia32_scalefss_round): Remove.
11409 (__builtin_ia32_scalefsd_mask_round,
11410 __builtin_ia32_scalefss_mask_round): New intrinsics.
11411 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11412 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11413 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11414 "<round_constraint>")): Changed to ...
11415 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11416 "<round_scalar_constraint>")): ... this.
11417 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11418 %0, %1, %2<round_op3>}"): Changed to ...
11419 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11420 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11421 %2<round_scalar_mask_op3>}"): ... this.
11422 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11424 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11426 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11427 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11428 (_mm_maskz_sqrt_round_ss): New intrinsics.
11429 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11430 (__builtin_ia32_sqrtsd_mask_round)
11431 (__builtin_ia32_sqrtss_mask_round): New builtins.
11432 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11433 (__builtin_ia32_sqrtss_round): Remove.
11434 (__builtin_ia32_sqrtsd_mask_round)
11435 (__builtin_ia32_sqrtss_mask_round): New builtins.
11436 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11437 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11438 ((match_operand:VF_128 1 "vector_operand"
11439 "xBm,<round_constraint>")): Changed to ...
11440 ((match_operand:VF_128 1 "vector_operand"
11441 "xBm,<round_scalar_constraint>")): ... this.
11442 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11443 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11444 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11445 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11446 %<iptr>1<round_scalar_mask_op3>}): ... this.
11447 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11448 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11450 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11453 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11454 Cast vec_cmpeq result to correct type.
11455 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11456 Cast vec_cmpgt result to correct type.
11458 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11460 * final.c (final_scan_insn_1): Renamed from...
11461 (final_scan_insn): ... this. New wrapper, to recover
11462 seen from the outermost call in recursive ones.
11463 * config/sparc/sparc.c (output_return): Drop seen from call.
11464 (output_sibcall): Likewise.
11465 * config/visium/visium.c (output_branch): Likewise.
11467 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11469 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11472 2018-02-10 Alan Modra <amodra@gmail.com>
11475 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11476 Specify LR as an input.
11478 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11481 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11482 remove_member_access_dummy_vars): New functions.
11483 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11484 lower_omp_1, execute_lower_omp): Use them.
11486 PR rtl-optimization/84308
11487 * shrink-wrap.c (spread_components): Release todo vector.
11489 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11491 PR rtl-optimization/57193
11492 * ira-color.c (struct allocno_color_data): Add member
11493 conflict_allocno_hard_prefs.
11494 (update_conflict_allocno_hard_prefs): New.
11495 (bucket_allocno_compare_func): Add a preference based on
11496 conflict_allocno_hard_prefs.
11497 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11498 (color_allocnos): Remove a dead code. Initiate
11499 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11501 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11504 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11505 constraint from =wa to wa. Avoid a subreg on the output operand,
11506 instead use a pseudo and subreg it in a move.
11507 (p9_xxbrd_<mode>): Changed to ...
11508 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11509 (p9_xxbrd_v2df): New expander.
11510 (p9_xxbrw_<mode>): Changed to ...
11511 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11512 (p9_xxbrw_v4sf): New expander.
11514 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11516 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11518 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11521 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11522 multiply in 32-bit mode.
11523 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11524 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11527 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11529 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11530 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11531 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11532 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11534 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11537 * dwarf2out.c (is_trivial_indirect_ref): New function.
11538 (dwarf2out_late_global_decl): Do not generate a location
11539 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11540 and that are not defined in the current unit.
11542 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11544 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11545 instead of a libcall for UNORDERED.
11547 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11550 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11551 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11553 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11556 * config/s390/s390.c (s390_set_current_function): Invoke
11557 s390_indirect_branch_settings also if fndecl didn't change.
11559 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11561 * config/rs6000/rs6000.md (blockage): Set length to zero.
11563 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11565 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11567 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11570 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11571 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11575 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11576 PARALLEL incoming that failed vt_get_decl_and_offset check.
11578 PR middle-end/84237
11579 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11580 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11582 (get_variable_section): For decls in named .bss* sections pass true as
11583 second argument to bss_initializer_p.
11585 2018-02-09 Marek Polacek <polacek@redhat.com>
11586 Jakub Jelinek <jakub@redhat.com>
11589 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11590 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11591 Sync some changes from cxx_fold_indirect_ref.
11593 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11595 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11597 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11598 (BLOCK_INLINE_ENTRY_LABEL): New.
11599 (dwarf2out_var_location): Disregard inline entry markers.
11600 (inline_entry_data): New struct.
11601 (inline_entry_data_hasher): New hashtable type.
11602 (inline_entry_data_hasher::hash): New.
11603 (inline_entry_data_hasher::equal): New.
11604 (inline_entry_data_table): New variable.
11605 (add_high_low_attributes): Add DW_AT_entry_pc and
11606 DW_AT_GNU_entry_view attributes if a pending entry is found
11607 in inline_entry_data_table. Add old entry_pc attribute only
11608 if debug nonbinding markers are disabled.
11609 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11610 markers are enabled.
11611 (block_within_block_p, dwarf2out_inline_entry): New.
11612 (dwarf2out_finish): Check that no entries remained in
11613 inline_entry_data_table.
11614 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11615 (final_scan_insn, notice_source_line): Likewise.
11616 (rest_of_clean_state): Skip inline entry markers.
11617 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11619 * gimple.c (gimple_build_debug_inline_entry): New.
11620 * gimple.h (enum gimple_debug_subcode): Add
11621 GIMPLE_DEBUG_INLINE_ENTRY.
11622 (gimple_build_debug_inline_entry): Declare.
11623 (gimple_debug_inline_entry_p): New.
11624 (gimple_debug_nonbind_marker_p): Adjust.
11625 * insn-notes.def (INLINE_ENTRY): New.
11626 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11627 inline entry marker notes.
11628 (print_insn): Likewise.
11629 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11630 (INSN_DEBUG_MARKER_KIND): Likewise.
11631 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11632 * tree-inline.c (expand_call_inline): Build and insert
11633 debug_inline_entry stmt.
11634 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11635 inline entry blocks early, if nonbind markers are enabled.
11636 (dump_scope_block): Dump fragment info.
11637 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11638 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11639 (gimple_build_debug_inline_entry): New.
11640 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11641 Enable/disable inline entry points too.
11642 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11643 (DEBUG_INSN): Describe inline entry markers.
11645 * common.opt (gvariable-location-views): New.
11646 (gvariable-location-views=incompat5): New.
11647 * config.in: Rebuilt.
11648 * configure: Rebuilt.
11649 * configure.ac: Test assembler for view support.
11650 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11651 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11652 * dwarf2out.c (var_loc_view): New typedef.
11653 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11654 (dwarf2out_locviews_in_attribute): New.
11655 (dwarf2out_locviews_in_loclist): New.
11656 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11657 (enum dw_line_info_opcode): Add LI_adv_address.
11658 (struct dw_line_info_table): Add view.
11659 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11660 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11661 (zero_view_p): New variable.
11662 (ZERO_VIEW_P): New macro.
11663 (output_asm_line_debug_info): New.
11664 (struct var_loc_node): Add view.
11665 (add_AT_view_list, AT_loc_list): New.
11666 (add_var_loc_to_decl): Add view param. Test it against last.
11667 (new_loc_list): Add view params. Record them.
11668 (AT_loc_list_ptr): Handle loc and view lists.
11669 (view_list_to_loc_list_val_node): New.
11670 (print_dw_val): Handle dw_val_class_view_list.
11671 (size_of_die): Likewise.
11672 (value_format): Likewise.
11673 (loc_list_has_views): New.
11674 (gen_llsym): Set vl_symbol too.
11675 (maybe_gen_llsym, skip_loc_list_entry): New.
11676 (dwarf2out_maybe_output_loclist_view_pair): New.
11677 (output_loc_list): Output view list or entries too.
11678 (output_view_list_offset): New.
11679 (output_die): Handle dw_val_class_view_list.
11680 (output_dwarf_version): New.
11681 (output_compilation_unit_header): Use it.
11682 (output_skeleton_debug_sections): Likewise.
11683 (output_rnglists, output_line_info): Likewise.
11684 (output_pubnames, output_aranges): Update version comments.
11685 (output_one_line_info_table): Output view numbers in asm comments.
11686 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11687 Call maybe_gen_llsym.
11688 (loc_list_from_tree_1): Adjust.
11689 (add_AT_location_description): Create view list attribute if
11690 needed, check it's absent otherwise.
11691 (convert_cfa_to_fb_loc_list): Adjust.
11692 (maybe_emit_file): Call output_asm_line_debug_info for test.
11693 (dwarf2out_var_location): Reset views as needed. Precompute
11694 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11695 attribute. Set view.
11696 (new_line_info_table): Reset next view.
11697 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11698 (dwarf2out_source_line): Likewise. Output view resets and labels to
11699 the assembler, or select appropriate line info opcodes.
11700 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11701 (optimize_string_length): Catch it. Adjust.
11702 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11703 dw_val_class_view_list, and remove it if no longer needed.
11704 (hash_loc_list): Hash view numbers.
11705 (loc_list_hasher::equal): Compare them.
11706 (optimize_location_lists): Check whether a view list symbol is
11707 needed, and whether the locview attribute is present, and
11708 whether they match. Remove the locview attribute if no longer
11710 (index_location_lists): Call skip_loc_list_entry for test.
11711 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11712 Use output_dwarf_version.
11713 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11714 (struct dw_val_node): Add val_view_list.
11715 * final.c (SEEN_NEXT_VIEW): New.
11716 (set_next_view_needed): New.
11717 (clear_next_view_needed): New.
11718 (maybe_output_next_view): New.
11719 (final_start_function): Rename to...
11720 (final_start_function_1): ... this. Take pointer to FIRST,
11721 add SEEN parameter. Emit param bindings in the initial view.
11722 (final_start_function): Reintroduce SEEN-less interface.
11723 (final): Rename to...
11724 (final_1): ... this. Take SEEN parameter. Output final pending
11725 next view at the end.
11726 (final): Reintroduce seen-less interface.
11727 (final_scan_insn): Output pending next view before switching
11728 sections or ending a block. Mark the next view as needed when
11729 outputting variable locations. Notify debug backend of section
11730 changes, and of location view changes.
11731 (rest_of_handle_final): Adjust.
11732 * toplev.c (process_options): Autodetect value for debug variable
11733 location views option. Warn on incompat5 without -gdwarf-5.
11734 * doc/invoke.texi (gvariable-location-views): New.
11735 (gvariable-location-views=incompat5): New.
11736 (gno-variable-location-views): New.
11738 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11740 PR tree-optimization/84136
11741 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11742 that the result of find_edge is non-NULL.
11744 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11747 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11748 storing integer register in SImode. Fix cost of 256 and 512
11749 byte aligned SSE register store.
11751 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11753 * config/i386/i386.c (ix86_multiplication_cost): Fix
11754 multiplication cost for TARGET_AVX512DQ.
11756 2018-02-08 Marek Polacek <polacek@redhat.com>
11758 PR tree-optimization/84238
11759 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11762 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11764 PR tree-optimization/84265
11765 * tree-vect-stmts.c (vectorizable_store): Don't treat
11766 VMAT_CONTIGUOUS accesses as grouped.
11767 (vectorizable_load): Likewise.
11769 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11771 PR tree-optimization/81635
11772 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11773 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11774 (test_round_for_mask): New functions.
11775 (wide_int_cc_tests): Call test_round_for_mask.
11776 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11777 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11778 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11779 range returned by get_range_info.
11781 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11784 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11785 * symtab.c: Include builtins.h
11786 (symtab_node::output_to_lto_symbol_table_p): Move here
11787 from lto-streamer-out.c:output_symbol_p.
11788 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11789 (output_symbol_p): Move all logic to symtab.c
11790 (produce_symtab): Update.
11792 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11794 * config/s390/s390-opts.h (enum indirect_branch): Define.
11795 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11796 (s390_indirect_branch_via_thunk)
11797 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11798 (enum s390_indirect_branch_type): Define.
11799 * config/s390/s390.c (struct s390_frame_layout, struct
11800 machine_function): Remove.
11801 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11802 (indirect_branch_table_label_no, indirect_branch_table_name):
11804 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11805 (enum s390_indirect_branch_option): Define.
11806 (s390_return_addr_from_memory): New function.
11807 (s390_handle_string_attribute): New function.
11808 (s390_attribute_table): Add new attribute handler.
11809 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11810 (s390_indirect_branch_via_thunk): New function.
11811 (s390_indirect_branch_via_inline_thunk): New function.
11812 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11813 sibling call optimization for non z10 compiles.
11814 (s390_emit_call): Force indirect branch target to be a single
11815 register. Add r1 clobber for non-z10 compiles.
11816 (s390_emit_epilogue): Emit return jump via return_use expander.
11817 (s390_reorg): Handle JUMP_INSNs as execute targets.
11818 (s390_option_override_internal): Perform validity checks for the
11819 new command line options.
11820 (s390_indirect_branch_attrvalue): New function.
11821 (s390_indirect_branch_settings): New function.
11822 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11823 (s390_output_indirect_thunk_function): New function.
11824 (s390_code_end): Implement target hook.
11825 (s390_case_values_threshold): Implement target hook.
11826 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11828 * config/s390/s390.h (struct s390_frame_layout)
11829 (struct machine_function): Move here from s390.c.
11830 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11831 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11832 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11833 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11834 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11835 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11836 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11837 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11838 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11839 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11840 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11841 (mnemonic attribute): Add values which aren't recognized
11843 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11844 pattern for branch conversion. Fix mnemonic attribute.
11845 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11846 indirect branch via thunk if requested.
11847 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11848 ("*indirect_jump"): Disable for branch conversion using out of
11850 ("indirect_jump_via_thunk<mode>_z10")
11851 ("indirect_jump_via_thunk<mode>")
11852 ("indirect_jump_via_inlinethunk<mode>_z10")
11853 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11854 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11855 ("casesi_jump_via_inlinethunk<mode>_z10")
11856 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11857 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11858 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11859 ("*indirect2_jump"): Disable for branch conversion.
11860 ("casesi_jump"): Turn into expander and expand patterns for branch
11862 ("return_use"): New expander.
11863 ("*return"): Emit return via thunk and rename it to ...
11864 ("*return<mode>"): ... this one.
11865 * config/s390/s390.opt: Add new options and and enum for the
11868 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11870 * lra-constraints.c (match_reload): Unconditionally use
11871 gen_lowpart_SUBREG, rather than selecting between that
11872 and equivalent gen_rtx_SUBREG code.
11874 2018-02-08 Richard Biener <rguenther@suse.de>
11876 PR tree-optimization/84233
11877 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11878 changed flag instead of boguously re-using phi_inserted.
11880 2018-02-08 Martin Jambor <mjambor@suse.cz>
11882 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11883 static local variables.
11885 2018-02-08 Richard Biener <rguenther@suse.de>
11887 PR tree-optimization/84278
11888 * tree-vect-stmts.c (vectorizable_store): When looking for
11889 smaller vector types to perform grouped strided loads/stores
11890 make sure the mode is supported by the target.
11891 (vectorizable_load): Likewise.
11893 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11895 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11896 Increase LDP/STP opportunities by adding adjacent callee-saves.
11898 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11900 PR rtl-optimization/84068
11901 PR rtl-optimization/83459
11902 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11904 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11906 PR tree-optimization/84224
11907 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11908 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11909 non-zero arguments.
11911 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11914 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11915 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11916 count, remove one USE.
11918 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11920 * doc/install.texi (Configuration): Document the
11921 --with-long-double-format={ibm,ieee} PowerPC configuration
11925 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11926 Convert from define_expand to be define_insn_and_split. Rework
11927 float/double/_Float128 conversions to QI/HI/SImode to work with
11928 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11929 conversions to QI/HImode types did a store and then a load to
11930 truncate the value. For conversions to VSX registers, don't split
11931 the insn, instead emit the code directly. Use the code iterator
11932 any_fix to combine signed and unsigned conversions.
11933 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11934 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11935 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11936 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11937 (fix_<mode>di2_hw): Likewise.
11938 (fixuns_<mode>di2_hw): Likewise.
11939 (fix_<mode>si2_hw): Likewise.
11940 (fixuns_<mode>si2_hw): Likewise.
11941 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11942 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11943 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11944 fix<uns>_trunc<SFDF:mode>si2_p8.
11945 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11946 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11947 (fix<uns>_<mode>_mem): Likewise.
11948 (fctiw<u>z_<mode>_mem): Likewise.
11949 (fix<uns>_<mode>_mem): Likewise.
11950 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11951 the register allocator from doing a direct move to the GPRs to do
11952 a store, and instead use the ISA 3.0 store byte/half-word from
11953 vector register instruction. For IEEE 128-bit floating point,
11954 also optimize stores of 32-bit ints.
11955 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11957 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11959 * genextract.c (push_pathstr_operand): New function to support
11961 (walk_rtx): Call push_pathstr_operand.
11962 (print_path): Support [a-zA-Z].
11964 2018-02-07 Richard Biener <rguenther@suse.de>
11966 PR tree-optimization/84037
11967 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11968 (cse_and_gimplify_to_preheader): Declare.
11969 (vect_get_place_in_interleaving_chain): Likewise.
11970 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11972 (_loop_vec_info::~_loop_vec_info): Delete it.
11973 (cse_and_gimplify_to_preheader): New function.
11974 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11975 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11976 (vectorizable_load): Likewise. For grouped stores always base
11977 the IV on the first element.
11978 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11979 condition before gimplifying.
11981 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11983 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11984 *DIV_EXPR and *MOD_EXPR.
11986 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11989 * config/i386/i386.c (ix86_option_override_internal): Mask out
11990 the CF_SET bit when checking -fcf-protection.
11992 2018-02-07 Tom de Vries <tom@codesourcery.com>
11995 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11998 2018-02-07 Richard Biener <rguenther@suse.de>
12000 PR tree-optimization/84204
12001 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12004 PR tree-optimization/84205
12005 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12006 special-case isl_ast_op_zdiv_r.
12008 PR tree-optimization/84223
12009 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12010 Only add conditions from within the region.
12011 (gather_bbs::after_dom_children): Adjust.
12013 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12016 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12017 * config/avr/avr.md: Only post-reload split REG-REG moves if
12018 either register is GENERAL_REG_P.
12020 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12022 PR tree-optimization/84235
12023 * tree-ssa-scopedtables.c
12024 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12025 if the subtraction is performed in floating point type where NaNs are
12026 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12027 build 1. Formatting fix.
12029 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12032 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12033 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12034 and skip it regardless of bb boundaries. Use CALL_P macro,
12035 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12036 unnecessarily, formatting fix.
12038 2018-02-06 Michael Collison <michael.collison@arm.com>
12040 * config/arm/thumb2.md:
12041 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12042 (*thumb_mov_notscc): Ditto.
12044 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12047 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12048 unsigned_fix, not "s".
12050 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12052 * configure.ac (gcc_fn_eh_frame_ro): New function.
12053 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12054 correct .eh_frame permissions.
12055 * configure: Regenerate.
12057 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12059 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12060 irrelevant options.
12062 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12064 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12065 Display warning message for -mno-speculate-indirect-jumps.
12067 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12069 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12071 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12073 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12075 PR tree-optimization/84225
12076 * tree-eh.c (find_trapping_overflow): Only call
12077 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12079 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12082 * config/i386/i386.c: Reimplement the check of possible options
12083 -mibt/-mshstk conbination. Change error messages.
12084 * doc/invoke.texi: Fix a typo: remove extra '='.
12086 2018-02-06 Marek Polacek <polacek@redhat.com>
12088 PR tree-optimization/84228
12089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12091 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12094 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12095 emitted arch directives.
12096 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12097 __ARM_FEATURE_COPROC before changing architectures.
12099 2018-02-06 Richard Biener <rguenther@suse.de>
12101 * config/i386/i386.c (print_reg): Fix typo.
12102 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12104 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12106 * configure: Regenerate.
12108 2018-02-05 Martin Sebor <msebor@redhat.com>
12110 PR tree-optimization/83369
12111 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12114 2018-02-05 Martin Liska <mliska@suse.cz>
12116 * doc/invoke.texi: Cherry-pick upstream r323995.
12118 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12120 * ira.c (ira_init_register_move_cost): Adjust comment.
12122 2018-02-05 Martin Liska <mliska@suse.cz>
12124 PR gcov-profile/84137
12125 * doc/gcov.texi: Fix typo in documentation.
12127 2018-02-05 Martin Liska <mliska@suse.cz>
12129 PR gcov-profile/83879
12130 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12131 using dlopen functionality.
12133 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12135 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12136 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12137 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12138 _mm_maskz_range_round_ss): New intrinsics.
12139 (__builtin_ia32_rangesd128_round)
12140 (__builtin_ia32_rangess128_round): Remove.
12141 (__builtin_ia32_rangesd128_mask_round,
12142 __builtin_ia32_rangess128_mask_round): New builtins.
12143 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12144 __builtin_ia32_rangess128_round): Remove.
12145 (__builtin_ia32_rangesd128_mask_round,
12146 __builtin_ia32_rangess128_mask_round): New builtins.
12147 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12148 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12149 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12150 "<round_saeonly_constraint>")): Changed to ...
12151 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12152 "<round_saeonly_scalar_constraint>")): ... this.
12153 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12154 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12155 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12156 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12157 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12159 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12161 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12163 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12164 Remove all values except native, 8540 and 8548.
12166 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12168 * config/i386/i386.c (ix86_output_function_return): Pass
12169 INVALID_REGNUM, instead of -1, as invalid register number to
12170 indirect_thunk_name and output_indirect_thunk.
12172 2018-02-02 Julia Koval <julia.koval@intel.com>
12174 * config.gcc: Add -march=icelake.
12175 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12176 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12177 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12178 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12179 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12180 (processor_target_table): Add icelake.
12181 (ix86_option_override_internal): Handle new PTAs.
12182 (get_builtin_code_for_version): Handle icelake.
12183 (M_INTEL_COREI7_ICELAKE): New.
12184 (fold_builtin_cpu): Handle icelake.
12185 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12186 * doc/invoke.texi: Add -march=icelake.
12188 2018-02-02 Julia Koval <julia.koval@intel.com>
12190 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12191 to wide_int_bitmask.
12192 * wide-int-bitmask.h: New.
12194 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12197 * config/i386/i386.md: Replace Pmode with word_mode in
12198 builtin_setjmp_setup and builtin_longjmp to support x32.
12200 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12204 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12206 (rs6000_supported_cpu_names): New static variable.
12207 (linux_cpu_translation_table): Likewise.
12208 (elf_platform) <cpu>: Define new static variable and use it.
12209 Translate kernel AT_PLATFORM name to canonical name if needed.
12210 Error if platform name is unknown.
12212 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12215 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12217 2018-02-01 Jeff Law <law@redhat.com>
12220 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12221 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12222 the scratch if RELEASE_VIA_POP is false.
12223 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12224 If we have to save a temporary register, decrement SIZE appropriately.
12225 Pass new arguments to release_scratch_register_on_entry.
12226 (ix86_adjust_stack_and_probe): Likewise.
12227 (ix86_emit_probe_stack_range): Pass new arguments to
12228 release_scratch_register_on_entry.
12230 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12232 PR rtl-optimization/84157
12233 * combine.c (change_zero_ext): Use REG_P predicate in
12234 front of HARD_REGISTER_P predicate.
12236 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12238 * config/avr/avr.c (avr_option_override): Move disabling of
12239 -fdelete-null-pointer-checks to...
12240 * common/config/avr/avr-common.c (avr_option_optimization_table):
12243 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12245 PR tree-optimization/81635
12246 * tree-data-ref.c (split_constant_offset_1): For types that
12247 wrap on overflow, try to use range info to prove that wrapping
12250 2018-02-01 Renlin Li <renlin.li@arm.com>
12253 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12254 TAILCALL_ADDR_REGS.
12255 (aarch64_register_move_cost): Likewise.
12256 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12257 TAILCALL_ADDR_REGS.
12258 (REG_CLASS_NAMES): Likewise.
12259 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12260 TAILCALL_ADDR_REGS. Remove IP registers.
12261 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12263 2018-02-01 Richard Biener <rguenther@suse.de>
12265 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12266 for specifying RPO order and allow NULL for that.
12267 * domwalk.c (dom_walker::dom_walker): Likewise.
12268 (dom_walker::walk): Handle NULL RPO order.
12269 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12271 (rewrite_update_dom_walker): Likewise.
12272 (mark_def_dom_walker): Likewise.
12274 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12276 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12277 (aarch64_maybe_expand_sve_subreg_move): Declare.
12278 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12279 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12281 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12282 that are semantically a reverse operation.
12283 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12284 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12285 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12287 (aarch64_can_change_mode_class): For big-endian, forbid changes
12288 between two SVE modes if they have different element sizes.
12290 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12292 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12293 the TImode handling for big-endian targets.
12295 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12297 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12298 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12300 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12301 Remove BSWAP handing for big-endian targets and use the form of
12302 LD1RQ appropariate for the mode.
12304 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12306 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12307 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12308 duplicated element.
12310 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12313 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12314 check for operands that need to go through aarch64_sve_reload_be.
12316 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12318 PR tree-optimization/81661
12319 PR tree-optimization/84117
12320 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12321 * tree-eh.c: Include gimplify.h.
12322 (find_trapping_overflow, replace_trapping_overflow,
12323 rewrite_to_non_trapping_overflow): New functions.
12324 * tree-vect-loop.c: Include tree-eh.h.
12325 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12326 * tree-data-ref.c: Include tree-eh.h.
12327 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12329 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12331 PR rtl-optimization/84123
12332 * combine.c (change_zero_ext): Check if hard register satisfies
12333 can_change_dest_mode before calling gen_lowpart_SUBREG.
12335 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12338 * ira.c (ira_init_register_move_cost): Remove assert.
12340 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12342 PR rtl-optimization/84071
12343 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12344 * doc/tm.texi: Regenerate.
12346 2018-01-31 Richard Biener <rguenther@suse.de>
12348 PR tree-optimization/84132
12349 * tree-data-ref.c (analyze_miv_subscript): Properly
12350 check whether evolution_function_is_affine_multivariate_p
12351 before calling gcd_of_steps_may_divide_p.
12353 2018-01-31 Julia Koval <julia.koval@intel.com>
12356 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12357 * config/i386/i386.md (rdpid_rex64) New.
12358 (rdpid): Make 32bit only.
12360 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12363 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12364 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12366 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12369 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12371 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12373 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12375 PR rtl-optimization/84071
12376 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12377 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12379 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12381 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12382 (arc_attribute_table): Add 'aux' attribute.
12383 (arc_in_small_data_p): Consider aux like variables.
12384 (arc_is_aux_reg_p): New function.
12385 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12386 (arc_get_aux_arg): New function.
12387 (prepare_move_operands): Handle aux-register access.
12388 (arc_handle_aux_attribute): New function.
12389 * doc/extend.texi (ARC Variable attributes): Add subsection.
12391 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12393 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12394 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12395 (arc_attribute_table): Add 'uncached' attribute.
12396 (arc_print_operand): Print '.di' flag for uncached memory
12398 (arc_in_small_data_p): Do not consider for small data the uncached
12400 (arc_is_uncached_mem_p): New function.
12401 * config/arc/predicates.md (compact_store_memory_operand): Check
12402 for uncached memory accesses.
12403 (nonvol_nonimm_operand): Likewise.
12404 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12406 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12409 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12410 falign-loops=): Add Optimization flag.
12412 2018-01-30 Jeff Law <law@redhat.com>
12415 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12416 INT_REGISTERS_SAVED. Check it prior to calling
12417 get_scratch_register_on_entry.
12418 (ix86_adjust_stack_and_probe): Similarly.
12419 (ix86_emit_probe_stack_range): Similarly.
12420 (ix86_expand_prologue): Corresponding changes.
12422 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12425 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12426 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12428 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12431 * lra-constraints.c (curr_insn_transform): Process AND in the
12434 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12436 PR rtl-optimization/83986
12437 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12438 dependence against last_pending_memory_flush in addition to
12439 pending_jump_insns.
12441 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12443 PR tree-optimization/81611
12444 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12447 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12450 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12453 2018-01-30 Richard Biener <rguenther@suse.de>
12454 Jakub Jelinek <jakub@redhat.com>
12456 PR tree-optimization/84111
12457 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12458 inner loops added during recursion, as they don't have up-to-date
12461 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12464 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12465 (can_inline_edge_by_limits_p): ... here.
12466 (can_early_inline_edge_p, check_callers,
12467 update_caller_keys, update_callee_keys, recursive_inlining,
12468 add_new_edges_to_heap, speculation_useful_p,
12469 inline_small_functions,
12470 inline_small_functions, flatten_function,
12471 inline_to_all_callers_1): Update.
12473 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12475 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12478 2018-01-30 Richard Biener <rguenther@suse.de>
12480 PR tree-optimization/83008
12481 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12482 invariant and constant vector uses in stmts when they need
12483 more than one stmt.
12485 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12488 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12489 * configure: Regenerate.
12491 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12493 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12495 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12496 Use gen_rtx_REG rather than gen_lowpart.
12498 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12500 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12501 rather than 0 when creating partial subregs.
12503 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12505 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12508 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12511 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12512 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12513 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12514 flags. This restores the settings used before the 2017-07-24.
12515 Turning off pre increment/decrement/modify allows IVOPTS to
12516 optimize DF/SF loops where the index is an int.
12518 2018-01-29 Richard Biener <rguenther@suse.de>
12519 Kelvin Nilsen <kelvin@gcc.gnu.org>
12522 * tree-vect-stmts.c (vectorizable_call): Don't call
12523 targetm.vectorize_builtin_md_vectorized_function if callee is
12526 2018-01-22 Carl Love <cel@us.ibm.com>
12528 * doc/extend.tex: Fix typo in second arg in
12529 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12531 2018-01-29 Richard Biener <rguenther@suse.de>
12533 PR tree-optimization/84086
12534 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12535 (flush_ssaname_freelist): When SSA names were released reset
12536 the SCEV hash table.
12538 2018-01-29 Richard Biener <rguenther@suse.de>
12540 PR tree-optimization/84057
12541 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12542 removed paths when removing edges.
12544 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12546 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12547 -mfunction-return=@var{choice}.
12549 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12551 PR diagnostic/84034
12552 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12553 Handle CR like TAB.
12554 (layout::print_source_line): Likewise.
12555 (test_get_line_width_without_trailing_whitespace): Add test cases.
12557 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12559 PR middle-end/84040
12560 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12563 2018-01-26 Jim Wilson <jimw@sifive.com>
12565 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12567 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12570 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12572 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12573 and CMP + SUB-immediate -> SUBS.
12575 2018-01-26 Martin Sebor <msebor@redhat.com>
12577 PR tree-optimization/83896
12578 * tree-ssa-strlen.c (get_string_len): Rename...
12579 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12580 Avoid assuming length is constant.
12581 (handle_char_store): Use HOST_WIDE_INT for string length.
12583 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12586 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12587 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12589 2018-01-26 Richard Biener <rguenther@suse.de>
12591 PR rtl-optimization/84003
12592 * dse.c (record_store): Only record redundant stores when
12593 the earlier store aliases at least all accesses the later one does.
12595 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12597 PR rtl-optimization/83985
12598 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12599 REG_CFA_RESTORE insns.
12600 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12601 REG_CFA_RESTORE insns here.
12604 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12605 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12607 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12609 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12610 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12611 (arc_init): Likewise.
12612 (arc_override_options): Likewise.
12613 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12615 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12617 * config/arc/arc.h (TARGET_DBNZ): Define.
12618 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12619 properly set the tune attribute.
12620 (dbnz): Use TARGET_DBNZ guard.
12621 * config/arc/arc.opt (mtune): Add core3 option.
12623 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12625 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12626 recognize new pic like addresses.
12627 (arc_delegitimize_address): Clean up.
12629 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12631 * config/arc/arc-arches.def: Option mrf16 valid for all
12633 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12634 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12635 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12636 * config/arc/arc-tables.opt: Regenerate.
12637 * config/arc/arc.c (arc_conditional_register_usage): Handle
12638 reduced register file case.
12639 (arc_file_start): Set must have build attributes.
12640 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12641 mrf16 option value.
12642 * config/arc/arc.opt (mrf16): Add new option.
12643 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12644 * config/arc/genmultilib.awk: Handle new mrf16 option.
12645 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12646 * config/arc/t-multilib: Regenerate.
12647 * doc/invoke.texi (ARC Options): Document mrf16 option.
12649 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12651 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12652 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12653 (arc_attribute_table): Add 'secure_call' attribute.
12654 (arc_print_operand): Print secure call operand.
12655 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12657 (arc_is_secure_call_p): New function. * config/arc/arc.md
12658 (call_i): Add support for sjli instruction.
12659 (call_value_i): Likewise.
12660 * config/arc/constraints.md (Csc): New constraint.
12662 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12663 John Eric Martin <John.Martin@emmicro-us.com>
12665 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12666 * config/arc/arc.c (_arc_jli_section): New struct.
12667 (arc_jli_section): New type.
12668 (rc_jli_sections): New static variable.
12669 (arc_handle_jli_attribute): New function.
12670 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12671 (arc_file_end): New function.
12672 (TARGET_ASM_FILE_END): Define.
12673 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12674 (arc_add_jli_section): New function.
12675 (jli_call_scan): Likewise.
12676 (arc_reorg): Call jli_call_scan.
12677 (arc_output_addsi): Remove 'S' from printing asm operand.
12678 (arc_is_jli_call_p): New function.
12679 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12681 (movhi_insn): Likewise.
12682 (movsi_insn): Likewise.
12683 (movsi_set_cc_insn): Likewise.
12684 (loadqi_update): Likewise.
12685 (load_zeroextendqisi_update): Likewise.
12686 (load_signextendqisi_update): Likewise.
12687 (loadhi_update): Likewise.
12688 (load_zeroextendhisi_update): Likewise.
12689 (load_signextendhisi_update): Likewise.
12690 (loadsi_update): Likewise.
12691 (loadsf_update): Likewise.
12692 (movsicc_insn): Likewise.
12693 (bset_insn): Likewise.
12694 (bxor_insn): Likewise.
12695 (bclr_insn): Likewise.
12696 (bmsk_insn): Likewise.
12697 (bicsi3_insn): Likewise.
12698 (cmpsi_cc_c_insn): Likewise.
12699 (movsi_ne): Likewise.
12700 (movsi_cond_exec): Likewise.
12701 (clrsbsi2): Likewise.
12702 (norm_f): Likewise.
12712 (call_i): Remove 'S' asm letter, add jli instruction.
12713 (call_value_i): Likewise.
12714 * config/arc/arc.op (mjli-always): New option.
12715 * config/arc/constraints.md (Cji): New constraint.
12716 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12718 (subsf3_fpx): Likewise.
12719 (mulsf3_fpx): Likewise.
12720 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12722 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12723 function attrbutes.
12724 * doc/invoke.texi (ARC): Document mjli-always option.
12726 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12728 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12729 avoid addition with 0 and use incw and decw where possible.
12731 2018-01-26 Richard Biener <rguenther@suse.de>
12733 PR tree-optimization/81082
12734 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12735 association if it requires casting to unsigned.
12736 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12737 from fold_plusminus_mult_expr to catch important cases late when
12738 range info is available.
12740 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12742 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12743 * configure.ac (hidden_linkonce): New test.
12744 * configure: Regenerate.
12745 * config.in: Regenerate.
12747 2018-01-26 Julia Koval <julia.koval@intel.com>
12749 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12750 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12751 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12752 _mm_mask_bitshuffle_epi64_mask): Fix type.
12753 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12754 USI_FTYPE_V4DI_V4DI_USI): Remove.
12755 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12756 __builtin_ia32_vpshufbitqmb256_mask,
12757 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12758 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12759 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12761 2018-01-26 Alan Modra <amodra@gmail.com>
12764 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12765 UNSPEC_VBPERMQ. Sort other unspecs.
12767 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12769 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12771 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12773 PR middle-end/83055
12774 * predict.c (drop_profile): Do not push/pop cfun; update also
12776 (handle_missing_profiles): Fix logic looking for zero profiles.
12778 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12780 PR middle-end/83977
12781 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12782 on functions with #pragma omp declare simd or functions with simd
12784 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12785 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12786 Remove trailing \n from warning_at calls.
12788 2018-01-25 Tom de Vries <tom@codesourcery.com>
12791 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12792 for neutered workers.
12794 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12797 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12798 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12800 2018-01-24 Jeff Law <law@redhat.com>
12803 * i386.c (get_probe_interval): Move to earlier point.
12804 (ix86_compute_frame_layout): If -fstack-clash-protection and
12805 the frame is larger than the probe interval, then use pushes
12806 to save registers rather than reg->mem moves.
12807 (ix86_expand_prologue): Remove conditional for int_registers_saved
12810 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12813 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12814 min/max for never referenced object.
12816 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12818 PR middle-end/83977
12819 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12821 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12822 attributes from DECL_ATTRIBUTES (decl) without affecting
12823 DECL_ATTRIBUTES (current_function_decl).
12824 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12825 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12827 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12829 PR tree-optimization/83979
12830 * fold-const.c (fold_comparison): Use constant_boolean_node
12831 instead of boolean_{true,false}_node.
12833 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12835 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12838 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12840 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12841 Simplify the clause that sets the length attribute.
12842 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12843 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12844 clause that sets the length attribute.
12845 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12847 2018-01-24 Tom de Vries <tom@codesourcery.com>
12850 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12851 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12852 Add strict parameter.
12853 (prevent_branch_around_nothing): Insert dummy insn between branch to
12854 label and label with no ptx insn inbetween.
12855 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12857 2018-01-24 Tom de Vries <tom@codesourcery.com>
12860 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12861 for neutered threads in warp.
12862 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12864 2018-01-24 Richard Biener <rguenther@suse.de>
12866 PR tree-optimization/83176
12867 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12870 2018-01-24 Richard Biener <rguenther@suse.de>
12872 PR tree-optimization/82819
12873 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12874 code generating pluses that are no-ops in the target precision.
12876 2018-01-24 Richard Biener <rguenther@suse.de>
12878 PR middle-end/84000
12879 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12881 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12883 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12884 to merge probabilities.
12885 * predict.c (probably_never_executed): Also mark as cold functions
12886 with global 0 profile and guessed local profile.
12887 * profile-count.c (profile_probability::combine_with_count): New
12889 * profile-count.h (profile_probability::operator*,
12890 profile_probability::operator*=, profile_probability::operator/,
12891 profile_probability::operator/=): Reduce precision to adjusted
12892 and set value to guessed on contradictory divisions.
12893 (profile_probability::combine_with_freq): Remove.
12894 (profile_probability::combine_wiht_count): Declare.
12895 (profile_count::force_nonzero):: Set to adjusted.
12896 (profile_count::probability_in):: Set quality to adjusted.
12897 * tree-ssa-tail-merge.c (replace_block_by): Use
12898 combine_with_count.
12900 2018-01-23 Andrew Waterman <andrew@sifive.com>
12901 Jim Wilson <jimw@sifive.com>
12903 * config/riscv/riscv.c (riscv_stack_boundary): New.
12904 (riscv_option_override): Set riscv_stack_boundary. Handle
12905 riscv_preferred_stack_boundary_arg.
12906 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12907 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12908 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12909 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12910 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12911 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12913 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12916 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12917 of struct ix86_frame.
12918 (ix86_expand_epilogue): Likewise. Add a local variable for
12919 the reg_save_offset field in struct ix86_frame.
12921 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12923 PR tree-optimization/82604
12924 * tree-loop-distribution.c (enum partition_kind): New enum item
12925 PKIND_PARTIAL_MEMSET.
12926 (partition_builtin_p): Support above new enum item.
12927 (generate_code_for_partition): Ditto.
12928 (compute_access_range): Differentiate cases that equality can be
12929 proven at all loops, the innermost loops or no loops.
12930 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12931 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12932 (finalize_partitions, distribute_loop): Don't fuse partition of
12933 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12934 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12935 parloop is enabled.
12937 2018-01-23 Martin Liska <mliska@suse.cz>
12939 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12940 order to ignore the predictor.
12941 (PRED_POLYMORPHIC_CALL): Likewise.
12942 (PRED_RECURSIVE_CALL): Likewise.
12944 2018-01-23 Martin Liska <mliska@suse.cz>
12946 * tree-profile.c (tree_profiling): Print function header to
12947 aware reader which function we are working on.
12948 * value-prof.c (gimple_find_values_to_profile): Do not print
12949 not interesting value histograms.
12951 2018-01-23 Martin Liska <mliska@suse.cz>
12953 * profile-count.h (enum profile_quality): Add
12954 profile_uninitialized as the first value. Do not number values
12955 as they are zero based.
12956 (profile_count::verify): Update sanity check.
12957 (profile_probability::verify): Likewise.
12959 2018-01-23 Nathan Sidwell <nathan@acm.org>
12961 * doc/invoke.texi (ffor-scope): Deprecate.
12963 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12965 PR tree-optimization/83510
12966 * domwalk.c (set_all_edges_as_executable): New function.
12967 (dom_walker::dom_walker): Convert bool param
12968 "skip_unreachable_blocks" to enum reachability. Move setup of
12969 edge flags to set_all_edges_as_executable and only do it when
12970 reachability is REACHABLE_BLOCKS.
12971 * domwalk.h (enum dom_walker::reachability): New enum.
12972 (dom_walker::dom_walker): Convert bool param
12973 "skip_unreachable_blocks" to enum reachability.
12974 (set_all_edges_as_executable): New decl.
12975 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12976 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12978 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12979 but converting true to REACHABLE_BLOCKS.
12980 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12982 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12983 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12984 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12986 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12987 if check_all_array_refs will be called.
12989 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12991 * tree.c (selftest::test_location_wrappers): Add more test
12994 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12996 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12997 (selftest::test_bit_in_range): Likewise.
12999 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13002 * doc/sourcebuild.texi (vect_float): Say that the selector
13003 only describes the situation when -funsafe-math-optimizations is on.
13004 (vect_float_strict): Document.
13006 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13008 PR tree-optimization/83965
13009 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13010 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13011 instead of checking only for a reduction.
13012 (vect_recog_widen_sum_pattern): Likewise.
13014 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13016 * predict.c (probably_never_executed): Only use precise profile info.
13017 (compute_function_frequency): Skip after inlining hack since we now
13018 have quality checking.
13020 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13022 * profile-count.h (profile_probability::very_unlikely,
13023 profile_probability::unlikely, profile_probability::even): Set
13024 precision to guessed.
13026 2018-01-23 Richard Biener <rguenther@suse.de>
13028 PR tree-optimization/83963
13029 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13030 Properly terminate dominator walk when crossing the exit edge not
13031 when visiting its source block.
13033 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13036 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13037 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13039 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13041 PR tree-optimization/83957
13042 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13043 semicolon after for body surrounded by braces.
13045 PR tree-optimization/83081
13046 * profile-count.h (profile_probability::split): New method.
13047 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13048 Use profile_probability::split.
13049 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13050 when splitting a single conditional jump into 2.
13052 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13054 PR tree-optimization/69452
13055 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13058 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13060 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13061 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13062 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13064 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13066 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13068 * config/rl78/rl78.md (movdi): New define_expand.
13069 * config/rl78/rl78.c (rl78_split_movdi): New function.
13071 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13074 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13076 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13077 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13078 128-bit to produce an UNSPEC move to get the double word with the
13079 signbit and then a shift directly to do signbit.
13080 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13081 implementation with a new version that just does either a direct
13082 move or a regular move. Move memory interface to separate insns.
13083 Move insns so they are next to the expander.
13084 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13085 with signbit move. Split big and little endian case.
13086 (signbit<mode>2_dm_mem_le): Likewise.
13087 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13088 (signbit<mode>2_dm2): Likewise.
13090 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13092 * config/rl78/rl78.md (anddi3): New define_expand.
13094 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13096 * config/rl78/rl78.md (umindi3): New define_expand.
13098 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13100 * config/rl78/rl78.md (smindi3): New define_expand.
13102 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13104 * config/rl78/rl78.md (smaxdi3): New define_expand.
13106 2018-01-22 Carl Love <cel@us.ibm.com>
13108 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13109 LVX_V1TI): Add macro expansion.
13110 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13111 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13112 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13113 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13114 Change check to determine if the instruction is a byte reversing
13115 entry. Fix typo in comment.
13116 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13117 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13118 Add def_builtin calls for new builtins.
13119 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13120 Add define_insn expansion.
13122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13124 * config/rl78/rl78.md (umaxdi3): New define_expand.
13126 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13128 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13129 for non-QImode registers.
13131 2018-01-22 Richard Biener <rguenther@suse.de>
13133 PR tree-optimization/83963
13134 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13135 including the loop exit block.
13136 (scop_detection::merge_sese): Likewise.
13137 (scop_detection::add_scop): Do it here instead.
13139 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13141 * doc/sourcebuild.texi (arm_softfloat): Document.
13143 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13146 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13147 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13148 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13150 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13151 David Edelsohn <dje.gcc@gmail.com>
13154 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13155 Change "crset eq" to "crset 2".
13156 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13157 (*call_indirect_aix<mode>_nospec): Likewise.
13158 (*call_value_indirect_aix<mode>_nospec): Likewise.
13159 (*call_indirect_elfv2<mode>_nospec): Likewise.
13160 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13161 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13162 change assembly output from . to $.
13163 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13164 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13165 (*tablejump<mode>_internal1_nospec): Likewise.
13167 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13170 * config/sh/sh_optimize_sett_clrt.cc:
13171 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13173 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13175 PR tree-optimization/83940
13176 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13177 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13178 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13179 use it to pass back the definition type.
13180 (vect_check_store_rhs): Likewise rhs_dt_out.
13181 (vect_build_gather_load_calls): Add a mask_dt argument and use
13182 it instead of a call to vect_is_simple_use.
13183 (vectorizable_store): Update calls to vect_check_load_store_mask
13184 and vect_check_store_rhs. Use the dt returned by the latter instead
13185 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13186 instead of calls to vect_is_simple_use. Pass the scalar rather
13187 than the vector operand to vect_is_simple_use when handling
13188 second and subsequent copies of an rhs value.
13189 (vectorizable_load): Update calls to vect_check_load_store_mask
13190 and vect_build_gather_load_calls. Use the cached mask_dt and
13191 gs_info.offset_dt instead of calls to vect_is_simple_use.
13193 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13195 PR middle-end/83945
13196 * tree-emutls.c: Include gimplify.h.
13197 (lower_emutls_2): New function.
13198 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13199 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13200 it before further processing.
13203 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13204 UINTVAL (trueop1) instead of INTVAL (op1).
13206 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13210 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13211 INCOMING_FRAME_SP_OFFSET if not defined.
13212 (scan_trace): Add ENTRY argument. If true and
13213 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13214 emit a note to adjust the CFA offset.
13215 (create_cfi_notes): Adjust scan_trace callers.
13216 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13217 INCOMING_FRAME_SP_OFFSET in the CIE.
13218 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13219 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13221 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13222 * doc/tm.texi: Regenerated.
13224 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13226 PR rtl-optimization/83147
13227 * lra-constraints.c (remove_inheritance_pseudos): Use
13228 lra_substitute_pseudo_within_insn.
13230 2018-01-19 Tom de Vries <tom@codesourcery.com>
13231 Cesar Philippidis <cesar@codesourcery.com>
13234 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13236 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13239 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13240 spaces for function labels.
13242 2018-01-19 Martin Liska <mliska@suse.cz>
13244 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13245 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13246 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13247 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13248 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13249 (PRED_CONST_RETURN): Change from 69 to 65.
13250 (PRED_NULL_RETURN): Change from 91 to 71.
13251 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13252 (PRED_LOOP_GUARD): Change from 66 to 73.
13254 2018-01-19 Martin Liska <mliska@suse.cz>
13256 * predict.c (predict_insn_def): Add new assert.
13257 (struct branch_predictor): Change type to signed integer.
13258 (test_prediction_value_range): Amend test to cover
13259 PROB_UNINITIALIZED.
13260 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13261 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13262 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13263 (PRED_LOOP_IV_COMPARE): Likewise.
13264 * predict.h (PROB_UNINITIALIZED): Define new constant.
13266 2018-01-19 Martin Liska <mliska@suse.cz>
13268 * predict.c (dump_prediction): Add new format for
13269 analyze_brprob.py script which is enabled with -details
13271 * profile-count.h (precise_p): New function.
13273 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13275 PR tree-optimization/83922
13276 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13277 there are no statements that need masking.
13278 (vect_active_double_reduction_p): New function.
13279 (vect_analyze_loop_operations): Use it when handling phis that
13280 are not in the loop header.
13282 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13284 PR tree-optimization/83914
13285 * tree-vect-loop.c (vectorizable_induction): Don't convert
13286 init_expr or apply the peeling adjustment for inductions
13287 that are nested within the vectorized loop.
13289 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13291 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13294 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13298 * function.h (gimplify_parameters): Add gimple_seq * argument.
13299 * function.c: Include gimple.h and options.h.
13300 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13301 for the added local temporaries if needed.
13302 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13303 if there are any parameter cleanups, wrap whole body into a
13304 try/finally with the cleanups.
13306 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13309 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13310 Use GET_MODE_CLASS for scalar floating point.
13312 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13316 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13317 Fix call of call_cgraph_insertion_hooks.
13319 2018-01-18 Martin Sebor <msebor@redhat.com>
13321 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13323 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13326 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13329 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13332 * common.opt: (-ffile-prefix-map): New option.
13333 * opts.c (common_handle_option): Defer it.
13334 * opts-global.c (handle_common_deferred_options): Handle it.
13335 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13336 * file-prefix-map.h: New file.
13337 (remap_debug_filename, add_debug_prefix_map): ...here.
13338 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13339 * final.c (debug_prefix_map, add_debug_prefix_map
13340 remap_debug_filename): Move to...
13341 * file-prefix-map.c: New file.
13342 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13343 generalize, get rid of alloca(), use strrchr() instead of strchr().
13344 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13345 Implement in terms of add_prefix_map().
13346 (remap_macro_filename, remap_debug_filename): Implement in term of
13348 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13349 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13350 * dbxout.c: Include file-prefix-map.h.
13351 * varasm.c: Likewise.
13352 * vmsdbgout.c: Likewise.
13353 * xcoffout.c: Likewise.
13354 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13355 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13356 * doc/invoke.texi (-ffile-prefix-map): Document.
13357 (-fdebug-prefix-map): Update description.
13359 2018-01-18 Martin Liska <mliska@suse.cz>
13361 * config/i386/i386.c (indirect_thunk_name): Document that also
13363 (output_indirect_thunk): Document why both instructions
13364 (pause and lfence) are generated.
13366 2018-01-18 Richard Biener <rguenther@suse.de>
13368 PR tree-optimization/83887
13369 * graphite-scop-detection.c
13370 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13371 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13372 (scop_detection::merge_sese): Re-implement with a flood-fill
13373 algorithm that properly finds a SESE region if it exists.
13375 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13378 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13379 pointer_diff optimizations use view_convert instead of convert.
13381 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13383 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13384 Generate different code for -mno-speculate-indirect-jumps.
13385 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13386 (*call_indirect_aix<mode>): Disable for
13387 -mno-speculate-indirect-jumps.
13388 (*call_indirect_aix<mode>_nospec): New define_insn.
13389 (*call_value_indirect_aix<mode>): Disable for
13390 -mno-speculate-indirect-jumps.
13391 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13392 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13393 -mno-speculate-indirect-jumps.
13394 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13396 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13398 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13399 long double type, set the flags for noting the default long double
13400 type, even if we don't pass or return a long double type.
13402 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13405 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13408 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13410 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13411 support for merge[hl].
13412 (fold_mergehl_helper): New helper function.
13413 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13414 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13415 (altivec_vmrglw_direct): Add xxmrglw insn.
13417 2018-01-17 Andrew Waterman <andrew@sifive.com>
13419 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13420 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13422 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13425 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13426 call the lto_location_cache before reading the
13427 DECL_SOURCE_LOCATION of the types.
13429 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13430 Richard Sandiford <richard.sandiford@linaro.org>
13432 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13433 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13434 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13435 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13436 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13438 * config/aarch64/constraints.md (aarch64_movti_operand):
13440 * config/aarch64/predicates.md (Uti): Add new constraint.
13442 2018-01-17 Carl Love <cel@us.ibm.com>
13444 * config/rs6000/vsx.md (define_expand xl_len_r,
13445 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13446 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13448 (define_expand, define_insn): Move the shift left from the
13449 define_insn to the define_expand for lxvl and stxvl instructions.
13450 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13451 and XL_LEN_R definitions to PURE.
13453 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13455 * config/i386/i386.c (indirect_thunk_name): Declare regno
13456 as unsigned int. Compare regno with INVALID_REGNUM.
13457 (output_indirect_thunk): Ditto.
13458 (output_indirect_thunk_function): Ditto.
13459 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13460 in the call to output_indirect_thunk_function.
13462 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13464 PR middle-end/83884
13465 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13466 rather than the size of inner_type to determine the stack slot size
13467 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13469 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13472 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13475 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13477 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13478 endian Linux systems to optionally enable multilibs for selecting
13479 the long double type if the user configured an explicit type.
13480 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13481 have no long double multilibs if not defined.
13482 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13483 warn if the user used -mabi={ieee,ibm}longdouble and we built
13484 multilibs for long double.
13485 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13486 appropriate multilib option.
13487 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13489 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13490 for building long double multilibs.
13491 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13493 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13495 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13498 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13500 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13503 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13506 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13509 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13511 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13512 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13514 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13516 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13517 different rtl trees depending on TARGET_64BIT.
13518 (rs6000_gen_lvx): Likewise.
13520 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13522 * config/visium/visium.md (nop): Tweak comment.
13523 (hazard_nop): Likewise.
13525 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13527 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13528 -mspeculate-indirect-jumps.
13529 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13530 for -mno-speculate-indirect-jumps.
13531 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13532 (*call_value_indirect_elfv2<mode>): Disable for
13533 -mno-speculate-indirect-jumps.
13534 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13535 (indirect_jump): Emit different RTL for
13536 -mno-speculate-indirect-jumps.
13537 (*indirect_jump<mode>): Disable for
13538 -mno-speculate-indirect-jumps.
13539 (*indirect_jump<mode>_nospec): New define_insn.
13540 (tablejump): Emit different RTL for
13541 -mno-speculate-indirect-jumps.
13542 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13543 (tablejumpsi_nospec): New define_expand.
13544 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13545 (tablejumpdi_nospec): New define_expand.
13546 (*tablejump<mode>_internal1): Disable for
13547 -mno-speculate-indirect-jumps.
13548 (*tablejump<mode>_internal1_nospec): New define_insn.
13549 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13552 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13554 * caller-save.c (insert_save): Drop unnecessary parameter. All
13557 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13558 Richard Biener <rguenth@suse.de>
13561 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13562 return early, inline manually is_gimple_sizepos. Make sure if we
13563 call gimplify_expr we don't end up with a gimple constant.
13564 * tree.c (variably_modified_type_p): Don't return true for
13565 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13566 * gimplify.h (is_gimple_sizepos): Remove.
13568 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13570 PR tree-optimization/83857
13571 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13572 vectorizable_live_operation for pure SLP statements.
13573 (vectorizable_live_operation): Handle PHIs.
13575 2018-01-16 Richard Biener <rguenther@suse.de>
13577 PR tree-optimization/83867
13578 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13579 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13581 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13584 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13585 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13586 If off is not INTEGER_CST, issue a may not be aligned warning
13587 rather than isn't aligned. Use isn%'t rather than isn't.
13588 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13590 <case MULT_EXPR>: Improve the case when bottom and one of the
13591 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13592 operand, in that case check if the other operand is multiple of
13593 bottom divided by the INTEGER_CST operand.
13595 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13598 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13599 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13600 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13601 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13602 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13603 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13604 (pa_function_arg_size): New function.
13606 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13608 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13609 in a separate statement.
13611 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13613 PR tree-optimization/83847
13614 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13615 group gathers and scatters.
13617 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13619 PR rtl-optimization/86620
13620 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13622 PR rtl-optimization/83213
13623 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13624 to last if both are JUMP_INSNs.
13626 PR tree-optimization/83843
13627 * gimple-ssa-store-merging.c
13628 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13629 store_immediate_info for bswap/nop orig_stores.
13631 2018-01-15 Andrew Waterman <andrew@sifive.com>
13633 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13635 <UDIV>: Increase cost if !TARGET_DIV.
13637 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13639 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13640 (define_attr "cr_logical_3op"): New.
13641 (cceq_ior_compare): Adjust.
13642 (cceq_ior_compare_complement): Adjust.
13643 (*cceq_rev_compare): Adjust.
13644 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13645 (is_cracked_insn): Adjust.
13646 (insn_must_be_first_in_group): Adjust.
13647 * config/rs6000/40x.md: Adjust.
13648 * config/rs6000/440.md: Adjust.
13649 * config/rs6000/476.md: Adjust.
13650 * config/rs6000/601.md: Adjust.
13651 * config/rs6000/603.md: Adjust.
13652 * config/rs6000/6xx.md: Adjust.
13653 * config/rs6000/7450.md: Adjust.
13654 * config/rs6000/7xx.md: Adjust.
13655 * config/rs6000/8540.md: Adjust.
13656 * config/rs6000/cell.md: Adjust.
13657 * config/rs6000/e300c2c3.md: Adjust.
13658 * config/rs6000/e500mc.md: Adjust.
13659 * config/rs6000/e500mc64.md: Adjust.
13660 * config/rs6000/e5500.md: Adjust.
13661 * config/rs6000/e6500.md: Adjust.
13662 * config/rs6000/mpc.md: Adjust.
13663 * config/rs6000/power4.md: Adjust.
13664 * config/rs6000/power5.md: Adjust.
13665 * config/rs6000/power6.md: Adjust.
13666 * config/rs6000/power7.md: Adjust.
13667 * config/rs6000/power8.md: Adjust.
13668 * config/rs6000/power9.md: Adjust.
13669 * config/rs6000/rs64.md: Adjust.
13670 * config/rs6000/titan.md: Adjust.
13672 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13674 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13675 ix86_indirect_branch_register logic.
13677 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13679 * config/i386/constraints.md (Bs): Update
13680 ix86_indirect_branch_register check. Don't check
13681 ix86_indirect_branch_register with GOT_memory_operand.
13683 * config/i386/predicates.md (GOT_memory_operand): Don't check
13684 ix86_indirect_branch_register here.
13685 (GOT32_symbol_operand): Likewise.
13687 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13689 * config/i386/predicates.md (constant_call_address_operand):
13690 Rewrite ix86_indirect_branch_register logic.
13691 (sibcall_insn_operand): Likewise.
13693 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13695 * config/i386/constraints.md (Bs): Replace
13696 ix86_indirect_branch_thunk_register with
13697 ix86_indirect_branch_register.
13699 * config/i386/i386.md (indirect_jump): Likewise.
13700 (tablejump): Likewise.
13701 (*sibcall_memory): Likewise.
13702 (*sibcall_value_memory): Likewise.
13703 Peepholes of indirect call and jump via memory: Likewise.
13704 * config/i386/i386.opt: Likewise.
13705 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13706 (GOT_memory_operand): Likewise.
13707 (call_insn_operand): Likewise.
13708 (sibcall_insn_operand): Likewise.
13709 (GOT32_symbol_operand): Likewise.
13711 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13713 PR middle-end/83837
13714 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13715 type rather than type addr's type points to.
13716 (expand_omp_atomic_mutex): Likewise.
13717 (expand_omp_atomic): Likewise.
13719 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13722 * config/i386/i386.c (output_indirect_thunk_function): Use
13723 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13724 for __x86_return_thunk.
13726 2018-01-15 Richard Biener <rguenther@suse.de>
13728 PR middle-end/83850
13729 * expmed.c (extract_bit_field_1): Fix typo.
13731 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13734 * config/arm/iterators.md (VF): New mode iterator.
13735 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13736 Remove integer-related logic from pattern.
13737 (neon_vabd<mode>_3): Likewise.
13739 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13741 PR middle-end/82694
13742 * common.opt (fstrict-overflow): No longer an alias.
13743 (fwrapv-pointer): New option.
13744 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13745 also for pointer types based on flag_wrapv_pointer.
13746 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13747 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13748 opts->x_flag_wrapv got set.
13749 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13750 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13751 POINTER_TYPE_OVERFLOW_UNDEFINED.
13752 * match.pd: Likewise in address comparison pattern.
13753 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13755 2018-01-15 Richard Biener <rguenther@suse.de>
13758 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13759 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13760 Reset type names to their identifier if their TYPE_DECL doesn't
13761 have linkage (and thus is used for ODR and devirt).
13762 (save_debug_info_for_decl): Remove.
13763 (save_debug_info_for_type): Likewise.
13764 (add_tree_to_fld_list): Adjust.
13765 * tree-pretty-print.c (dump_generic_node): Make dumping of
13766 type names more robust.
13768 2018-01-15 Richard Biener <rguenther@suse.de>
13770 * BASE-VER: Bump to 8.0.1.
13772 2018-01-14 Martin Sebor <msebor@redhat.com>
13775 * builtins.c (check_access): Avoid warning when the no-warning bit
13778 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13780 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13781 * ira-color (allocno_hard_regs_compare): Likewise.
13783 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13786 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13787 Use .pushsection/.popsection.
13789 2018-01-14 Martin Sebor <msebor@redhat.com>
13792 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13794 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13796 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13797 entry from extra_headers.
13798 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13799 extra_headers, make the list bitwise identical to the i?86-*-* one.
13801 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13803 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13804 -mcmodel=large with -mindirect-branch=thunk,
13805 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13806 -mfunction-return=thunk-extern.
13807 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13808 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13809 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13811 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13813 * config/i386/i386.c (print_reg): Print the name of the full
13814 integer register without '%'.
13815 (ix86_print_operand): Handle 'V'.
13816 * doc/extend.texi: Document 'V' modifier.
13818 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13820 * config/i386/constraints.md (Bs): Disallow memory operand for
13821 -mindirect-branch-register.
13823 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13824 (GOT_memory_operand): Likewise.
13825 (call_insn_operand): Likewise.
13826 (sibcall_insn_operand): Likewise.
13827 (GOT32_symbol_operand): Likewise.
13828 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13829 for -mindirect-branch-register.
13830 (tablejump): Likewise.
13831 (*sibcall_memory): Likewise.
13832 (*sibcall_value_memory): Likewise.
13833 Disallow peepholes of indirect call and jump via memory for
13834 -mindirect-branch-register.
13835 (*call_pop): Replace m with Bw.
13836 (*call_value_pop): Likewise.
13837 (*sibcall_pop_memory): Replace m with Bs.
13838 * config/i386/i386.opt (mindirect-branch-register): New option.
13839 * doc/invoke.texi: Document -mindirect-branch-register option.
13841 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13843 * config/i386/i386-protos.h (ix86_output_function_return): New.
13844 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13845 set function_return_type.
13846 (indirect_thunk_name): Add ret_p to indicate thunk for function
13848 (output_indirect_thunk_function): Pass false to
13849 indirect_thunk_name.
13850 (ix86_output_indirect_branch_via_reg): Likewise.
13851 (ix86_output_indirect_branch_via_push): Likewise.
13852 (output_indirect_thunk_function): Create alias for function
13853 return thunk if regno < 0.
13854 (ix86_output_function_return): New function.
13855 (ix86_handle_fndecl_attribute): Handle function_return.
13856 (ix86_attribute_table): Add function_return.
13857 * config/i386/i386.h (machine_function): Add
13858 function_return_type.
13859 * config/i386/i386.md (simple_return_internal): Use
13860 ix86_output_function_return.
13861 (simple_return_internal_long): Likewise.
13862 * config/i386/i386.opt (mfunction-return=): New option.
13863 (indirect_branch): Mention -mfunction-return=.
13864 * doc/extend.texi: Document function_return function attribute.
13865 * doc/invoke.texi: Document -mfunction-return= option.
13867 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13869 * config/i386/i386-opts.h (indirect_branch): New.
13870 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13871 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13872 with local indirect jump when converting indirect call and jump.
13873 (ix86_set_indirect_branch_type): New.
13874 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13875 (indirectlabelno): New.
13876 (indirect_thunk_needed): Likewise.
13877 (indirect_thunk_bnd_needed): Likewise.
13878 (indirect_thunks_used): Likewise.
13879 (indirect_thunks_bnd_used): Likewise.
13880 (INDIRECT_LABEL): Likewise.
13881 (indirect_thunk_name): Likewise.
13882 (output_indirect_thunk): Likewise.
13883 (output_indirect_thunk_function): Likewise.
13884 (ix86_output_indirect_branch_via_reg): Likewise.
13885 (ix86_output_indirect_branch_via_push): Likewise.
13886 (ix86_output_indirect_branch): Likewise.
13887 (ix86_output_indirect_jmp): Likewise.
13888 (ix86_code_end): Call output_indirect_thunk_function if needed.
13889 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13891 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13892 (ix86_attribute_table): Add indirect_branch.
13893 * config/i386/i386.h (machine_function): Add indirect_branch_type
13894 and has_local_indirect_jump.
13895 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13897 (tablejump): Likewise.
13898 (*indirect_jump): Use ix86_output_indirect_jmp.
13899 (*tablejump_1): Likewise.
13900 (simple_return_indirect_internal): Likewise.
13901 * config/i386/i386.opt (mindirect-branch=): New option.
13902 (indirect_branch): New.
13905 (thunk-inline): Likewise.
13906 (thunk-extern): Likewise.
13907 * doc/extend.texi: Document indirect_branch function attribute.
13908 * doc/invoke.texi: Document -mindirect-branch= option.
13910 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13913 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13915 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13917 * ipa-inline.c (want_inline_small_function_p): Return false if
13918 inlining has already failed with CIF_FINAL_ERROR.
13919 (update_caller_keys): Call want_inline_small_function_p before
13921 (update_callee_keys): Likewise.
13923 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13925 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13927 (rs6000_quadword_masked_address_p): Likewise.
13928 (quad_aligned_load_p): Likewise.
13929 (quad_aligned_store_p): Likewise.
13930 (const_load_sequence_p): Add comment to describe the outer-most loop.
13931 (mimic_memory_attributes_and_flags): New function.
13932 (rs6000_gen_stvx): Likewise.
13933 (replace_swapped_aligned_store): Likewise.
13934 (rs6000_gen_lvx): Likewise.
13935 (replace_swapped_aligned_load): Likewise.
13936 (replace_swapped_load_constant): Capitalize argument name in
13937 comment describing this function.
13938 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13939 and stores that access quad-word aligned addresses and replace
13940 with stvx or lvx instructions when appropriate.
13941 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13942 New function prototype.
13943 (rs6000_quadword_masked_address_p): Likewise.
13944 (rs6000_gen_lvx): Likewise.
13945 (rs6000_gen_stvx): Likewise.
13946 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13947 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13948 when memory address is aligned.
13949 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13950 this split to select lvx instruction when memory address is aligned.
13951 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13952 instruction when memory address is aligned.
13953 (*vsx_le_perm_load_v16qi): Likewise.
13954 (four unnamed splitters): Modify to select the stvx instruction
13955 when memory is aligned.
13957 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13959 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13960 which appears in the queue multiple times.
13962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13963 Alan Hayward <alan.hayward@arm.com>
13964 David Sherwood <david.sherwood@arm.com>
13966 * tree-vectorizer.h (vec_lower_bound): New structure.
13967 (_loop_vec_info): Add check_nonzero and lower_bounds.
13968 (LOOP_VINFO_CHECK_NONZERO): New macro.
13969 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13970 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13971 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13972 fields. Make seg_len the distance travelled, not including the
13974 (dr_direction_indicator): Declare.
13975 (dr_zero_step_indicator): Likewise.
13976 (dr_known_forward_stride_p): Likewise.
13977 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13979 (runtime_alias_check_p): Allow runtime alias checks with
13981 (operator ==): Compare access_size and align.
13982 (prune_runtime_alias_test_list): Rework for new distinction between
13983 the access_size and seg_len.
13984 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13986 (get_segment_min_max): New function.
13987 (create_intersect_range_checks): Use it.
13988 (dr_step_indicator): New function.
13989 (dr_direction_indicator): Likewise.
13990 (dr_zero_step_indicator): Likewise.
13991 (dr_known_forward_stride_p): Likewise.
13992 * tree-loop-distribution.c (data_ref_segment_size): Return
13993 DR_STEP * (niters - 1).
13994 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13996 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13997 (vect_preserves_scalar_order_p): New function, split out from...
13998 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13999 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14000 (vect_vfa_access_size): New function.
14001 (vect_vfa_align): Likewise.
14002 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14003 (dump_lower_bound): New function.
14004 (vect_check_lower_bound): Likewise.
14005 (vect_small_gap_p): Likewise.
14006 (vectorizable_with_step_bound_p): Likewise.
14007 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14008 depencies if the vectorization factor is 1. Convert the checks
14009 for nonzero steps into checks on the bounds of DR_STEP. Try using
14010 a bunds check for variable steps if the minimum required step is
14011 relatively small. Update calls to the dr_with_seg_len
14012 constructor and to vect_compile_time_alias.
14013 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14015 (vect_loop_versioning): Call it.
14016 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14018 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14021 Alan Hayward <alan.hayward@arm.com>
14022 David Sherwood <david.sherwood@arm.com>
14024 * doc/sourcebuild.texi (vect_scatter_store): Document.
14025 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14027 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14029 * genopinit.c (main): Add supports_vec_scatter_store and
14030 supports_vec_scatter_store_cached to target_optabs.
14031 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14032 IFN_MASK_SCATTER_STORE.
14033 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14035 * internal-fn.h (internal_store_fn_p): Declare.
14036 (internal_fn_stored_value_index): Likewise.
14037 * internal-fn.c (scatter_store_direct): New macro.
14038 (expand_scatter_store_optab_fn): New function.
14039 (direct_scatter_store_optab_supported_p): New macro.
14040 (internal_store_fn_p): New function.
14041 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14042 IFN_MASK_SCATTER_STORE.
14043 (internal_fn_mask_index): Likewise.
14044 (internal_fn_stored_value_index): New function.
14045 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14046 for scatter stores.
14047 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14048 * optabs-query.c (supports_vec_scatter_store_p): New function.
14049 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14050 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14051 true for scatter stores.
14052 (vect_gather_scatter_fn_p): Handle scatter stores too.
14053 (vect_check_gather_scatter): Consider using scatter stores if
14054 supports_vec_scatter_store_p.
14055 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14056 scatter stores too.
14057 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14058 internal_fn_stored_value_index.
14059 (check_load_store_masking): Handle scatter stores too.
14060 (vect_get_store_rhs): Make public.
14061 (vectorizable_call): Use internal_store_fn_p.
14062 (vectorizable_store): Handle scatter store internal functions.
14063 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14064 when deciding whether the end of the group has been reached.
14065 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14066 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14067 (mask_scatter_store<mode>): New insns.
14069 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14070 Alan Hayward <alan.hayward@arm.com>
14071 David Sherwood <david.sherwood@arm.com>
14073 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14074 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14075 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14077 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14078 Use vect_truncate_gather_scatter_offset if we can't treat the
14079 operation as a normal gather load or scatter store.
14080 (get_group_load_store_type): Take the gather_scatter_info
14081 as argument. Try using a gather load or scatter store for
14082 single-element groups.
14083 (get_load_store_type): Update calls to get_group_load_store_type
14084 and vect_use_strided_gather_scatters_p.
14086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14087 Alan Hayward <alan.hayward@arm.com>
14088 David Sherwood <david.sherwood@arm.com>
14090 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14091 optional tree argument.
14092 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14094 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14095 but continue to use the current value as a fallback.
14096 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14097 to compare the updates.
14098 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14099 (get_load_store_type): Use it when handling a strided access.
14100 (vect_get_strided_load_store_ops): New function.
14101 (vect_get_data_ptr_increment): Likewise.
14102 (vectorizable_load): Handle strided gather loads. Always pass
14103 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14106 Alan Hayward <alan.hayward@arm.com>
14107 David Sherwood <david.sherwood@arm.com>
14109 * doc/md.texi (gather_load@var{m}): Document.
14110 (mask_gather_load@var{m}): Likewise.
14111 * genopinit.c (main): Add supports_vec_gather_load and
14112 supports_vec_gather_load_cached to target_optabs.
14113 * optabs-tree.c (init_tree_optimization_optabs): Use
14114 ggc_cleared_alloc to allocate target_optabs.
14115 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14116 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14118 * internal-fn.h (internal_load_fn_p): Declare.
14119 (internal_gather_scatter_fn_p): Likewise.
14120 (internal_fn_mask_index): Likewise.
14121 (internal_gather_scatter_fn_supported_p): Likewise.
14122 * internal-fn.c (gather_load_direct): New macro.
14123 (expand_gather_load_optab_fn): New function.
14124 (direct_gather_load_optab_supported_p): New macro.
14125 (direct_internal_fn_optab): New function.
14126 (internal_load_fn_p): Likewise.
14127 (internal_gather_scatter_fn_p): Likewise.
14128 (internal_fn_mask_index): Likewise.
14129 (internal_gather_scatter_fn_supported_p): Likewise.
14130 * optabs-query.c (supports_at_least_one_mode_p): New function.
14131 (supports_vec_gather_load_p): Likewise.
14132 * optabs-query.h (supports_vec_gather_load_p): Declare.
14133 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14134 and memory_type field.
14135 (NUM_PATTERNS): Bump to 15.
14136 * tree-vect-data-refs.c: Include internal-fn.h.
14137 (vect_gather_scatter_fn_p): New function.
14138 (vect_describe_gather_scatter_call): Likewise.
14139 (vect_check_gather_scatter): Try using internal functions for
14140 gather loads. Recognize existing calls to a gather load function.
14141 (vect_analyze_data_refs): Consider using gather loads if
14142 supports_vec_gather_load_p.
14143 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14144 (vect_get_gather_scatter_offset_type): Likewise.
14145 (vect_convert_mask_for_vectype): Likewise.
14146 (vect_add_conversion_to_patterm): Likewise.
14147 (vect_try_gather_scatter_pattern): Likewise.
14148 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14149 (vect_vect_recog_func_ptrs): Add it.
14150 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14151 internal_fn_mask_index and internal_gather_scatter_fn_p.
14152 (check_load_store_masking): Take the gather_scatter_info as an
14153 argument and handle gather loads.
14154 (vect_get_gather_scatter_ops): New function.
14155 (vectorizable_call): Check internal_load_fn_p.
14156 (vectorizable_load): Likewise. Handle gather load internal
14158 (vectorizable_store): Update call to check_load_store_masking.
14159 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14160 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14161 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14162 (aarch64_gather_scale_operand_d): New predicates.
14163 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14164 (mask_gather_load<mode>): New insns.
14166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14167 Alan Hayward <alan.hayward@arm.com>
14168 David Sherwood <david.sherwood@arm.com>
14170 * optabs.def (fold_left_plus_optab): New optab.
14171 * doc/md.texi (fold_left_plus_@var{m}): Document.
14172 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14173 * internal-fn.c (fold_left_direct): Define.
14174 (expand_fold_left_optab_fn): Likewise.
14175 (direct_fold_left_optab_supported_p): Likewise.
14176 * fold-const-call.c (fold_const_fold_left): New function.
14177 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14178 * tree-parloops.c (valid_reduction_p): New function.
14179 (gather_scalar_reductions): Use it.
14180 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14181 (vect_finish_replace_stmt): Declare.
14182 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14183 (needs_fold_left_reduction_p): New function, split out from...
14184 (vect_is_simple_reduction): ...here. Accept reductions that
14185 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14186 (vect_force_simple_reduction): Also store the reduction type in
14187 the assignment's STMT_VINFO_REDUC_TYPE.
14188 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14189 (merge_with_identity): New function.
14190 (vect_expand_fold_left): Likewise.
14191 (vectorize_fold_left_reduction): Likewise.
14192 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14193 scalar phi in place for it. Check for target support and reject
14194 cases that would reassociate the operation. Defer the transform
14195 phase to vectorize_fold_left_reduction.
14196 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14197 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14198 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14202 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14203 call to ifc_temp_var.
14205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14206 Alan Hayward <alan.hayward@arm.com>
14207 David Sherwood <david.sherwood@arm.com>
14209 * target.def (legitimize_address_displacement): Take the original
14210 offset as a poly_int.
14211 * targhooks.h (default_legitimize_address_displacement): Update
14213 * targhooks.c (default_legitimize_address_displacement): Likewise.
14214 * doc/tm.texi: Regenerate.
14215 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14216 as an argument, moving assert of ad->disp == ad->disp_term to...
14217 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14218 Try calling targetm.legitimize_address_displacement before expanding
14219 the address rather than afterwards, and adjust for the new interface.
14220 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14221 Match the new hook interface. Handle SVE addresses.
14222 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14223 new hook interface.
14225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14227 * Makefile.in (OBJS): Add early-remat.o.
14228 * target.def (select_early_remat_modes): New hook.
14229 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14230 * doc/tm.texi: Regenerate.
14231 * targhooks.h (default_select_early_remat_modes): Declare.
14232 * targhooks.c (default_select_early_remat_modes): New function.
14233 * timevar.def (TV_EARLY_REMAT): New timevar.
14234 * passes.def (pass_early_remat): New pass.
14235 * tree-pass.h (make_pass_early_remat): Declare.
14236 * early-remat.c: New file.
14237 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14239 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14242 Alan Hayward <alan.hayward@arm.com>
14243 David Sherwood <david.sherwood@arm.com>
14245 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14246 vfm1 with a bound_epilog parameter.
14247 (vect_do_peeling): Update calls accordingly, and move the prologue
14248 call earlier in the function. Treat the base bound_epilog as 0 for
14249 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14250 this base when peeling for gaps.
14251 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14252 with fully-masked loops.
14253 (vect_estimate_min_profitable_iters): Handle the single peeled
14254 iteration in that case.
14256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14257 Alan Hayward <alan.hayward@arm.com>
14258 David Sherwood <david.sherwood@arm.com>
14260 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14261 single-element interleaving even if the size is not a power of 2.
14262 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14263 accesses for single-element interleaving if the group size is
14266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14267 Alan Hayward <alan.hayward@arm.com>
14268 David Sherwood <david.sherwood@arm.com>
14270 * doc/md.texi (fold_extract_last_@var{m}): Document.
14271 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14272 * optabs.def (fold_extract_last_optab): New optab.
14273 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14274 * internal-fn.c (fold_extract_direct): New macro.
14275 (expand_fold_extract_optab_fn): Likewise.
14276 (direct_fold_extract_optab_supported_p): Likewise.
14277 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14278 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14279 EXTRACT_LAST_REDUCTION.
14280 (get_initial_def_for_reduction): Do not create an initial vector
14281 for EXTRACT_LAST_REDUCTION reductions.
14282 (vectorizable_reduction): Leave the scalar phi in place for
14283 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14284 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14285 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14286 transform phase to vectorizable_condition.
14287 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14289 (vect_finish_stmt_generation): ...here.
14290 (vect_finish_replace_stmt): New function.
14291 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14292 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14294 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14297 Alan Hayward <alan.hayward@arm.com>
14298 David Sherwood <david.sherwood@arm.com>
14300 * doc/md.texi (extract_last_@var{m}): Document.
14301 * optabs.def (extract_last_optab): New optab.
14302 * internal-fn.def (EXTRACT_LAST): New internal function.
14303 * internal-fn.c (cond_unary_direct): New macro.
14304 (expand_cond_unary_optab_fn): Likewise.
14305 (direct_cond_unary_optab_supported_p): Likewise.
14306 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14307 loops using EXTRACT_LAST.
14308 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14309 (extract_last_<mode>): ...this optab.
14310 (vec_extract<mode><Vel>): Update accordingly.
14312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14313 Alan Hayward <alan.hayward@arm.com>
14314 David Sherwood <david.sherwood@arm.com>
14316 * target.def (empty_mask_is_expensive): New hook.
14317 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14318 * doc/tm.texi: Regenerate.
14319 * targhooks.h (default_empty_mask_is_expensive): Declare.
14320 * targhooks.c (default_empty_mask_is_expensive): New function.
14321 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14322 if the target says that empty masks are expensive.
14323 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14325 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14328 Alan Hayward <alan.hayward@arm.com>
14329 David Sherwood <david.sherwood@arm.com>
14331 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14332 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14333 (vect_use_loop_mask_for_alignment_p): New function.
14334 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14335 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14336 niters_skip argument. Make sure that the first niters_skip elements
14337 of the first iteration are inactive.
14338 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14339 Update call to vect_set_loop_masks_directly.
14340 (get_misalign_in_elems): New function, split out from...
14341 (vect_gen_prolog_loop_niters): ...here.
14342 (vect_update_init_of_dr): Take a code argument that specifies whether
14343 the adjustment should be added or subtracted.
14344 (vect_update_init_of_drs): Likewise.
14345 (vect_prepare_for_masked_peels): New function.
14346 (vect_do_peeling): Skip prologue peeling if we're using a mask
14347 instead. Update call to vect_update_inits_of_drs.
14348 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14350 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14351 alignment. Do not include the number of peeled iterations in
14352 the minimum threshold in that case.
14353 (vectorizable_induction): Adjust the start value down by
14354 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14355 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14356 Take the number of skipped iterations into account when calculating
14358 * tree-vect-stmts.c (vect_gen_while_not): New function.
14360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14361 Alan Hayward <alan.hayward@arm.com>
14362 David Sherwood <david.sherwood@arm.com>
14364 * doc/sourcebuild.texi (vect_fully_masked): Document.
14365 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14366 default value to 0.
14367 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14369 (vect_analyze_loop_2): ...here. Don't check the vectorization
14370 factor against the number of loop iterations if the loop is
14373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14374 Alan Hayward <alan.hayward@arm.com>
14375 David Sherwood <david.sherwood@arm.com>
14377 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14378 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14379 (dump_groups): Update accordingly.
14380 (iv_use::mem_type): New member variable.
14381 (address_p): New function.
14382 (record_use): Add a mem_type argument and initialize the new
14384 (record_group_use): Add a mem_type argument. Use address_p.
14385 Remove obsolete null checks of base_object. Update call to record_use.
14386 (find_interesting_uses_op): Update call to record_group_use.
14387 (find_interesting_uses_cond): Likewise.
14388 (find_interesting_uses_address): Likewise.
14389 (get_mem_type_for_internal_fn): New function.
14390 (find_address_like_use): Likewise.
14391 (find_interesting_uses_stmt): Try find_address_like_use before
14392 calling find_interesting_uses_op.
14393 (addr_offset_valid_p): Use the iv mem_type field as the type
14394 of the addressed memory.
14395 (add_autoinc_candidates): Likewise.
14396 (get_address_cost): Likewise.
14397 (split_small_address_groups_p): Use address_p.
14398 (split_address_groups): Likewise.
14399 (add_iv_candidate_for_use): Likewise.
14400 (autoinc_possible_for_pair): Likewise.
14401 (rewrite_groups): Likewise.
14402 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14403 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14404 (get_alias_ptr_type_for_ptr_address): New function.
14405 (rewrite_use_address): Rewrite address uses in calls that were
14406 identified by find_address_like_use.
14408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14409 Alan Hayward <alan.hayward@arm.com>
14410 David Sherwood <david.sherwood@arm.com>
14412 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14414 * gimple-expr.h (is_gimple_addressable: Likewise.
14415 * gimple-expr.c (is_gimple_address): Likewise.
14416 * internal-fn.c (expand_call_mem_ref): New function.
14417 (expand_mask_load_optab_fn): Use it.
14418 (expand_mask_store_optab_fn): Likewise.
14420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14421 Alan Hayward <alan.hayward@arm.com>
14422 David Sherwood <david.sherwood@arm.com>
14424 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14425 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14426 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14427 (cond_umax@var{mode}): Document.
14428 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14429 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14430 (cond_umin_optab, cond_umax_optab): New optabs.
14431 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14432 (COND_IOR, COND_XOR): New internal functions.
14433 * internal-fn.h (get_conditional_internal_fn): Declare.
14434 * internal-fn.c (cond_binary_direct): New macro.
14435 (expand_cond_binary_optab_fn): Likewise.
14436 (direct_cond_binary_optab_supported_p): Likewise.
14437 (get_conditional_internal_fn): New function.
14438 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14439 Cope with reduction statements that are vectorized as calls rather
14441 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14442 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14443 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14444 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14445 (UNSPEC_COND_EOR): New unspecs.
14446 (optab): Add mappings for them.
14447 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14448 (sve_int_op, sve_fp_op): New int attributes.
14450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14451 Alan Hayward <alan.hayward@arm.com>
14452 David Sherwood <david.sherwood@arm.com>
14454 * optabs.def (while_ult_optab): New optab.
14455 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14456 * internal-fn.def (WHILE_ULT): New internal function.
14457 * internal-fn.h (direct_internal_fn_supported_p): New override
14458 that takes two types as argument.
14459 * internal-fn.c (while_direct): New macro.
14460 (expand_while_optab_fn): New function.
14461 (convert_optab_supported_p): Likewise.
14462 (direct_while_optab_supported_p): New macro.
14463 * wide-int.h (wi::udiv_ceil): New function.
14464 * tree-vectorizer.h (rgroup_masks): New structure.
14465 (vec_loop_masks): New typedef.
14466 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14467 and fully_masked_p.
14468 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14469 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14470 (vect_max_vf): New function.
14471 (slpeel_make_loop_iterate_ntimes): Delete.
14472 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14473 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14474 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14475 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14476 internal-fn.h, stor-layout.h and optabs-query.h.
14477 (vect_set_loop_mask): New function.
14478 (add_preheader_seq): Likewise.
14479 (add_header_seq): Likewise.
14480 (interleave_supported_p): Likewise.
14481 (vect_maybe_permute_loop_masks): Likewise.
14482 (vect_set_loop_masks_directly): Likewise.
14483 (vect_set_loop_condition_masked): Likewise.
14484 (vect_set_loop_condition_unmasked): New function, split out from
14485 slpeel_make_loop_iterate_ntimes.
14486 (slpeel_make_loop_iterate_ntimes): Rename to..
14487 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14488 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14489 (vect_do_peeling): Update call accordingly.
14490 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14492 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14493 mask_compare_type, can_fully_mask_p and fully_masked_p.
14494 (release_vec_loop_masks): New function.
14495 (_loop_vec_info): Use it to free the loop masks.
14496 (can_produce_all_loop_masks_p): New function.
14497 (vect_get_max_nscalars_per_iter): Likewise.
14498 (vect_verify_full_masking): Likewise.
14499 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14500 retries, and free the mask rgroups before retrying. Check loop-wide
14501 reasons for disallowing fully-masked loops. Make the final decision
14502 about whether use a fully-masked loop or not.
14503 (vect_estimate_min_profitable_iters): Do not assume that peeling
14504 for the number of iterations will be needed for fully-masked loops.
14505 (vectorizable_reduction): Disable fully-masked loops.
14506 (vectorizable_live_operation): Likewise.
14507 (vect_halve_mask_nunits): New function.
14508 (vect_double_mask_nunits): Likewise.
14509 (vect_record_loop_mask): Likewise.
14510 (vect_get_loop_mask): Likewise.
14511 (vect_transform_loop): Handle the case in which the final loop
14512 iteration might handle a partial vector. Call vect_set_loop_condition
14513 instead of slpeel_make_loop_iterate_ntimes.
14514 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14515 (check_load_store_masking): New function.
14516 (prepare_load_store_mask): Likewise.
14517 (vectorizable_store): Handle fully-masked loops.
14518 (vectorizable_load): Likewise.
14519 (supportable_widening_operation): Use vect_halve_mask_nunits for
14521 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14522 (vect_gen_while): New function.
14523 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14524 (aarch64_uqdec<mode>): New insn.
14526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14527 Alan Hayward <alan.hayward@arm.com>
14528 David Sherwood <david.sherwood@arm.com>
14530 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14531 (reduc_xor_scal_optab): New optabs.
14532 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14533 (reduc_xor_scal_@var{m}): Document.
14534 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14535 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14536 internal functions.
14537 * fold-const-call.c (fold_const_call): Handle them.
14538 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14539 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14540 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14541 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14542 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14543 (UNSPEC_XORV): New unspecs.
14544 (optab): Add entries for them.
14545 (BITWISEV): New int iterator.
14546 (bit_reduc_op): New int attributes.
14548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14549 Alan Hayward <alan.hayward@arm.com>
14550 David Sherwood <david.sherwood@arm.com>
14552 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14553 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14554 * optabs.def (vec_shl_insert_optab): New optab.
14555 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14556 (duplicate_and_interleave): Likewise.
14557 * tree-vect-loop.c: Include internal-fn.h.
14558 (neutral_op_for_slp_reduction): New function, split out from
14559 get_initial_defs_for_reduction.
14560 (get_initial_def_for_reduction): Handle option 2 for variable-length
14561 vectors by loading the neutral value into a vector and then shifting
14562 the initial value into element 0.
14563 (get_initial_defs_for_reduction): Replace the code argument with
14564 the neutral value calculated by neutral_op_for_slp_reduction.
14565 Use gimple_build_vector for constant-length vectors.
14566 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14567 but the first group_size elements have a neutral value.
14568 Use duplicate_and_interleave otherwise.
14569 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14570 Update call to get_initial_defs_for_reduction. Handle SLP
14571 reductions for variable-length vectors by creating one vector
14572 result for each scalar result, with the elements associated
14573 with other scalar results stubbed out with the neutral value.
14574 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14575 Require IFN_VEC_SHL_INSERT for double reductions on
14576 variable-length vectors, or SLP reductions that have
14577 a neutral value. Require can_duplicate_and_interleave_p
14578 support for variable-length unchained SLP reductions if there
14579 is no neutral value, such as for MIN/MAX reductions. Also require
14580 the number of vector elements to be a multiple of the number of
14581 SLP statements when doing variable-length unchained SLP reductions.
14582 Update call to vect_create_epilog_for_reduction.
14583 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14584 and remove initial values.
14585 (duplicate_and_interleave): Make public.
14586 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14587 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14590 Alan Hayward <alan.hayward@arm.com>
14591 David Sherwood <david.sherwood@arm.com>
14593 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14594 (can_duplicate_and_interleave_p): New function.
14595 (vect_get_and_check_slp_defs): Take the vector of statements
14596 rather than just the current one. Remove excess parentheses.
14597 Restriction rejectinon of vect_constant_def and vect_external_def
14598 for variable-length vectors to boolean types, or types for which
14599 can_duplicate_and_interleave_p is false.
14600 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14601 (duplicate_and_interleave): New function.
14602 (vect_get_constant_vectors): Use gimple_build_vector for
14603 constant-length vectors and suitable variable-length constant
14604 vectors. Use duplicate_and_interleave for other variable-length
14605 vectors. Don't defer the update when inserting new statements.
14607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14608 Alan Hayward <alan.hayward@arm.com>
14609 David Sherwood <david.sherwood@arm.com>
14611 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14612 min_profitable_iters doesn't go negative.
14614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14615 Alan Hayward <alan.hayward@arm.com>
14616 David Sherwood <david.sherwood@arm.com>
14618 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14619 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14620 * optabs.def (vec_mask_load_lanes_optab): New optab.
14621 (vec_mask_store_lanes_optab): Likewise.
14622 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14623 (MASK_STORE_LANES): Likewise.
14624 * internal-fn.c (mask_load_lanes_direct): New macro.
14625 (mask_store_lanes_direct): Likewise.
14626 (expand_mask_load_optab_fn): Handle masked operations.
14627 (expand_mask_load_lanes_optab_fn): New macro.
14628 (expand_mask_store_optab_fn): Handle masked operations.
14629 (expand_mask_store_lanes_optab_fn): New macro.
14630 (direct_mask_load_lanes_optab_supported_p): Likewise.
14631 (direct_mask_store_lanes_optab_supported_p): Likewise.
14632 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14634 (vect_load_lanes_supported): Likewise.
14635 * tree-vect-data-refs.c (strip_conversion): New function.
14636 (can_group_stmts_p): Likewise.
14637 (vect_analyze_data_ref_accesses): Use it instead of checking
14638 for a pair of assignments.
14639 (vect_store_lanes_supported): Take a masked_p parameter.
14640 (vect_load_lanes_supported): Likewise.
14641 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14642 vect_store_lanes_supported and vect_load_lanes_supported.
14643 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14644 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14645 parameter. Don't allow gaps for masked accesses.
14646 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14647 and vect_load_lanes_supported.
14648 (get_load_store_type): Take a masked_p parameter and update
14649 call to get_group_load_store_type.
14650 (vectorizable_store): Update call to get_load_store_type.
14651 Handle IFN_MASK_STORE_LANES.
14652 (vectorizable_load): Update call to get_load_store_type.
14653 Handle IFN_MASK_LOAD_LANES.
14655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14656 Alan Hayward <alan.hayward@arm.com>
14657 David Sherwood <david.sherwood@arm.com>
14659 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14661 * config/aarch64/aarch64-protos.h
14662 (aarch64_sve_struct_memory_operand_p): Declare.
14663 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14664 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14665 (VPRED, vpred): Handle SVE structure modes.
14666 * config/aarch64/constraints.md (Utx): New constraint.
14667 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14668 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14669 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14670 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14671 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14672 structure modes. Split into pieces after RA.
14673 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14674 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14676 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14677 SVE structure modes.
14678 (aarch64_classify_address): Likewise.
14679 (sizetochar): Move earlier in file.
14680 (aarch64_print_operand): Handle SVE register lists.
14681 (aarch64_array_mode): New function.
14682 (aarch64_sve_struct_memory_operand_p): Likewise.
14683 (TARGET_ARRAY_MODE): Redefine.
14685 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14686 Alan Hayward <alan.hayward@arm.com>
14687 David Sherwood <david.sherwood@arm.com>
14689 * target.def (array_mode): New target hook.
14690 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14691 * doc/tm.texi: Regenerate.
14692 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14693 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14694 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14695 targetm.array_mode.
14696 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14700 Alan Hayward <alan.hayward@arm.com>
14701 David Sherwood <david.sherwood@arm.com>
14703 * fold-const.c (fold_binary_loc): Check the argument types
14704 rather than the result type when testing for a vector operation.
14706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14708 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14709 * doc/tm.texi: Regenerate.
14711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14712 Alan Hayward <alan.hayward@arm.com>
14713 David Sherwood <david.sherwood@arm.com>
14715 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14716 (sve): Document new AArch64 extension.
14717 * doc/md.texi (w): Extend the description of the AArch64
14718 constraint to include SVE vectors.
14719 (Upl, Upa): Document new AArch64 predicate constraints.
14720 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14722 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14723 (msve-vector-bits=): New option.
14724 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14725 SVE when these are disabled.
14726 (sve): New extension.
14727 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14728 modes. Adjust their number of units based on aarch64_sve_vg.
14729 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14730 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14731 aarch64_addr_query_type.
14732 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14733 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14734 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14735 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14736 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14737 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14738 (aarch64_simd_imm_zero_p): Delete.
14739 (aarch64_check_zero_based_sve_index_immediate): Declare.
14740 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14741 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14742 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14743 (aarch64_sve_float_mul_immediate_p): Likewise.
14744 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14745 rather than an rtx.
14746 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14747 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14748 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14749 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14750 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14751 (aarch64_regmode_natural_size): Likewise.
14752 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14753 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14755 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14756 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14757 for VG and the SVE predicate registers.
14758 (V_ALIASES): Add a "z"-prefixed alias.
14759 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14760 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14761 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14762 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14763 (REG_CLASS_NAMES): Add entries for them.
14764 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14765 and the predicate registers.
14766 (aarch64_sve_vg): Declare.
14767 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14768 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14769 (REGMODE_NATURAL_SIZE): Define.
14770 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14772 * config/aarch64/aarch64.c: Include cfgrtl.h.
14773 (simd_immediate_info): Add a constructor for series vectors,
14774 and an associated step field.
14775 (aarch64_sve_vg): New variable.
14776 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14777 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14778 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14779 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14780 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14781 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14782 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14783 (aarch64_get_mask_mode): New functions.
14784 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14785 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14786 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14787 predicate modes and predicate registers. Explicitly restrict
14788 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14789 to store a vector mode if it is recognized by
14790 aarch64_classify_vector_mode.
14791 (aarch64_regmode_natural_size): New function.
14792 (aarch64_hard_regno_caller_save_mode): Return the original mode
14794 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14795 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14796 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14797 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14799 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14800 does not overlap dest if the function is frame-related. Handle
14802 (aarch64_split_add_offset): New function.
14803 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14804 them aarch64_add_offset.
14805 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14806 and update call to aarch64_sub_sp.
14807 (aarch64_add_cfa_expression): New function.
14808 (aarch64_expand_prologue): Pass extra temporary registers to the
14809 functions above. Handle the case in which we need to emit new
14810 DW_CFA_expressions for registers that were originally saved
14811 relative to the stack pointer, but now have to be expressed
14812 relative to the frame pointer.
14813 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14815 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14816 IP0 and IP1 values for SVE frames.
14817 (aarch64_expand_vec_series): New function.
14818 (aarch64_expand_sve_widened_duplicate): Likewise.
14819 (aarch64_expand_sve_const_vector): Likewise.
14820 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14821 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14822 into the register, rather than emitting a SET directly.
14823 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14824 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14825 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14826 (offset_9bit_signed_scaled_p): New functions.
14827 (aarch64_replicate_bitmask_imm): New function.
14828 (aarch64_bitmask_imm): Use it.
14829 (aarch64_cannot_force_const_mem): Reject expressions involving
14830 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14831 (aarch64_classify_index): Handle SVE indices, by requiring
14832 a plain register index with a scale that matches the element size.
14833 (aarch64_classify_address): Handle SVE addresses. Assert that
14834 the mode of the address is VOIDmode or an integer mode.
14835 Update call to aarch64_classify_symbol.
14836 (aarch64_classify_symbolic_expression): Update call to
14837 aarch64_classify_symbol.
14838 (aarch64_const_vec_all_in_range_p): New function.
14839 (aarch64_print_vector_float_operand): Likewise.
14840 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14841 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14842 and the FP immediates 1.0 and 0.5.
14843 (aarch64_print_address_internal): Handle SVE addresses.
14844 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14845 (aarch64_regno_regclass): Handle predicate registers.
14846 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14848 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14849 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14850 (aarch64_convert_sve_vector_bits): New function.
14851 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14852 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14853 rather than an rtx.
14854 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14855 Handle SVE vector and predicate modes. Accept VL-based constants
14856 that need only one temporary register, and VL offsets that require
14857 no temporary registers.
14858 (aarch64_conditional_register_usage): Mark the predicate registers
14859 as fixed if SVE isn't available.
14860 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14861 Return true for SVE vector and predicate modes.
14862 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14863 rather than an unsigned int. Handle SVE modes.
14864 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14866 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14868 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14869 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14870 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14871 (aarch64_sve_float_mul_immediate_p): New functions.
14872 (aarch64_sve_valid_immediate): New function.
14873 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14874 Explicitly reject structure modes. Check for INDEX constants.
14875 Handle PTRUE and PFALSE constants.
14876 (aarch64_check_zero_based_sve_index_immediate): New function.
14877 (aarch64_simd_imm_zero_p): Delete.
14878 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14879 vector modes. Accept constants in the range of CNT[BHWD].
14880 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14881 ask for an Advanced SIMD mode.
14882 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14883 (aarch64_simd_vector_alignment): Handle SVE predicates.
14884 (aarch64_vectorize_preferred_vector_alignment): New function.
14885 (aarch64_simd_vector_alignment_reachable): Use it instead of
14887 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14888 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14890 (MAX_VECT_LEN): Delete.
14891 (expand_vec_perm_d): Add a vec_flags field.
14892 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14893 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14894 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14896 (aarch64_evpc_rev): Rename to...
14897 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14898 (aarch64_evpc_rev_global): New function.
14899 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14900 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14902 (aarch64_evpc_sve_tbl): New function.
14903 (aarch64_expand_vec_perm_const_1): Update after rename of
14904 aarch64_evpc_rev. Handle SVE permutes too, trying
14905 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14906 than aarch64_evpc_tbl.
14907 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14908 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14909 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14910 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14911 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14912 (aarch64_expand_sve_vcond): New functions.
14913 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14914 of aarch64_vector_mode_p.
14915 (aarch64_dwarf_poly_indeterminate_value): New function.
14916 (aarch64_compute_pressure_classes): Likewise.
14917 (aarch64_can_change_mode_class): Likewise.
14918 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14919 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14920 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14921 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14922 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14923 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14924 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14925 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14927 (Dn, Dl, Dr): Accept const as well as const_vector.
14928 (Dz): Likewise. Compare against CONST0_RTX.
14929 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14930 of "vector" where appropriate.
14931 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14932 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14933 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14934 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14935 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14936 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14937 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14938 (v_int_equiv): Extend to SVE modes.
14939 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14941 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14942 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14943 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14944 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14945 (SVE_COND_FP_CMP): New int iterators.
14946 (perm_hilo): Handle the new unpack unspecs.
14947 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14949 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14950 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14951 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14952 (aarch64_equality_operator, aarch64_constant_vector_operand)
14953 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14954 (aarch64_sve_nonimmediate_operand): Likewise.
14955 (aarch64_sve_general_operand): Likewise.
14956 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14957 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14958 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14959 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14960 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14961 (aarch64_sve_float_arith_immediate): Likewise.
14962 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14963 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14964 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14965 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14966 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14967 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14968 (aarch64_sve_float_arith_operand): Likewise.
14969 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14970 (aarch64_sve_float_mul_operand): Likewise.
14971 (aarch64_sve_vec_perm_operand): Likewise.
14972 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14973 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14974 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14975 as well as const_vector.
14976 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14977 in file. Use CONST0_RTX and CONSTM1_RTX.
14978 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14979 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14980 Use aarch64_simd_imm_zero.
14981 * config/aarch64/aarch64-sve.md: New file.
14982 * config/aarch64/aarch64.md: Include it.
14983 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14984 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14985 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14986 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14987 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14988 (sve): New attribute.
14989 (enabled): Disable instructions with the sve attribute unless
14991 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14992 aarch64_expand_mov_immediate.
14993 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14994 CNT[BHSD] immediates.
14995 (movti): Split CONST_POLY_INT moves into two halves.
14996 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14997 Split additions that need a temporary here if the destination
14998 is the stack pointer.
14999 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15000 (*add<mode>3_poly_1): New instruction.
15001 (set_clobber_cc): New expander.
15003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15005 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15006 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15007 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15008 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15009 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15010 Change innermode from fixed_mode_size to machine_mode.
15011 (simplify_subreg): Update call accordingly. Handle a constant-sized
15012 subreg of a variable-length CONST_VECTOR.
15014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15015 Alan Hayward <alan.hayward@arm.com>
15016 David Sherwood <david.sherwood@arm.com>
15018 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15019 (add_offset_to_base): New function, split out from...
15020 (create_mem_ref): ...here. When handling a scale other than 1,
15021 check first whether the address is valid without the offset.
15022 Add it into the base if so, leaving the index and scale as-is.
15024 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15027 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15028 fold_for_warn before checking if arg2 is INTEGER_CST.
15030 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15032 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15033 (store_multiple_operation): Delete.
15034 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15035 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15036 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15037 guarded by TARGET_STRING.
15038 (rs6000_output_load_multiple): Delete.
15039 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15040 OPTION_MASK_STRING / TARGET_STRING handling.
15041 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15042 (const rs6000_opt_masks) <"string">: Change mask to 0.
15043 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15044 (MASK_STRING): Delete.
15045 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15047 (load_multiple): Delete.
15054 (store_multiple): Delete.
15061 (movmemsi_8reg): Delete.
15062 (corresponding unnamed define_insn): Delete.
15063 (movmemsi_6reg): Delete.
15064 (corresponding unnamed define_insn): Delete.
15065 (movmemsi_4reg): Delete.
15066 (corresponding unnamed define_insn): Delete.
15067 (movmemsi_2reg): Delete.
15068 (corresponding unnamed define_insn): Delete.
15069 (movmemsi_1reg): Delete.
15070 (corresponding unnamed define_insn): Delete.
15071 * config/rs6000/rs6000.opt (mno-string): New.
15072 (mstring): Replace by deprecation warning stub.
15073 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15075 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15077 * regrename.c (regrename_do_replace): If replacing the same
15078 reg multiple times, try to reuse last created gen_raw_REG.
15081 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15082 main to workaround a bug in GDB.
15084 2018-01-12 Tom de Vries <tom@codesourcery.com>
15087 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15089 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15091 PR rtl-optimization/80481
15092 * ira-color.c (get_cap_member): New function.
15093 (allocnos_conflict_by_live_ranges_p): Use it.
15094 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15095 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15097 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15100 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15101 (*saddl_se_1): Ditto.
15102 (*ssubsi_1): Ditto.
15103 (*ssubl_se_1): Ditto.
15105 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15107 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15108 rather than wi::to_widest for DR_INITs.
15109 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15110 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15111 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15113 (vect_analyze_group_access_1): Note that here.
15115 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15117 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15118 polynomial type sizes.
15120 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15122 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15123 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15124 (gimple_add_tmp_var): Likewise.
15126 2018-01-12 Martin Liska <mliska@suse.cz>
15128 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15129 (gimple_alloc_sizes): Likewise.
15130 (dump_gimple_statistics): Use PRIu64 in printf format.
15131 * gimple.h: Change uint64_t to int.
15133 2018-01-12 Martin Liska <mliska@suse.cz>
15135 * tree-core.h: Use uint64_t instead of int.
15136 * tree.c (tree_node_counts): Likewise.
15137 (tree_node_sizes): Likewise.
15138 (dump_tree_statistics): Use PRIu64 in printf format.
15140 2018-01-12 Martin Liska <mliska@suse.cz>
15142 * Makefile.in: As qsort_chk is implemented in vec.c, add
15143 vec.o to linkage of gencfn-macros.
15144 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15145 passing the info to record_node_allocation_statistics.
15146 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15148 * ggc-common.c (struct ggc_usage): Add operator== and use
15149 it in operator< and compare function.
15150 * mem-stats.h (struct mem_usage): Likewise.
15151 * vec.c (struct vec_usage): Remove operator< and compare
15152 function. Can be simply inherited.
15154 2018-01-12 Martin Jambor <mjambor@suse.cz>
15157 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15158 * tree-ssa-math-opts.c: Include domwalk.h.
15159 (convert_mult_to_fma_1): New function.
15160 (fma_transformation_info): New type.
15161 (fma_deferring_state): Likewise.
15162 (cancel_fma_deferring): New function.
15163 (result_of_phi): Likewise.
15164 (last_fma_candidate_feeds_initial_phi): Likewise.
15165 (convert_mult_to_fma): Added deferring logic, split actual
15166 transformation to convert_mult_to_fma_1.
15167 (math_opts_dom_walker): New type.
15168 (math_opts_dom_walker::after_dom_children): New method, body moved
15169 here from pass_optimize_widening_mul::execute, added deferring logic
15171 (pass_optimize_widening_mul::execute): Moved most of code to
15172 math_opts_dom_walker::after_dom_children.
15173 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15174 * config/i386/i386.c (ix86_option_override_internal): Added
15175 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15177 2018-01-12 Richard Biener <rguenther@suse.de>
15180 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15181 inline instance vars.
15183 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15186 * config/rx/rx.c (rx_is_restricted_memory_address):
15187 Handle SUBREG case.
15189 2018-01-12 Richard Biener <rguenther@suse.de>
15191 PR tree-optimization/80846
15192 * target.def (split_reduction): New target hook.
15193 * targhooks.c (default_split_reduction): New function.
15194 * targhooks.h (default_split_reduction): Declare.
15195 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15196 target requests first reduce vectors by combining low and high
15198 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15199 (get_vectype_for_scalar_type_and_size): Export.
15200 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15201 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15202 * doc/tm.texi: Regenerate.
15203 * config/i386/i386.c (ix86_split_reduction): Implement
15204 TARGET_VECTORIZE_SPLIT_REDUCTION.
15206 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15209 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15210 in PIC mode except for TARGET_VXWORKS_RTP.
15211 * config/sparc/sparc.c: Include cfgrtl.h.
15212 (TARGET_INIT_PIC_REG): Define.
15213 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15214 (sparc_pic_register_p): New predicate.
15215 (sparc_legitimate_address_p): Use it.
15216 (sparc_legitimize_pic_address): Likewise.
15217 (sparc_delegitimize_address): Likewise.
15218 (sparc_mode_dependent_address_p): Likewise.
15219 (gen_load_pcrel_sym): Remove 4th parameter.
15220 (load_got_register): Adjust call to above. Remove obsolete stuff.
15221 (sparc_expand_prologue): Do not call load_got_register here.
15222 (sparc_flat_expand_prologue): Likewise.
15223 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15224 (sparc_use_pseudo_pic_reg): New function.
15225 (sparc_init_pic_reg): Likewise.
15226 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15227 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15229 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15231 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15232 Add item for branch_cost.
15234 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15236 PR rtl-optimization/83565
15237 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15238 not extend the result to a larger mode for rotate operations.
15239 (num_sign_bit_copies1): Likewise.
15241 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15244 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15246 Use values-Xc.o for -pedantic.
15247 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15249 2018-01-12 Martin Liska <mliska@suse.cz>
15252 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15254 (possible_polymorphic_call_targets): Use it.
15255 (ipa_devirt): Likewise.
15257 2018-01-12 Martin Liska <mliska@suse.cz>
15259 * profile-count.h (enum profile_quality): Use 0 as invalid
15260 enum value of profile_quality.
15262 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15264 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15265 -mext-string options.
15267 2018-01-12 Richard Biener <rguenther@suse.de>
15269 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15270 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15271 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15273 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15275 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15277 * configure.ac (--with-long-double-format): Add support for the
15278 configuration option to change the default long double format on
15280 * config.gcc (powerpc*-linux*-*): Likewise.
15281 * configure: Regenerate.
15282 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15283 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15284 used without modification.
15286 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15288 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15289 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15290 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15291 MISC_BUILTIN_SPEC_BARRIER.
15292 (rs6000_init_builtins): Likewise.
15293 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15295 (speculation_barrier): New define_insn.
15296 * doc/extend.texi: Document __builtin_speculation_barrier.
15298 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15301 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15302 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15303 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15305 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15306 integral modes instead of "ss" and "sd".
15307 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15308 vectors with 32-bit and 64-bit elements.
15309 (vecdupssescalarmodesuffix): New mode attribute.
15310 (vec_dup<mode>): Use it.
15312 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15315 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15316 frame if argument is passed on stack.
15318 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15321 * ree.c (combine_reaching_defs): Optimize also
15322 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15323 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15325 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15327 PR middle-end/83189
15328 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15330 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15332 PR middle-end/83718
15333 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15334 after they are computed.
15336 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15338 PR tree-optimization/83695
15339 * gimple-loop-linterchange.cc
15340 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15341 reset cached scev information after interchange.
15342 (pass_linterchange::execute): Remove call to scev_reset_htab.
15344 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15346 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15347 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15348 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15349 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15350 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15351 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15352 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15353 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15354 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15355 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15356 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15357 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15358 (V_lane_reg): Likewise.
15359 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15361 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15362 (vfmal_lane_low<mode>_intrinsic,
15363 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15364 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15365 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15366 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15367 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15368 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15370 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15372 * config/arm/arm-cpus.in (fp16fml): New feature.
15373 (ALL_SIMD): Add fp16fml.
15374 (armv8.2-a): Add fp16fml as an option.
15375 (armv8.3-a): Likewise.
15376 (armv8.4-a): Add fp16fml as part of fp16.
15377 * config/arm/arm.h (TARGET_FP16FML): Define.
15378 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15380 * config/arm/arm-modes.def (V2HF): Define.
15381 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15382 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15383 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15384 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15385 vfmsl_low, vfmsl_high): New set of builtins.
15386 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15387 (vfml_op): New code attribute.
15388 (VFMLHALVES): New int iterator.
15389 (VFML, VFMLSEL): New mode attributes.
15390 (V_reg): Define mapping for V2HF.
15391 (V_hi, V_lo): New mode attributes.
15392 (VF_constraint): Likewise.
15393 (vfml_half, vfml_half_selector): New int attributes.
15394 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15396 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15397 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15399 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15400 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15401 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15402 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15404 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15405 Document new effective target and option set.
15407 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15409 * config/arm/arm-cpus.in (armv8_4): New feature.
15410 (ARMv8_4a): New fgroup.
15411 (armv8.4-a): New arch.
15412 * config/arm/arm-tables.opt: Regenerate.
15413 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15414 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15415 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15416 Add matching rules for -march=armv8.4-a and extensions.
15417 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15419 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15422 * config/rx/rx.md (BW): New mode attribute.
15423 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15425 2018-01-11 Richard Biener <rguenther@suse.de>
15427 PR tree-optimization/83435
15428 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15429 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15430 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15432 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15433 Alan Hayward <alan.hayward@arm.com>
15434 David Sherwood <david.sherwood@arm.com>
15436 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15438 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15439 (aarch64_print_address_internal): Use it to check for a zero offset.
15441 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15442 Alan Hayward <alan.hayward@arm.com>
15443 David Sherwood <david.sherwood@arm.com>
15445 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15446 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15447 Return a poly_int64 rather than a HOST_WIDE_INT.
15448 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15449 rather than a HOST_WIDE_INT.
15450 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15451 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15452 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15453 final_offset from HOST_WIDE_INT to poly_int64.
15454 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15455 to_constant when getting the number of units in an Advanced SIMD
15457 (aarch64_builtin_vectorized_function): Check for a constant number
15459 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15461 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15462 attribute instead of GET_MODE_NUNITS.
15463 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15464 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15465 GET_MODE_SIZE for fixed-size registers.
15466 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15467 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15468 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15469 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15470 (aarch64_print_operand, aarch64_print_address_internal)
15471 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15472 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15473 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15474 Handle polynomial GET_MODE_SIZE.
15475 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15476 wider than SImode without modification.
15477 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15478 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15479 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15480 passing and returning SVE modes.
15481 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15482 rather than GEN_INT.
15483 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15484 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15485 (aarch64_allocate_and_probe_stack_space): Likewise.
15486 (aarch64_layout_frame): Cope with polynomial offsets.
15487 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15488 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15489 polynomial offsets.
15490 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15491 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15492 poly_int64 rather than a HOST_WIDE_INT.
15493 (aarch64_get_separate_components, aarch64_process_components)
15494 (aarch64_expand_prologue, aarch64_expand_epilogue)
15495 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15496 (aarch64_anchor_offset): New function, split out from...
15497 (aarch64_legitimize_address): ...here.
15498 (aarch64_builtin_vectorization_cost): Handle polynomial
15499 TYPE_VECTOR_SUBPARTS.
15500 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15502 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15503 number of elements from the PARALLEL rather than the mode.
15504 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15505 rather than GET_MODE_BITSIZE.
15506 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15507 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15508 (aarch64_expand_vec_perm_const_1): Handle polynomial
15509 d->perm.length () and d->perm elements.
15510 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15511 Apply to_constant to d->perm elements.
15512 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15513 polynomial CONST_VECTOR_NUNITS.
15514 (aarch64_move_pointer): Take amount as a poly_int64 rather
15516 (aarch64_progress_pointer): Avoid temporary variable.
15517 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15518 the mode attribute instead of GET_MODE.
15520 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15521 Alan Hayward <alan.hayward@arm.com>
15522 David Sherwood <david.sherwood@arm.com>
15524 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15525 x exists before using it.
15526 (aarch64_add_constant_internal): Rename to...
15527 (aarch64_add_offset_1): ...this. Replace regnum with separate
15528 src and dest rtxes. Handle the case in which they're different,
15529 including when the offset is zero. Replace scratchreg with an rtx.
15530 Use 2 additions if there is no spare register into which we can
15531 move a 16-bit constant.
15532 (aarch64_add_constant): Delete.
15533 (aarch64_add_offset): Replace reg with separate src and dest
15534 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15535 Use aarch64_add_offset_1.
15536 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15537 an rtx rather than an int. Take the delta as a poly_int64
15538 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15539 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15540 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15541 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15542 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15543 and aarch64_add_sp.
15544 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15545 aarch64_add_constant.
15547 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15549 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15550 Use scalar_float_mode.
15552 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15554 * config/aarch64/aarch64-simd.md
15555 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15556 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15557 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15558 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15559 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15560 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15561 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15562 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15563 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15564 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15566 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15569 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15570 targ_options->x_arm_arch_string is non NULL.
15572 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15574 * config/aarch64/aarch64.h
15575 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15577 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15580 * expmed.c (emit_store_flag_force): Swap if const op0
15581 and change VOIDmode to mode of op0.
15583 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15585 PR rtl-optimization/83761
15586 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15587 than bytes to mode_for_size.
15589 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15591 PR middle-end/83189
15592 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15593 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15596 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15598 PR middle-end/83575
15599 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15600 when in layout mode.
15601 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15602 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15605 2018-01-10 Michael Collison <michael.collison@arm.com>
15607 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15608 * config/aarch64/aarch64-option-extension.def: Add
15609 AARCH64_OPT_EXTENSION of 'fp16fml'.
15610 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15611 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15612 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15613 * config/aarch64/constraints.md (Ui7): New constraint.
15614 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15615 (VFMLA_SEL_W): Ditto.
15618 (VFMLA16_LOW): New int iterator.
15619 (VFMLA16_HIGH): Ditto.
15620 (UNSPEC_FMLAL): New unspec.
15621 (UNSPEC_FMLSL): Ditto.
15622 (UNSPEC_FMLAL2): Ditto.
15623 (UNSPEC_FMLSL2): Ditto.
15624 (f16mac): New code attribute.
15625 * config/aarch64/aarch64-simd-builtins.def
15626 (aarch64_fmlal_lowv2sf): Ditto.
15627 (aarch64_fmlsl_lowv2sf): Ditto.
15628 (aarch64_fmlalq_lowv4sf): Ditto.
15629 (aarch64_fmlslq_lowv4sf): Ditto.
15630 (aarch64_fmlal_highv2sf): Ditto.
15631 (aarch64_fmlsl_highv2sf): Ditto.
15632 (aarch64_fmlalq_highv4sf): Ditto.
15633 (aarch64_fmlslq_highv4sf): Ditto.
15634 (aarch64_fmlal_lane_lowv2sf): Ditto.
15635 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15636 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15637 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15638 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15639 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15640 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15641 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15642 (aarch64_fmlal_lane_highv2sf): Ditto.
15643 (aarch64_fmlsl_lane_highv2sf): Ditto.
15644 (aarch64_fmlal_laneq_highv2sf): Ditto.
15645 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15646 (aarch64_fmlalq_lane_highv4sf): Ditto.
15647 (aarch64_fmlsl_lane_highv4sf): Ditto.
15648 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15649 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15650 * config/aarch64/aarch64-simd.md:
15651 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15652 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15653 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15654 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15655 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15656 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15657 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15658 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15659 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15660 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15661 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15662 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15663 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15664 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15665 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15666 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15667 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15668 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15669 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15670 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15671 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15672 (vfmlsl_low_u32): Ditto.
15673 (vfmlalq_low_u32): Ditto.
15674 (vfmlslq_low_u32): Ditto.
15675 (vfmlal_high_u32): Ditto.
15676 (vfmlsl_high_u32): Ditto.
15677 (vfmlalq_high_u32): Ditto.
15678 (vfmlslq_high_u32): Ditto.
15679 (vfmlal_lane_low_u32): Ditto.
15680 (vfmlsl_lane_low_u32): Ditto.
15681 (vfmlal_laneq_low_u32): Ditto.
15682 (vfmlsl_laneq_low_u32): Ditto.
15683 (vfmlalq_lane_low_u32): Ditto.
15684 (vfmlslq_lane_low_u32): Ditto.
15685 (vfmlalq_laneq_low_u32): Ditto.
15686 (vfmlslq_laneq_low_u32): Ditto.
15687 (vfmlal_lane_high_u32): Ditto.
15688 (vfmlsl_lane_high_u32): Ditto.
15689 (vfmlal_laneq_high_u32): Ditto.
15690 (vfmlsl_laneq_high_u32): Ditto.
15691 (vfmlalq_lane_high_u32): Ditto.
15692 (vfmlslq_lane_high_u32): Ditto.
15693 (vfmlalq_laneq_high_u32): Ditto.
15694 (vfmlslq_laneq_high_u32): Ditto.
15695 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15696 (AARCH64_FL_FOR_ARCH8_4): New.
15697 (AARCH64_ISA_F16FML): New ISA flag.
15698 (TARGET_F16FML): New feature flag for fp16fml.
15699 (doc/invoke.texi): Document new fp16fml option.
15701 2018-01-10 Michael Collison <michael.collison@arm.com>
15703 * config/aarch64/aarch64-builtins.c:
15704 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15705 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15706 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15707 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15708 (AARCH64_ISA_SHA3): New ISA flag.
15709 (TARGET_SHA3): New feature flag for sha3.
15710 * config/aarch64/iterators.md (sha512_op): New int attribute.
15711 (CRYPTO_SHA512): New int iterator.
15712 (UNSPEC_SHA512H): New unspec.
15713 (UNSPEC_SHA512H2): Ditto.
15714 (UNSPEC_SHA512SU0): Ditto.
15715 (UNSPEC_SHA512SU1): Ditto.
15716 * config/aarch64/aarch64-simd-builtins.def
15717 (aarch64_crypto_sha512hqv2di): New builtin.
15718 (aarch64_crypto_sha512h2qv2di): Ditto.
15719 (aarch64_crypto_sha512su0qv2di): Ditto.
15720 (aarch64_crypto_sha512su1qv2di): Ditto.
15721 (aarch64_eor3qv8hi): Ditto.
15722 (aarch64_rax1qv2di): Ditto.
15723 (aarch64_xarqv2di): Ditto.
15724 (aarch64_bcaxqv8hi): Ditto.
15725 * config/aarch64/aarch64-simd.md:
15726 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15727 (aarch64_crypto_sha512su0qv2di): Ditto.
15728 (aarch64_crypto_sha512su1qv2di): Ditto.
15729 (aarch64_eor3qv8hi): Ditto.
15730 (aarch64_rax1qv2di): Ditto.
15731 (aarch64_xarqv2di): Ditto.
15732 (aarch64_bcaxqv8hi): Ditto.
15733 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15734 (vsha512h2q_u64): Ditto.
15735 (vsha512su0q_u64): Ditto.
15736 (vsha512su1q_u64): Ditto.
15737 (veor3q_u16): Ditto.
15738 (vrax1q_u64): Ditto.
15739 (vxarq_u64): Ditto.
15740 (vbcaxq_u16): Ditto.
15741 * config/arm/types.md (crypto_sha512): New type attribute.
15742 (crypto_sha3): Ditto.
15743 (doc/invoke.texi): Document new sha3 option.
15745 2018-01-10 Michael Collison <michael.collison@arm.com>
15747 * config/aarch64/aarch64-builtins.c:
15748 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15749 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15750 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15751 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15752 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15753 (AARCH64_ISA_SM4): New ISA flag.
15754 (TARGET_SM4): New feature flag for sm4.
15755 * config/aarch64/aarch64-simd-builtins.def
15756 (aarch64_sm3ss1qv4si): Ditto.
15757 (aarch64_sm3tt1aq4si): Ditto.
15758 (aarch64_sm3tt1bq4si): Ditto.
15759 (aarch64_sm3tt2aq4si): Ditto.
15760 (aarch64_sm3tt2bq4si): Ditto.
15761 (aarch64_sm3partw1qv4si): Ditto.
15762 (aarch64_sm3partw2qv4si): Ditto.
15763 (aarch64_sm4eqv4si): Ditto.
15764 (aarch64_sm4ekeyqv4si): Ditto.
15765 * config/aarch64/aarch64-simd.md:
15766 (aarch64_sm3ss1qv4si): Ditto.
15767 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15768 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15769 (aarch64_sm4eqv4si): Ditto.
15770 (aarch64_sm4ekeyqv4si): Ditto.
15771 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15772 (sm3part_op): Ditto.
15773 (CRYPTO_SM3TT): Ditto.
15774 (CRYPTO_SM3PART): Ditto.
15775 (UNSPEC_SM3SS1): New unspec.
15776 (UNSPEC_SM3TT1A): Ditto.
15777 (UNSPEC_SM3TT1B): Ditto.
15778 (UNSPEC_SM3TT2A): Ditto.
15779 (UNSPEC_SM3TT2B): Ditto.
15780 (UNSPEC_SM3PARTW1): Ditto.
15781 (UNSPEC_SM3PARTW2): Ditto.
15782 (UNSPEC_SM4E): Ditto.
15783 (UNSPEC_SM4EKEY): Ditto.
15784 * config/aarch64/constraints.md (Ui2): New constraint.
15785 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15786 * config/arm/types.md (crypto_sm3): New type attribute.
15787 (crypto_sm4): Ditto.
15788 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15789 (vsm3tt1aq_u32): Ditto.
15790 (vsm3tt1bq_u32): Ditto.
15791 (vsm3tt2aq_u32): Ditto.
15792 (vsm3tt2bq_u32): Ditto.
15793 (vsm3partw1q_u32): Ditto.
15794 (vsm3partw2q_u32): Ditto.
15795 (vsm4eq_u32): Ditto.
15796 (vsm4ekeyq_u32): Ditto.
15797 (doc/invoke.texi): Document new sm4 option.
15799 2018-01-10 Michael Collison <michael.collison@arm.com>
15801 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15802 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15803 (AARCH64_FL_FOR_ARCH8_4): New.
15804 (AARCH64_FL_V8_4): New flag.
15805 (doc/invoke.texi): Document new armv8.4-a option.
15807 2018-01-10 Michael Collison <michael.collison@arm.com>
15809 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15810 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15811 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15812 * config/aarch64/aarch64-option-extension.def: Add
15813 AARCH64_OPT_EXTENSION of 'sha2'.
15814 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15815 (crypto): Disable sha2 and aes if crypto disabled.
15816 (crypto): Enable aes and sha2 if enabled.
15817 (simd): Disable sha2 and aes if simd disabled.
15818 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15820 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15821 (TARGET_SHA2): New feature flag for sha2.
15822 (TARGET_AES): New feature flag for aes.
15823 * config/aarch64/aarch64-simd.md:
15824 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15825 conditional on TARGET_AES.
15826 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15827 (aarch64_crypto_sha1hsi): Make pattern conditional
15829 (aarch64_crypto_sha1hv4si): Ditto.
15830 (aarch64_be_crypto_sha1hv4si): Ditto.
15831 (aarch64_crypto_sha1su1v4si): Ditto.
15832 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15833 (aarch64_crypto_sha1su0v4si): Ditto.
15834 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15835 (aarch64_crypto_sha256su0v4si): Ditto.
15836 (aarch64_crypto_sha256su1v4si): Ditto.
15837 (doc/invoke.texi): Document new aes and sha2 options.
15839 2018-01-10 Martin Sebor <msebor@redhat.com>
15841 PR tree-optimization/83781
15842 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15845 2018-01-11 Martin Sebor <msebor@gmail.com>
15846 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15848 PR tree-optimization/83501
15849 PR tree-optimization/81703
15851 * tree-ssa-strlen.c (get_string_cst): Rename...
15852 (get_string_len): ...to this. Handle global constants.
15853 (handle_char_store): Adjust.
15855 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15856 Jim Wilson <jimw@sifive.com>
15858 * config/riscv/riscv-protos.h (riscv_output_return): New.
15859 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15860 (riscv_attribute_table, riscv_output_return),
15861 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15862 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15863 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15865 (riscv_expand_prologue): Add early return for naked function.
15866 (riscv_expand_epilogue): Likewise.
15867 (riscv_function_ok_for_sibcall): Return false for naked function.
15868 (riscv_set_current_function): New.
15869 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15870 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15871 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15872 * doc/extend.texi (RISC-V Function Attributes): New.
15874 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15876 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15877 check for 128-bit long double before checking TCmode.
15878 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15879 128-bit long doubles before checking TFmode or TCmode.
15880 (FLOAT128_IBM_P): Likewise.
15882 2018-01-10 Martin Sebor <msebor@redhat.com>
15884 PR tree-optimization/83671
15885 * builtins.c (c_strlen): Unconditionally return zero for the empty
15887 Use -Warray-bounds for warnings.
15888 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15889 for non-constant array indices with COMPONENT_REF, arrays of
15890 arrays, and pointers to arrays.
15891 (gimple_fold_builtin_strlen): Determine and set length range for
15892 non-constant character arrays.
15894 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15896 PR middle-end/81897
15897 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15900 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15902 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15904 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15907 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15908 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15909 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15910 indexed_or_indirect_operand predicate.
15911 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15912 (*vsx_le_perm_load_v8hi): Likewise.
15913 (*vsx_le_perm_load_v16qi): Likewise.
15914 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15915 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15916 (*vsx_le_perm_store_v8hi): Likewise.
15917 (*vsx_le_perm_store_v16qi): Likewise.
15918 (eight unnamed splitters): Likewise.
15920 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15922 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15923 * config/rs6000/emmintrin.h: Likewise.
15924 * config/rs6000/mmintrin.h: Likewise.
15925 * config/rs6000/xmmintrin.h: Likewise.
15927 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15930 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15932 * tree.c (tree_nop_conversion): Return true for location wrapper
15934 (maybe_wrap_with_location): New function.
15935 (selftest::check_strip_nops): New function.
15936 (selftest::test_location_wrappers): New function.
15937 (selftest::tree_c_tests): Call it.
15938 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15939 (maybe_wrap_with_location): New decl.
15940 (EXPR_LOCATION_WRAPPER_P): New macro.
15941 (location_wrapper_p): New inline function.
15942 (tree_strip_any_location_wrapper): New inline function.
15944 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15947 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15948 stack_realign_offset for the largest alignment of stack slot
15950 (ix86_find_max_used_stack_alignment): New function.
15951 (ix86_finalize_stack_frame_flags): Use it. Set
15952 max_used_stack_alignment if we don't realign stack.
15953 * config/i386/i386.h (machine_function): Add
15954 max_used_stack_alignment.
15956 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15958 * config/arm/arm.opt (-mbranch-cost): New option.
15959 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15962 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15965 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15966 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15968 2018-01-10 Richard Biener <rguenther@suse.de>
15971 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15972 early out so it also covers the case where we have a non-NULL
15975 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15977 PR tree-optimization/83753
15978 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15979 for non-strided grouped accesses if the number of elements is 1.
15981 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15984 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15985 * i386.h (TARGET_USE_GATHER): Define.
15986 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15988 2018-01-10 Martin Liska <mliska@suse.cz>
15991 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15992 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15994 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15995 CLEANUP_NO_PARTITIONING is not set.
15997 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15999 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16000 for vectors, as a partial revert of r254296.
16001 * rtl.h (const_vec_p): Delete.
16002 (const_vec_duplicate_p): Don't test for vector CONSTs.
16003 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16004 * expmed.c (make_tree): Likewise.
16007 * common.md (E, F): Use CONSTANT_P instead of checking for
16009 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16010 checking for CONST_VECTOR.
16012 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16014 PR middle-end/83575
16015 * predict.c (force_edge_cold): Handle in more sane way edges
16016 with no prediction.
16018 2018-01-09 Carl Love <cel@us.ibm.com>
16020 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16022 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16023 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16024 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16025 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16026 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16027 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16028 * config/rs6000/rs6000-protos.h: Add extern defition for
16029 rs6000_generate_float2_double_code.
16030 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16032 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16033 (float2_v2df): Add define_expand.
16035 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16038 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16039 op_mode in the force_to_mode call.
16041 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16043 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16044 instead of checking each element individually.
16045 (aarch64_evpc_uzp): Likewise.
16046 (aarch64_evpc_zip): Likewise.
16047 (aarch64_evpc_ext): Likewise.
16048 (aarch64_evpc_rev): Likewise.
16049 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16050 instead of checking each element individually. Return true without
16052 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16053 whether all selected elements come from the same input, instead of
16054 checking each element individually. Remove calls to gen_rtx_REG,
16055 start_sequence and end_sequence and instead assert that no rtl is
16058 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16060 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16061 order of HIGH and CONST checks.
16063 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16065 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16066 if the destination isn't an SSA_NAME.
16068 2018-01-09 Richard Biener <rguenther@suse.de>
16070 PR tree-optimization/83668
16071 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16073 (canonicalize_loop_form): ... here, renamed from ...
16074 (canonicalize_loop_closed_ssa_form): ... this and amended to
16075 swap successor edges for loop exit blocks to make us use
16076 the RPO order we need for initial schedule generation.
16078 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16080 PR tree-optimization/64811
16081 * match.pd: When optimizing comparisons with Inf, avoid
16082 introducing or losing exceptions from comparisons with NaN.
16084 2018-01-09 Martin Liska <mliska@suse.cz>
16087 * asan.c (shadow_mem_size): Add gcc_assert.
16089 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16091 Don't save registers in main().
16094 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16095 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16096 * config/avr/avr.c (avr_set_current_function): Don't error if
16097 naked, OS_task or OS_main are specified at the same time.
16098 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16100 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16102 * common/config/avr/avr-common.c (avr_option_optimization_table):
16103 Switch on -mmain-is-OS_task for optimizing compilations.
16105 2018-01-09 Richard Biener <rguenther@suse.de>
16107 PR tree-optimization/83572
16108 * graphite.c: Include cfganal.h.
16109 (graphite_transform_loops): Connect infinite loops to exit
16110 and remove fake edges at the end.
16112 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16114 * ipa-inline.c (edge_badness): Revert accidental checkin.
16116 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16119 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16120 symbols; not inline clones.
16122 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16125 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16126 hard registers. Formatting fixes.
16128 PR preprocessor/83722
16129 * gcc.c (try_generate_repro): Pass
16130 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16131 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16134 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16135 Kito Cheng <kito.cheng@gmail.com>
16137 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16138 (riscv_leaf_function_p): Delete.
16139 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16141 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16143 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16145 (do_ifelse): New function.
16146 (do_isel): New function.
16147 (do_sub3): New function.
16148 (do_add3): New function.
16149 (do_load_mask_compare): New function.
16150 (do_overlap_load_compare): New function.
16151 (expand_compare_loop): New function.
16152 (expand_block_compare): Call expand_compare_loop() when appropriate.
16153 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16154 option description.
16155 (-mblock-compare-inline-loop-limit): New option.
16157 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16160 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16161 Reverse order of second and third operands in first alternative.
16162 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16163 of first and second elements in UNSPEC_VPERMR vector.
16164 (altivec_expand_vec_perm_le): Likewise.
16166 2018-01-08 Jeff Law <law@redhat.com>
16168 PR rtl-optimizatin/81308
16169 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16170 (process_switch): If group_case_labels makes a change, then set
16172 (pass_convert_switch::execute): If a switch is converted, then
16173 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16175 PR rtl-optimization/81308
16176 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16179 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16181 PR target/83663 - Revert r255946
16182 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16183 generation for cases where splatting a value is not useful.
16184 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16185 across a vec_duplicate and a paradoxical subreg forming a vector
16186 mode to a vec_concat.
16188 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16190 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16191 -march=armv8.3-a variants.
16192 * config/arm/t-multilib: Likewise.
16193 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16195 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16197 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16199 (cceq_ior_compare_complement): Give it a name so I can use it, and
16200 change boolean_or_operator predicate to boolean_operator so it can
16201 be used to generate a crand.
16202 (eqne): New code iterator.
16203 (bd/bd_neg): New code_attrs.
16204 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16205 a single define_insn.
16206 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16207 decrement (bdnzt/bdnzf/bdzt/bdzf).
16208 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16209 with the new names of the branch decrement patterns, and added the
16210 names of the branch decrement conditional patterns.
16212 2018-01-08 Richard Biener <rguenther@suse.de>
16214 PR tree-optimization/83563
16215 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16218 2018-01-08 Richard Biener <rguenther@suse.de>
16220 PR middle-end/83713
16221 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16223 2018-01-08 Richard Biener <rguenther@suse.de>
16225 PR tree-optimization/83685
16226 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16227 references to abnormals.
16229 2018-01-08 Richard Biener <rguenther@suse.de>
16232 * dwarf2out.c (output_indirect_strings): Handle empty
16233 skeleton_debug_str_hash.
16234 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16236 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16238 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16239 (emit_store_direct): Likewise.
16240 (arc_trampoline_adjust_address): Likewise.
16241 (arc_asm_trampoline_template): New function.
16242 (arc_initialize_trampoline): Use asm_trampoline_template.
16243 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16244 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16245 * config/arc/arc.md (flush_icache): Delete pattern.
16247 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16249 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16250 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16253 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16256 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16257 by not USED_FOR_TARGET.
16258 (make_pass_resolve_sw_modes): Likewise.
16260 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16262 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16265 2018-01-08 Richard Biener <rguenther@suse.de>
16267 PR middle-end/83580
16268 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16270 2018-01-08 Richard Biener <rguenther@suse.de>
16272 PR middle-end/83517
16273 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16275 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16277 PR middle-end/81897
16278 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16279 basic blocks with a small number of successors.
16280 (convert_control_dep_chain_into_preds): Improve handling of
16282 (dump_predicates): Split apart into...
16283 (dump_pred_chain): ...here...
16284 (dump_pred_info): ...and here.
16285 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16286 (can_chain_union_be_invalidated_p): Improve check for invalidation
16288 (uninit_uses_cannot_happen): Avoid unnecessary if
16289 convert_control_dep_chain_into_preds yielded nothing.
16291 2018-01-06 Martin Sebor <msebor@redhat.com>
16293 PR tree-optimization/83640
16294 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16295 subtracting negative offset from size.
16296 (builtin_access::overlap): Adjust offset bounds of the access to fall
16297 within the size of the object if possible.
16299 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16301 PR rtl-optimization/83699
16302 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16303 extract_bit_field_as_subreg to cases in which the extracted
16304 value is also a vector.
16306 * lra-constraints.c (process_alt_operands): Test for the equivalence
16307 substitutions when detecting a possible reload cycle.
16309 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16312 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16313 by default if flag_selective_schedling{,2}. Formatting fixes.
16315 PR rtl-optimization/83682
16316 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16317 if it has non-VECTOR_MODE element mode.
16318 (vec_duplicate_p): Likewise.
16320 PR middle-end/83694
16321 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16322 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16324 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16327 * config/i386/i386-builtin.def
16328 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16329 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16330 Require also OPTION_MASK_ISA_AVX512F in addition to
16331 OPTION_MASK_ISA_GFNI.
16332 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16333 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16334 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16335 to OPTION_MASK_ISA_GFNI.
16336 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16337 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16338 OPTION_MASK_ISA_AVX512BW.
16339 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16340 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16341 addition to OPTION_MASK_ISA_GFNI.
16342 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16343 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16344 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16345 to OPTION_MASK_ISA_GFNI.
16346 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16347 a requirement for all ISAs rather than any of them with a few
16349 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16351 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16352 bitmasks to be enabled with 3 exceptions, instead of requiring any
16353 enabled ISA with lots of exceptions.
16354 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16355 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16356 Change avx512bw in isa attribute to avx512f.
16357 * config/i386/sgxintrin.h: Add license boilerplate.
16358 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16359 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16360 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16361 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16363 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16364 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16365 temporarily sse2 rather than sse if not enabled already.
16368 * config/i386/sse.md (VI248_VLBW): Rename to ...
16369 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16370 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16371 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16372 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16373 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16374 mode iterator instead of VI248_VLBW.
16376 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16378 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16379 (record_modified): Skip clobbers; add debug output.
16380 (param_change_prob): Use sreal frequencies.
16382 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16384 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16385 punt for user-aligned variables.
16387 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16389 * tree-chrec.c (chrec_contains_symbols): Return true for
16392 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16395 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16396 of (x|y) == x for BICS pattern.
16398 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16400 PR tree-optimization/83605
16401 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16402 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16405 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16407 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16408 * config/epiphany/rtems.h: New file.
16410 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16411 Uros Bizjak <ubizjak@gmail.com>
16414 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16415 QIreg_operand instead of register_operand predicate.
16416 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16417 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16418 comments instead of -fmitigate[-_]rop.
16420 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16423 * cgraphunit.c (symbol_table::compile): Switch to text_section
16424 before calling assembly_start debug hook.
16425 * run-rtl-passes.c (run_rtl_passes): Likewise.
16428 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16430 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16431 range_int_cst_p rather than !symbolic_range_p before calling
16432 extract_range_from_multiplicative_op_1.
16434 2018-01-04 Jeff Law <law@redhat.com>
16436 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16437 redundant test in assertion.
16439 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16441 * doc/rtl.texi: Document machine_mode wrapper classes.
16443 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16445 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16446 using tree_to_uhwi.
16448 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16450 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16451 the VEC_PERM_EXPR fold to fail.
16453 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16456 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16457 to switched_sections.
16459 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16462 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16463 test for d.testing.
16465 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16468 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16469 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16471 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16474 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16475 is BLKmode and bitpos not zero or mode change is needed.
16477 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16480 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16483 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16486 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16487 instead of MULT rtx. Update all corresponding splitters.
16488 (*saddl_se): Ditto.
16489 (*ssub<modesuffix>): Ditto.
16490 (*ssubl_se): Ditto.
16491 (*cmp_sadd_di): Update split patterns.
16492 (*cmp_sadd_si): Ditto.
16493 (*cmp_sadd_sidi): Ditto.
16494 (*cmp_ssub_di): Ditto.
16495 (*cmp_ssub_si): Ditto.
16496 (*cmp_ssub_sidi): Ditto.
16497 * config/alpha/predicates.md (const23_operand): New predicate.
16498 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16499 Look for ASHIFT, not MULT inner operand.
16500 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16502 2018-01-04 Martin Liska <mliska@suse.cz>
16504 PR gcov-profile/83669
16505 * gcov.c (output_intermediate_file): Add version to intermediate
16507 * doc/gcov.texi: Document new field 'version' in intermediate
16508 file format. Fix location of '-k' option of gcov command.
16510 2018-01-04 Martin Liska <mliska@suse.cz>
16513 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16515 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16517 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16519 2018-01-03 Martin Sebor <msebor@redhat.com>
16521 PR tree-optimization/83655
16522 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16523 checking calls with invalid arguments.
16525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16527 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16528 (vectorizable_mask_load_store): Delete.
16529 (vectorizable_call): Return false for masked loads and stores.
16530 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16531 instead of gimple_assign_rhs1.
16532 (vectorizable_load): Handle IFN_MASK_LOAD.
16533 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16537 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16539 (vectorizable_mask_load_store): ...here.
16540 (vectorizable_load): ...and here.
16542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16544 * tree-vect-stmts.c (vect_build_all_ones_mask)
16545 (vect_build_zero_merge_argument): New functions, split out from...
16546 (vectorizable_load): ...here.
16548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16550 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16552 (vectorizable_mask_load_store): ...here.
16553 (vectorizable_store): ...and here.
16555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16557 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16559 (vectorizable_mask_load_store): ...here.
16561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16563 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16564 (vect_model_store_cost): Take a vec_load_store_type instead of a
16566 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16567 (vect_model_store_cost): Take a vec_load_store_type instead of a
16569 (vectorizable_mask_load_store): Update accordingly.
16570 (vectorizable_store): Likewise.
16571 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16575 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16576 IFN_MASK_LOAD calls here rather than...
16577 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16580 Alan Hayward <alan.hayward@arm.com>
16581 David Sherwood <david.sherwood@arm.com>
16583 * expmed.c (extract_bit_field_1): For vector extracts,
16584 fall back to extract_bit_field_as_subreg if vec_extract
16587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16588 Alan Hayward <alan.hayward@arm.com>
16589 David Sherwood <david.sherwood@arm.com>
16591 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16592 they are variable or constant sized.
16593 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16594 slots for constant-sized data.
16596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16597 Alan Hayward <alan.hayward@arm.com>
16598 David Sherwood <david.sherwood@arm.com>
16600 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16601 handling COND_EXPRs with boolean comparisons, try to find a better
16602 basis for the mask type than the boolean itself.
16604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16606 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16607 is calculated and how it can be overridden.
16608 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16609 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16611 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16615 Alan Hayward <alan.hayward@arm.com>
16616 David Sherwood <david.sherwood@arm.com>
16618 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16619 Remove the mode argument.
16620 (aarch64_simd_valid_immediate): Remove the mode and inverse
16622 * config/aarch64/iterators.md (bitsize): New iterator.
16623 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16624 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16625 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16626 aarch64_simd_valid_immediate.
16627 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16628 (aarch64_reg_or_bic_imm): Likewise.
16629 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16630 with an insn_type enum and msl with a modifier_type enum.
16631 Replace element_width with a scalar_mode. Change the shift
16632 to unsigned int. Add constructors for scalar_float_mode and
16633 scalar_int_mode elements.
16634 (aarch64_vect_float_const_representable_p): Delete.
16635 (aarch64_can_const_movi_rtx_p)
16636 (aarch64_simd_scalar_immediate_valid_for_move)
16637 (aarch64_simd_make_constant): Update call to
16638 aarch64_simd_valid_immediate.
16639 (aarch64_advsimd_valid_immediate_hs): New function.
16640 (aarch64_advsimd_valid_immediate): Likewise.
16641 (aarch64_simd_valid_immediate): Remove mode and inverse
16642 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16643 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16644 and aarch64_float_const_representable_p on the result.
16645 (aarch64_output_simd_mov_immediate): Remove mode argument.
16646 Update call to aarch64_simd_valid_immediate and use of
16647 simd_immediate_info.
16648 (aarch64_output_scalar_simd_mov_immediate): Update call
16651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16652 Alan Hayward <alan.hayward@arm.com>
16653 David Sherwood <david.sherwood@arm.com>
16655 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16656 (mode_nunits): Likewise CONST_MODE_NUNITS.
16657 * machmode.def (ADJUST_NUNITS): Document.
16658 * genmodes.c (mode_data::need_nunits_adj): New field.
16659 (blank_mode): Update accordingly.
16660 (adj_nunits): New variable.
16661 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16663 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16664 listed in adj_nunits.
16665 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16666 listed in adj_nunits. Don't emit case statements for such modes.
16667 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16668 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16669 nothing if adj_nunits is nonnull.
16670 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16671 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16672 (emit_mode_fbit): Update use of print_maybe_const_decl.
16673 (emit_move_size): Likewise. Treat the array as non-const
16675 (emit_mode_adjustments): Handle adj_nunits.
16677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16679 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16680 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16681 (VECTOR_MODES): Use it.
16682 (make_vector_modes): Take the prefix as an argument.
16684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16685 Alan Hayward <alan.hayward@arm.com>
16686 David Sherwood <david.sherwood@arm.com>
16688 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16689 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16690 for MODE_VECTOR_BOOL.
16691 * machmode.def (VECTOR_BOOL_MODE): Document.
16692 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16693 (make_vector_bool_mode): New function.
16694 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16696 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16697 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16699 * stor-layout.c (int_mode_for_mode): Likewise.
16700 * tree.c (build_vector_type_for_mode): Likewise.
16701 * varasm.c (output_constant_pool_2): Likewise.
16702 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16703 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16704 for MODE_VECTOR_BOOL.
16705 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16706 of mode class checks.
16707 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16708 instead of a list of mode class checks.
16709 (expand_vector_scalar_condition): Likewise.
16710 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16713 Alan Hayward <alan.hayward@arm.com>
16714 David Sherwood <david.sherwood@arm.com>
16716 * machmode.h (mode_size): Change from unsigned short to
16718 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16719 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16720 or if measurement_type is not polynomial.
16721 (fixed_size_mode::includes_p): Check for constant-sized modes.
16722 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16723 return a poly_uint16 rather than an unsigned short.
16724 (emit_mode_size): Change the type of mode_size from unsigned short
16725 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16726 (emit_mode_adjustments): Cope with polynomial vector sizes.
16727 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16729 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16731 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16732 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16733 * caller-save.c (setup_save_areas): Likewise.
16734 (replace_reg_with_saved_mem): Likewise.
16735 * calls.c (emit_library_call_value_1): Likewise.
16736 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16737 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16738 (gen_lowpart_for_combine): Likewise.
16739 * convert.c (convert_to_integer_1): Likewise.
16740 * cse.c (equiv_constant, cse_insn): Likewise.
16741 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16742 (cselib_subst_to_values): Likewise.
16743 * dce.c (word_dce_process_block): Likewise.
16744 * df-problems.c (df_word_lr_mark_ref): Likewise.
16745 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16746 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16747 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16748 (rtl_for_decl_location): Likewise.
16749 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16750 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16751 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16752 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16753 (expand_expr_real_1): Likewise.
16754 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16755 (pad_below): Likewise.
16756 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16757 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16758 * ira.c (get_subreg_tracking_sizes): Likewise.
16759 * ira-build.c (ira_create_allocno_objects): Likewise.
16760 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16761 (ira_sort_regnos_for_alter_reg): Likewise.
16762 * ira-costs.c (record_operand_costs): Likewise.
16763 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16764 (resolve_simple_move): Likewise.
16765 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16766 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16767 (lra_constraints): Likewise.
16768 (CONST_POOL_OK_P): Reject variable-sized modes.
16769 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16770 (add_pseudo_to_slot, lra_spill): Likewise.
16771 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16772 * optabs-query.c (get_best_extraction_insn): Likewise.
16773 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16774 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16775 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16776 * recog.c (offsettable_address_addr_space_p): Likewise.
16777 * regcprop.c (maybe_mode_change): Likewise.
16778 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16779 * regrename.c (build_def_use): Likewise.
16780 * regstat.c (dump_reg_info): Likewise.
16781 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16782 (find_reloads, find_reloads_subreg_address): Likewise.
16783 * reload1.c (eliminate_regs_1): Likewise.
16784 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16785 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16786 (simplify_binary_operation_1, simplify_subreg): Likewise.
16787 * targhooks.c (default_function_arg_padding): Likewise.
16788 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16789 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16790 (verify_gimple_assign_ternary): Likewise.
16791 * tree-inline.c (estimate_move_cost): Likewise.
16792 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16793 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16794 (get_address_cost_ainc): Likewise.
16795 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16796 (vect_supportable_dr_alignment): Likewise.
16797 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16798 (vectorizable_reduction): Likewise.
16799 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16800 (vectorizable_operation, vectorizable_load): Likewise.
16801 * tree.c (build_same_sized_truth_vector_type): Likewise.
16802 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16803 * var-tracking.c (emit_note_insn_var_location): Likewise.
16804 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16805 (ADDR_VEC_ALIGN): Likewise.
16807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16808 Alan Hayward <alan.hayward@arm.com>
16809 David Sherwood <david.sherwood@arm.com>
16811 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16813 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16814 or if measurement_type is polynomial.
16815 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16816 * combine.c (make_extraction): Likewise.
16817 * dse.c (find_shift_sequence): Likewise.
16818 * dwarf2out.c (mem_loc_descriptor): Likewise.
16819 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16820 (extract_bit_field, extract_low_bits): Likewise.
16821 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16822 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16823 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16824 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16825 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16826 * reload.c (find_reloads): Likewise.
16827 * reload1.c (alter_reg): Likewise.
16828 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16829 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16830 * tree-if-conv.c (predicate_mem_writes): Likewise.
16831 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16832 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16833 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16834 * valtrack.c (dead_debug_insert_temp): Likewise.
16835 * varasm.c (mergeable_constant_section): Likewise.
16836 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16839 Alan Hayward <alan.hayward@arm.com>
16840 David Sherwood <david.sherwood@arm.com>
16842 * expr.c (expand_assignment): Cope with polynomial mode sizes
16843 when assigning to a CONCAT.
16845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16846 Alan Hayward <alan.hayward@arm.com>
16847 David Sherwood <david.sherwood@arm.com>
16849 * machmode.h (mode_precision): Change from unsigned short to
16851 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16853 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16854 or if measurement_type is not polynomial.
16855 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16856 in which the mode is already known to be a scalar_int_mode.
16857 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16858 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16860 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16861 for GET_MODE_PRECISION.
16862 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16863 for GET_MODE_PRECISION.
16864 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16866 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16867 (expand_field_assignment, make_extraction): Likewise.
16868 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16869 (get_last_value): Likewise.
16870 * convert.c (convert_to_integer_1): Likewise.
16871 * cse.c (cse_insn): Likewise.
16872 * expr.c (expand_expr_real_1): Likewise.
16873 * lra-constraints.c (simplify_operand_subreg): Likewise.
16874 * optabs-query.c (can_atomic_load_p): Likewise.
16875 * optabs.c (expand_atomic_load): Likewise.
16876 (expand_atomic_store): Likewise.
16877 * ree.c (combine_reaching_defs): Likewise.
16878 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16879 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16880 * tree.h (type_has_mode_precision_p): Likewise.
16881 * ubsan.c (instrument_si_overflow): Likewise.
16883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16884 Alan Hayward <alan.hayward@arm.com>
16885 David Sherwood <david.sherwood@arm.com>
16887 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16888 polynomial numbers of units.
16889 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16890 (valid_vector_subparts_p): New function.
16891 (build_vector_type): Remove temporary shim and take the number
16892 of units as a poly_uint64 rather than an int.
16893 (build_opaque_vector_type): Take the number of units as a
16894 poly_uint64 rather than an int.
16895 * tree.c (build_vector_from_ctor): Handle polynomial
16896 TYPE_VECTOR_SUBPARTS.
16897 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16898 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16899 (build_vector_from_val): If the number of units is variable,
16900 use build_vec_duplicate_cst for constant operands and
16901 VEC_DUPLICATE_EXPR otherwise.
16902 (make_vector_type): Remove temporary is_constant ().
16903 (build_vector_type, build_opaque_vector_type): Take the number of
16904 units as a poly_uint64 rather than an int.
16905 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16907 * cfgexpand.c (expand_debug_expr): Likewise.
16908 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16909 (store_constructor, expand_expr_real_1): Likewise.
16910 (const_scalar_mask_from_tree): Likewise.
16911 * fold-const-call.c (fold_const_reduction): Likewise.
16912 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16913 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16914 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16915 (fold_relational_const): Likewise.
16916 (native_interpret_vector): Likewise. Change the size from an
16917 int to an unsigned int.
16918 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16919 TYPE_VECTOR_SUBPARTS.
16920 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16921 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16922 duplicating a non-constant operand into a variable-length vector.
16923 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16924 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16925 * ipa-icf.c (sem_variable::equals): Likewise.
16926 * match.pd: Likewise.
16927 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16928 * print-tree.c (print_node): Likewise.
16929 * stor-layout.c (layout_type): Likewise.
16930 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16931 * tree-cfg.c (verify_gimple_comparison): Likewise.
16932 (verify_gimple_assign_binary): Likewise.
16933 (verify_gimple_assign_ternary): Likewise.
16934 (verify_gimple_assign_single): Likewise.
16935 * tree-pretty-print.c (dump_generic_node): Likewise.
16936 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16937 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16938 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16939 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16940 (vect_shift_permute_load_chain): Likewise.
16941 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16942 (expand_vector_condition, optimize_vector_constructor): Likewise.
16943 (lower_vec_perm, get_compute_type): Likewise.
16944 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16945 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16946 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16947 (vect_recog_mask_conversion_pattern): Likewise.
16948 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16949 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16950 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16951 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16952 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16953 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16954 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16955 (supportable_widening_operation): Likewise.
16956 (supportable_narrowing_operation): Likewise.
16957 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16959 * varasm.c (output_constant): Likewise.
16961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16962 Alan Hayward <alan.hayward@arm.com>
16963 David Sherwood <david.sherwood@arm.com>
16965 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16966 so that both the length == 3 and length != 3 cases set up their
16967 own permute vectors. Add comments explaining why we know the
16968 number of elements is constant.
16969 (vect_permute_load_chain): Likewise.
16971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16972 Alan Hayward <alan.hayward@arm.com>
16973 David Sherwood <david.sherwood@arm.com>
16975 * machmode.h (mode_nunits): Change from unsigned char to
16977 (ONLY_FIXED_SIZE_MODES): New macro.
16978 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16979 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16980 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16982 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16983 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16984 or if measurement_type is not polynomial.
16985 * genmodes.c (ZERO_COEFFS): New macro.
16986 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16988 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16989 Use ZERO_COEFFS when emitting initializers.
16990 * data-streamer.h (bp_pack_poly_value): New function.
16991 (bp_unpack_poly_value): Likewise.
16992 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16993 for GET_MODE_NUNITS.
16994 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16995 for GET_MODE_NUNITS.
16996 * tree.c (make_vector_type): Remove temporary shim and make
16997 the real function take the number of units as a poly_uint64
16998 rather than an int.
16999 (build_vector_type_for_mode): Handle polynomial nunits.
17000 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17001 * emit-rtl.c (const_vec_series_p_1): Likewise.
17002 (gen_rtx_CONST_VECTOR): Likewise.
17003 * fold-const.c (test_vec_duplicate_folding): Likewise.
17004 * genrecog.c (validate_pattern): Likewise.
17005 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17006 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17007 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17008 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17009 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17010 * rtlanal.c (subreg_get_info): Likewise.
17011 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17012 (vect_grouped_load_supported): Likewise.
17013 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17014 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17015 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17016 (simplify_const_unary_operation, simplify_binary_operation_1)
17017 (simplify_const_binary_operation, simplify_ternary_operation)
17018 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17019 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17020 instead of CONST_VECTOR_NUNITS.
17021 * varasm.c (output_constant_pool_2): Likewise.
17022 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17023 explicit-encoded elements in the XVEC for variable-length vectors.
17025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17027 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17030 Alan Hayward <alan.hayward@arm.com>
17031 David Sherwood <david.sherwood@arm.com>
17033 * coretypes.h (fixed_size_mode): Declare.
17034 (fixed_size_mode_pod): New typedef.
17035 * builtins.h (target_builtins::x_apply_args_mode)
17036 (target_builtins::x_apply_result_mode): Change type to
17037 fixed_size_mode_pod.
17038 * builtins.c (apply_args_size, apply_result_size, result_vector)
17039 (expand_builtin_apply_args_1, expand_builtin_apply)
17040 (expand_builtin_return): Update accordingly.
17042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17044 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17045 * cselib.c (cselib_hash_rtx): Likewise.
17046 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17047 CONST_VECTOR encoding.
17049 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17050 Jeff Law <law@redhat.com>
17053 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17054 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17055 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17056 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17059 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17060 explicitly probe *sp in a noreturn function if there were any callee
17061 register saves or frame pointer is needed.
17063 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17066 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17067 BLKmode for ternary, binary or unary expressions.
17070 * var-tracking.c (delete_vta_debug_insn): New inline function.
17071 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17072 insns from get_insns () to NULL instead of each bb separately.
17073 Use delete_vta_debug_insn. No longer static.
17074 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17075 delete_vta_debug_insns callers.
17076 * rtl.h (delete_vta_debug_insns): Declare.
17077 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17078 instead of variable_tracking_main.
17080 2018-01-03 Martin Sebor <msebor@redhat.com>
17082 PR tree-optimization/83603
17083 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17084 arguments past the endof the argument list in functions declared
17085 without a prototype.
17086 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17087 Avoid checking when arguments are null.
17089 2018-01-03 Martin Sebor <msebor@redhat.com>
17092 * doc/extend.texi (attribute const): Fix a typo.
17093 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17094 issuing -Wsuggest-attribute for void functions.
17096 2018-01-03 Martin Sebor <msebor@redhat.com>
17098 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17099 offset_int::from instead of wide_int::to_shwi.
17100 (maybe_diag_overlap): Remove assertion.
17101 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17102 * gimple-ssa-sprintf.c (format_directive): Same.
17103 (parse_directive): Same.
17104 (sprintf_dom_walker::compute_format_length): Same.
17105 (try_substitute_return_value): Same.
17107 2018-01-03 Jeff Law <law@redhat.com>
17109 PR middle-end/83654
17110 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17111 non-constant residual for zero at runtime and avoid probing in
17112 that case. Reorganize code for trailing problem to mirror handling
17115 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17117 PR tree-optimization/83501
17118 * tree-ssa-strlen.c (get_string_cst): New.
17119 (handle_char_store): Call get_string_cst.
17121 2018-01-03 Martin Liska <mliska@suse.cz>
17123 PR tree-optimization/83593
17124 * tree-ssa-strlen.c: Include tree-cfg.h.
17125 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17126 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17127 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17129 (strlen_dom_walker::before_dom_children): Call
17130 gimple_purge_dead_eh_edges. Dump tranformation with details
17132 (strlen_dom_walker::before_dom_children): Update call by adding
17133 new argument cleanup_eh.
17134 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17136 2018-01-03 Martin Liska <mliska@suse.cz>
17139 * cif-code.def (VARIADIC_THUNK): New enum value.
17140 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17143 2018-01-03 Jan Beulich <jbeulich@suse.com>
17145 * sse.md (mov<mode>_internal): Tighten condition for when to use
17146 vmovdqu<ssescalarsize> for TI and OI modes.
17148 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17150 Update copyright years.
17152 2018-01-03 Martin Liska <mliska@suse.cz>
17155 * ipa-visibility.c (function_and_variable_visibility): Skip
17156 functions with noipa attribure.
17158 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17160 * gcc.c (process_command): Update copyright notice dates.
17161 * gcov-dump.c (print_version): Ditto.
17162 * gcov.c (print_version): Ditto.
17163 * gcov-tool.c (print_version): Ditto.
17164 * gengtype.c (create_file): Ditto.
17165 * doc/cpp.texi: Bump @copying's copyright year.
17166 * doc/cppinternals.texi: Ditto.
17167 * doc/gcc.texi: Ditto.
17168 * doc/gccint.texi: Ditto.
17169 * doc/gcov.texi: Ditto.
17170 * doc/install.texi: Ditto.
17171 * doc/invoke.texi: Ditto.
17173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17175 * vector-builder.h (vector_builder::m_full_nelts): Change from
17176 unsigned int to poly_uint64.
17177 (vector_builder::full_nelts): Update prototype accordingly.
17178 (vector_builder::new_vector): Likewise.
17179 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17180 (vector_builder::operator ==): Likewise.
17181 (vector_builder::finalize): Likewise.
17182 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17183 Take the number of elements as a poly_uint64 rather than an
17185 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17186 from unsigned int to poly_uint64.
17187 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17188 (vec_perm_indices::new_vector): Likewise.
17189 (vec_perm_indices::length): Likewise.
17190 (vec_perm_indices::nelts_per_input): Likewise.
17191 (vec_perm_indices::input_nelts): Likewise.
17192 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17193 number of elements per input as a poly_uint64 rather than an
17194 unsigned int. Use the original encoding for variable-length
17195 vectors, rather than clamping each individual element.
17196 For the second and subsequent elements in each pattern,
17197 clamp the step and base before clamping their sum.
17198 (vec_perm_indices::series_p): Handle polynomial element counts.
17199 (vec_perm_indices::all_in_range_p): Likewise.
17200 (vec_perm_indices_to_tree): Likewise.
17201 (vec_perm_indices_to_rtx): Likewise.
17202 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17203 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17204 (tree_vector_builder::new_binary_operation): Handle polynomial
17205 element counts. Return false if we need to know the number
17206 of elements at compile time.
17207 * fold-const.c (fold_vec_perm): Punt if the number of elements
17208 isn't known at compile time.
17210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17212 * vec-perm-indices.h (vec_perm_builder): Change element type
17213 from HOST_WIDE_INT to poly_int64.
17214 (vec_perm_indices::element_type): Update accordingly.
17215 (vec_perm_indices::clamp): Handle polynomial element_types.
17216 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17217 (vec_perm_indices::all_in_range_p): Likewise.
17218 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17220 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17221 polynomial vec_perm_indices element types.
17222 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17223 * fold-const.c (fold_vec_perm): Likewise.
17224 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17225 * tree-vect-generic.c (lower_vec_perm): Likewise.
17226 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17227 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17228 element type to HOST_WIDE_INT.
17230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17231 Alan Hayward <alan.hayward@arm.com>
17232 David Sherwood <david.sherwood@arm.com>
17234 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17235 rather than an int. Use plus_constant.
17236 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17237 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17240 Alan Hayward <alan.hayward@arm.com>
17241 David Sherwood <david.sherwood@arm.com>
17243 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17244 a HOST_WIDE_INT to a poly_int64.
17246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17247 Alan Hayward <alan.hayward@arm.com>
17248 David Sherwood <david.sherwood@arm.com>
17250 * calls.c (load_register_parameters): Cope with polynomial
17251 mode sizes. Require a constant size for BLKmode parameters
17252 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17253 forces a parameter to be padded at the lsb end in order to
17254 fill a complete number of words, require the parameter size
17255 to be ordered wrt UNITS_PER_WORD.
17257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17258 Alan Hayward <alan.hayward@arm.com>
17259 David Sherwood <david.sherwood@arm.com>
17261 * reload1.c (spill_stack_slot_width): Change element type
17262 from unsigned int to poly_uint64_pod.
17263 (alter_reg): Treat mode sizes as polynomial.
17265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17266 Alan Hayward <alan.hayward@arm.com>
17267 David Sherwood <david.sherwood@arm.com>
17269 * reload.c (complex_word_subreg_p): New function.
17270 (reload_inner_reg_of_subreg, push_reload): Use it.
17272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17273 Alan Hayward <alan.hayward@arm.com>
17274 David Sherwood <david.sherwood@arm.com>
17276 * lra-constraints.c (process_alt_operands): Reject matched
17277 operands whose sizes aren't ordered.
17278 (match_reload): Refer to this check here.
17280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17281 Alan Hayward <alan.hayward@arm.com>
17282 David Sherwood <david.sherwood@arm.com>
17284 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17285 that the mode size is in the set {1, 2, 4, 8, 16}.
17287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17288 Alan Hayward <alan.hayward@arm.com>
17289 David Sherwood <david.sherwood@arm.com>
17291 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17292 Use plus_constant instead of gen_rtx_PLUS.
17294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17295 Alan Hayward <alan.hayward@arm.com>
17296 David Sherwood <david.sherwood@arm.com>
17298 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17299 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17300 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17301 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17302 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17303 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17304 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17305 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17306 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17307 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17309 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17310 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17311 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17312 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17313 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17314 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17315 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17316 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17317 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17318 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17320 * expr.c (emit_move_resolve_push): Treat the input and result
17321 of PUSH_ROUNDING as a poly_int64.
17322 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17323 (emit_push_insn): Likewise.
17324 * lra-eliminations.c (mark_not_eliminable): Likewise.
17325 * recog.c (push_operand): Likewise.
17326 * reload1.c (elimination_effects): Likewise.
17327 * rtlanal.c (nonzero_bits1): Likewise.
17328 * calls.c (store_one_arg): Likewise. Require the padding to be
17329 known at compile time.
17331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17332 Alan Hayward <alan.hayward@arm.com>
17333 David Sherwood <david.sherwood@arm.com>
17335 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17336 Use plus_constant instead of gen_rtx_PLUS.
17338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17339 Alan Hayward <alan.hayward@arm.com>
17340 David Sherwood <david.sherwood@arm.com>
17342 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17343 rather than an int.
17345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17346 Alan Hayward <alan.hayward@arm.com>
17347 David Sherwood <david.sherwood@arm.com>
17349 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17350 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17351 via stack temporaries. Treat the mode size as polynomial too.
17353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17354 Alan Hayward <alan.hayward@arm.com>
17355 David Sherwood <david.sherwood@arm.com>
17357 * expr.c (expand_expr_real_2): When handling conversions involving
17358 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17359 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17360 as a poly_uint64 too.
17362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17363 Alan Hayward <alan.hayward@arm.com>
17364 David Sherwood <david.sherwood@arm.com>
17366 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17369 Alan Hayward <alan.hayward@arm.com>
17370 David Sherwood <david.sherwood@arm.com>
17372 * combine.c (can_change_dest_mode): Handle polynomial
17373 REGMODE_NATURAL_SIZE.
17374 * expmed.c (store_bit_field_1): Likewise.
17375 * expr.c (store_constructor): Likewise.
17376 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17377 and polynomial REGMODE_NATURAL_SIZE.
17378 (gen_lowpart_common): Likewise.
17379 * reginfo.c (record_subregs_of_mode): Likewise.
17380 * rtlanal.c (read_modify_subreg_p): Likewise.
17382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17383 Alan Hayward <alan.hayward@arm.com>
17384 David Sherwood <david.sherwood@arm.com>
17386 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17387 numbers of elements.
17389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17390 Alan Hayward <alan.hayward@arm.com>
17391 David Sherwood <david.sherwood@arm.com>
17393 * match.pd: Cope with polynomial numbers of vector elements.
17395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17396 Alan Hayward <alan.hayward@arm.com>
17397 David Sherwood <david.sherwood@arm.com>
17399 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17400 in a POINTER_PLUS_EXPR.
17402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17403 Alan Hayward <alan.hayward@arm.com>
17404 David Sherwood <david.sherwood@arm.com>
17406 * omp-simd-clone.c (simd_clone_subparts): New function.
17407 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17408 (ipa_simd_modify_function_body): Likewise.
17410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17411 Alan Hayward <alan.hayward@arm.com>
17412 David Sherwood <david.sherwood@arm.com>
17414 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17415 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17416 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17417 (expand_vector_condition, vector_element): Likewise.
17418 (subparts_gt): New function.
17419 (get_compute_type): Use subparts_gt.
17420 (count_type_subparts): Delete.
17421 (expand_vector_operations_1): Use subparts_gt instead of
17422 count_type_subparts.
17424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17425 Alan Hayward <alan.hayward@arm.com>
17426 David Sherwood <david.sherwood@arm.com>
17428 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17429 (vect_compile_time_alias): ...this new function. Do the calculation
17430 on poly_ints rather than trees.
17431 (vect_prune_runtime_alias_test_list): Update call accordingly.
17433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17434 Alan Hayward <alan.hayward@arm.com>
17435 David Sherwood <david.sherwood@arm.com>
17437 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17439 (vect_schedule_slp_instance): Likewise.
17441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17442 Alan Hayward <alan.hayward@arm.com>
17443 David Sherwood <david.sherwood@arm.com>
17445 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17446 constant and extern definitions for variable-length vectors.
17447 (vect_get_constant_vectors): Note that the number of units
17448 is known to be constant.
17450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17451 Alan Hayward <alan.hayward@arm.com>
17452 David Sherwood <david.sherwood@arm.com>
17454 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17455 of units as polynomial. Choose between WIDE and NARROW based
17458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17459 Alan Hayward <alan.hayward@arm.com>
17460 David Sherwood <david.sherwood@arm.com>
17462 * tree-vect-stmts.c (simd_clone_subparts): New function.
17463 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17466 Alan Hayward <alan.hayward@arm.com>
17467 David Sherwood <david.sherwood@arm.com>
17469 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17470 vectors as polynomial. Use build_index_vector for
17471 IFN_GOMP_SIMD_LANE.
17473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17474 Alan Hayward <alan.hayward@arm.com>
17475 David Sherwood <david.sherwood@arm.com>
17477 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17478 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17479 for variable-length vectors.
17480 (vectorizable_mask_load_store): Treat the number of units as
17481 polynomial, asserting that it is constant if the condition has
17482 already been enforced.
17483 (vectorizable_store, vectorizable_load): Likewise.
17485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17486 Alan Hayward <alan.hayward@arm.com>
17487 David Sherwood <david.sherwood@arm.com>
17489 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17490 of units as polynomial. Punt if we can't tell at compile time
17491 which vector contains the final result.
17493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17494 Alan Hayward <alan.hayward@arm.com>
17495 David Sherwood <david.sherwood@arm.com>
17497 * tree-vect-loop.c (vectorizable_induction): Treat the number
17498 of units as polynomial. Punt on SLP inductions. Use an integer
17499 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17500 cast of such a series for variable-length floating-point
17503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17504 Alan Hayward <alan.hayward@arm.com>
17505 David Sherwood <david.sherwood@arm.com>
17507 * tree.h (build_index_vector): Declare.
17508 * tree.c (build_index_vector): New function.
17509 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17510 of units as polynomial, forcibly converting it to a constant if
17511 vectorizable_reduction has already enforced the condition.
17512 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17513 to create a {1,2,3,...} vector.
17514 (vectorizable_reduction): Treat the number of units as polynomial.
17515 Choose vectype_in based on the largest scalar element size rather
17516 than the smallest number of units. Enforce the restrictions
17519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17520 Alan Hayward <alan.hayward@arm.com>
17521 David Sherwood <david.sherwood@arm.com>
17523 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17524 number of units as polynomial.
17526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17527 Alan Hayward <alan.hayward@arm.com>
17528 David Sherwood <david.sherwood@arm.com>
17530 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17531 * target.def (autovectorize_vector_sizes): Return the vector sizes
17532 by pointer, using vector_sizes rather than a bitmask.
17533 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17534 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17535 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17537 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17538 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17539 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17540 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17541 * omp-general.c (omp_max_vf): Likewise.
17542 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17543 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17544 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17545 * tree-vect-slp.c (vect_slp_bb): Likewise.
17546 * doc/tm.texi: Regenerate.
17547 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17549 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17550 the vector size as a poly_uint64 rather than an unsigned int.
17551 (current_vector_size): Change from an unsigned int to a poly_uint64.
17552 (get_vectype_for_scalar_type): Update accordingly.
17553 * tree.h (build_truth_vector_type): Take the size and number of
17554 units as a poly_uint64 rather than an unsigned int.
17555 (build_vector_type): Add a temporary overload that takes
17556 the number of units as a poly_uint64 rather than an unsigned int.
17557 * tree.c (make_vector_type): Likewise.
17558 (build_truth_vector_type): Take the number of units as a poly_uint64
17559 rather than an unsigned int.
17561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17562 Alan Hayward <alan.hayward@arm.com>
17563 David Sherwood <david.sherwood@arm.com>
17565 * target.def (get_mask_mode): Take the number of units and length
17566 as poly_uint64s rather than unsigned ints.
17567 * targhooks.h (default_get_mask_mode): Update accordingly.
17568 * targhooks.c (default_get_mask_mode): Likewise.
17569 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17570 * doc/tm.texi: Regenerate.
17572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17573 Alan Hayward <alan.hayward@arm.com>
17574 David Sherwood <david.sherwood@arm.com>
17576 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17577 * omp-general.c (omp_max_vf): Likewise.
17578 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17579 (expand_omp_simd): Handle polynomial safelen.
17580 * omp-low.c (omplow_simd_context): Add a default constructor.
17581 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17582 (lower_rec_simd_input_clauses): Update accordingly.
17583 (lower_rec_input_clauses): Likewise.
17585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17586 Alan Hayward <alan.hayward@arm.com>
17587 David Sherwood <david.sherwood@arm.com>
17589 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17590 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17591 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17592 (vect_analyze_slp_cost): Likewise.
17593 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17594 (vect_model_load_cost): Likewise.
17596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17597 Alan Hayward <alan.hayward@arm.com>
17598 David Sherwood <david.sherwood@arm.com>
17600 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17601 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17602 from an unsigned int * to a poly_uint64_pod *.
17603 (calculate_unrolling_factor): New function.
17604 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17607 Alan Hayward <alan.hayward@arm.com>
17608 David Sherwood <david.sherwood@arm.com>
17610 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17611 from an unsigned int to a poly_uint64.
17612 (_loop_vec_info::slp_unrolling_factor): Likewise.
17613 (_loop_vec_info::vectorization_factor): Change from an int
17615 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17616 (vect_get_num_vectors): New function.
17617 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17618 (vect_get_num_copies): Use vect_get_num_vectors.
17619 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17620 to an unsigned int *.
17621 (vect_analyze_data_refs): Change min_vf from an int * to a
17623 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17624 than an unsigned HOST_WIDE_INT.
17625 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17626 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17627 to an unsigned int *.
17628 (vect_analyze_data_ref_dependences): Likewise.
17629 (vect_compute_data_ref_alignment): Handle polynomial vf.
17630 (vect_enhance_data_refs_alignment): Likewise.
17631 (vect_prune_runtime_alias_test_list): Likewise.
17632 (vect_shift_permute_load_chain): Likewise.
17633 (vect_supportable_dr_alignment): Likewise.
17634 (dependence_distance_ge_vf): Take the vectorization factor as a
17635 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17636 (vect_analyze_data_refs): Change min_vf from an int * to a
17638 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17639 vfm1 as a poly_uint64 rather than an int. Make the same change
17640 for the returned bound_scalar.
17641 (vect_gen_vector_loop_niters): Handle polynomial vf.
17642 (vect_do_peeling): Likewise. Update call to
17643 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17644 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17646 * tree-vect-loop.c (vect_determine_vectorization_factor)
17647 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17648 (vect_get_known_peeling_cost): Likewise.
17649 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17650 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17651 (vect_transform_loop): Likewise. Use the lowest possible VF when
17652 updating the upper bounds of the loop.
17653 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17654 rather than an int.
17655 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17656 polynomial unroll factors.
17657 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17658 (vect_make_slp_decision): Likewise.
17659 (vect_supported_load_permutation_p): Likewise, and polynomial
17661 (vect_analyze_slp_cost): Handle polynomial vf.
17662 (vect_slp_analyze_node_operations): Likewise.
17663 (vect_slp_analyze_bb_1): Likewise.
17664 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17665 than an unsigned HOST_WIDE_INT.
17666 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17667 (vectorizable_load): Handle polynomial vf.
17668 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17670 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17673 Alan Hayward <alan.hayward@arm.com>
17674 David Sherwood <david.sherwood@arm.com>
17676 * match.pd: Handle bit operations involving three constants
17677 and try to fold one pair.
17679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17681 * tree-vect-loop-manip.c: Include gimple-fold.h.
17682 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17683 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17684 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17685 Add a path that uses a step of VF instead of 1, but disable it
17687 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17688 and niters_no_overflow parameters. Update calls to
17689 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17690 Create a new SSA name if the latter choses to use a ste other
17691 than zero, and return it via niters_vector_mult_vf_var.
17692 * tree-vect-loop.c (vect_transform_loop): Update calls to
17693 vect_do_peeling, vect_gen_vector_loop_niters and
17694 slpeel_make_loop_iterate_ntimes.
17695 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17696 (vect_gen_vector_loop_niters): Update declarations after above changes.
17698 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17700 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17701 128-bit round to integer instructions.
17702 (ceil<mode>2): Likewise.
17703 (btrunc<mode>2): Likewise.
17704 (round<mode>2): Likewise.
17706 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17708 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17709 unaligned VSX load/store on P8/P9.
17710 (expand_block_clear): Allow the use of unaligned VSX
17711 load/store on P8/P9.
17713 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17715 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17717 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17718 swap associated with both a load and a store.
17720 2018-01-02 Andrew Waterman <andrew@sifive.com>
17722 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17723 * config/riscv/riscv.md (clear_cache): Use it.
17725 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17727 * web.c: Remove out-of-date comment.
17729 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17731 * expr.c (fixup_args_size_notes): Check that any existing
17732 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17733 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17734 (emit_single_push_insn): ...here.
17736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17738 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17739 (const_vector_encoded_nelts): New function.
17740 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17741 (const_vector_int_elt, const_vector_elt): Declare.
17742 * emit-rtl.c (const_vector_int_elt_1): New function.
17743 (const_vector_elt): Likewise.
17744 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17745 of CONST_VECTOR_ELT.
17747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17749 * expr.c: Include rtx-vector-builder.h.
17750 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17751 directly on the tree encoding.
17752 (const_vector_from_tree): Likewise.
17753 * optabs.c: Include rtx-vector-builder.h.
17754 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17755 sequence of "u" values.
17756 * vec-perm-indices.c: Include rtx-vector-builder.h.
17757 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17758 directly on the vec_perm_indices encoding.
17760 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17762 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17763 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17764 * rtx-vector-builder.h: New file.
17765 * rtx-vector-builder.c: Likewise.
17766 * rtl.h (rtx_def::u2): Add a const_vector field.
17767 (CONST_VECTOR_NPATTERNS): New macro.
17768 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17769 (CONST_VECTOR_DUPLICATE_P): Likewise.
17770 (CONST_VECTOR_STEPPED_P): Likewise.
17771 (CONST_VECTOR_ENCODED_ELT): Likewise.
17772 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17773 (unwrap_const_vec_duplicate): Likewise.
17774 (const_vec_series_p): Check for a non-duplicated vector encoding.
17775 Say that the function only returns true for integer vectors.
17776 * emit-rtl.c: Include rtx-vector-builder.h.
17777 (gen_const_vec_duplicate_1): Delete.
17778 (gen_const_vector): Call gen_const_vec_duplicate instead of
17779 gen_const_vec_duplicate_1.
17780 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17781 (gen_const_vec_duplicate): Use rtx_vector_builder.
17782 (gen_const_vec_series): Likewise.
17783 (gen_rtx_CONST_VECTOR): Likewise.
17784 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17785 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17786 Build a new vector rather than modifying a CONST_VECTOR in-place.
17787 (handle_special_swappables): Update call accordingly.
17788 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17789 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17790 Build a new vector rather than modifying a CONST_VECTOR in-place.
17791 (handle_special_swappables): Update call accordingly.
17793 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17795 * simplify-rtx.c (simplify_const_binary_operation): Use
17796 CONST_VECTOR_ELT instead of XVECEXP.
17798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17800 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17801 the selector elements to be different from the data elements
17802 if the selector is a VECTOR_CST.
17803 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17804 ssizetype for the selector.
17806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17808 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17809 before testing each element individually.
17810 * tree-vect-generic.c (lower_vec_perm): Likewise.
17812 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17814 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17815 * selftest-run-tests.c (selftest::run_tests): Call it.
17816 * vector-builder.h (vector_builder::operator ==): New function.
17817 (vector_builder::operator !=): Likewise.
17818 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17819 (vec_perm_indices::all_from_input_p): New function.
17820 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17821 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17822 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17823 instead of reading the VECTOR_CST directly. Detect whether both
17824 vector inputs are the same before constructing the vec_perm_indices,
17825 and update the number of inputs argument accordingly. Use the
17826 utility functions added above. Only construct sel2 if we need to.
17828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17830 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17831 the broadcast of the low byte.
17832 (expand_mult_highpart): Use an explicit encoding for the permutes.
17833 * optabs-query.c (can_mult_highpart_p): Likewise.
17834 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17835 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17836 (vectorizable_bswap): Likewise.
17837 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17838 explicit encoding for the power-of-2 permutes.
17839 (vect_permute_store_chain): Likewise.
17840 (vect_grouped_load_supported): Likewise.
17841 (vect_permute_load_chain): Likewise.
17843 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17845 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17846 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17847 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17848 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17849 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17850 (vect_gen_perm_mask_any): Likewise.
17852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17854 * int-vector-builder.h: New file.
17855 * vec-perm-indices.h: Include int-vector-builder.h.
17856 (vec_perm_indices): Redefine as an int_vector_builder.
17857 (auto_vec_perm_indices): Delete.
17858 (vec_perm_builder): Redefine as a stand-alone class.
17859 (vec_perm_indices::vec_perm_indices): New function.
17860 (vec_perm_indices::clamp): Likewise.
17861 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17862 (vec_perm_indices::new_vector): New function.
17863 (vec_perm_indices::new_expanded_vector): Update for new
17864 vec_perm_indices class.
17865 (vec_perm_indices::rotate_inputs): New function.
17866 (vec_perm_indices::all_in_range_p): Operate directly on the
17867 encoded form, without computing elided elements.
17868 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17869 encoding. Update for new vec_perm_indices class.
17870 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17871 the given vec_perm_builder.
17872 (expand_vec_perm_var): Update vec_perm_builder constructor.
17873 (expand_mult_highpart): Use vec_perm_builder instead of
17874 auto_vec_perm_indices.
17875 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17876 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17877 or double series encoding as appropriate.
17878 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17879 vec_perm_indices instead of auto_vec_perm_indices.
17880 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17881 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17882 (vect_permute_store_chain): Likewise.
17883 (vect_grouped_load_supported): Likewise.
17884 (vect_permute_load_chain): Likewise.
17885 (vect_shift_permute_load_chain): Likewise.
17886 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17887 (vect_transform_slp_perm_load): Likewise.
17888 (vect_schedule_slp_instance): Likewise.
17889 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17890 (vectorizable_mask_load_store): Likewise.
17891 (vectorizable_bswap): Likewise.
17892 (vectorizable_store): Likewise.
17893 (vectorizable_load): Likewise.
17894 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17895 vec_perm_indices instead of auto_vec_perm_indices. Use
17896 tree_to_vec_perm_builder to read the vector from a tree.
17897 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17898 vec_perm_builder instead of a vec_perm_indices.
17899 (have_whole_vector_shift): Use vec_perm_builder and
17900 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17901 truncation to calc_vec_perm_mask_for_shift.
17902 (vect_create_epilog_for_reduction): Likewise.
17903 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17904 from auto_vec_perm_indices to vec_perm_indices.
17905 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17906 instead of changing individual elements.
17907 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17908 the vector in d.perm.
17909 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17910 from auto_vec_perm_indices to vec_perm_indices.
17911 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17912 instead of changing individual elements.
17913 (arm_vectorize_vec_perm_const): Use new_vector to install
17914 the vector in d.perm.
17915 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17916 Update vec_perm_builder constructor.
17917 (rs6000_expand_interleave): Likewise.
17918 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17919 (rs6000_expand_interleave): Likewise.
17921 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17923 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17924 to qimode could truncate the indices.
17925 * optabs.c (expand_vec_perm_var): Likewise.
17927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17929 * Makefile.in (OBJS): Add vec-perm-indices.o.
17930 * vec-perm-indices.h: New file.
17931 * vec-perm-indices.c: Likewise.
17932 * target.h (vec_perm_indices): Replace with a forward class
17934 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17935 * optabs.h: Include vec-perm-indices.h.
17936 (expand_vec_perm): Delete.
17937 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17938 (expand_vec_perm_const): Declare.
17939 * target.def (vec_perm_const_ok): Replace with...
17940 (vec_perm_const): ...this new hook.
17941 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17942 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17943 * doc/tm.texi: Regenerate.
17944 * optabs.def (vec_perm_const): Delete.
17945 * doc/md.texi (vec_perm_const): Likewise.
17946 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17947 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17948 expand_vec_perm for constant permutation vectors. Assert that
17949 the mode of variable permutation vectors is the integer equivalent
17950 of the mode that is being permuted.
17951 * optabs-query.h (selector_fits_mode_p): Declare.
17952 * optabs-query.c: Include vec-perm-indices.h.
17953 (selector_fits_mode_p): New function.
17954 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17955 is defined, instead of checking whether the vec_perm_const_optab
17956 exists. Use targetm.vectorize.vec_perm_const instead of
17957 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17958 fit in the vector mode before using a variable permute.
17959 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17960 vec_perm_indices instead of an rtx.
17961 (expand_vec_perm): Replace with...
17962 (expand_vec_perm_const): ...this new function. Take the selector
17963 as a vec_perm_indices rather than an rtx. Also take the mode of
17964 the selector. Update call to shift_amt_for_vec_perm_mask.
17965 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17966 Use vec_perm_indices::new_expanded_vector to expand the original
17967 selector into bytes. Check whether the indices fit in the vector
17968 mode before using a variable permute.
17969 (expand_vec_perm_var): Make global.
17970 (expand_mult_highpart): Use expand_vec_perm_const.
17971 * fold-const.c: Includes vec-perm-indices.h.
17972 * tree-ssa-forwprop.c: Likewise.
17973 * tree-vect-data-refs.c: Likewise.
17974 * tree-vect-generic.c: Likewise.
17975 * tree-vect-loop.c: Likewise.
17976 * tree-vect-slp.c: Likewise.
17977 * tree-vect-stmts.c: Likewise.
17978 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17980 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17981 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17982 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17983 (aarch64_vectorize_vec_perm_const): ...this new function.
17984 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17985 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17986 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17987 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17988 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17989 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17990 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17992 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17993 check for NEON modes.
17994 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17995 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17996 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17997 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17999 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18000 the old VEC_PERM_CONST conditions.
18001 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18002 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18003 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18004 (ia64_vectorize_vec_perm_const_ok): Merge into...
18005 (ia64_vectorize_vec_perm_const): ...this new function.
18006 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18007 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18008 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18009 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18010 * config/mips/mips.c (mips_expand_vec_perm_const)
18011 (mips_vectorize_vec_perm_const_ok): Merge into...
18012 (mips_vectorize_vec_perm_const): ...this new function.
18013 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18014 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18015 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18016 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18017 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18018 (rs6000_expand_vec_perm_const): Delete.
18019 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18021 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18022 (altivec_expand_vec_perm_const_le): Take each operand individually.
18023 Operate on constant selectors rather than rtxes.
18024 (altivec_expand_vec_perm_const): Likewise. Update call to
18025 altivec_expand_vec_perm_const_le.
18026 (rs6000_expand_vec_perm_const): Delete.
18027 (rs6000_vectorize_vec_perm_const_ok): Delete.
18028 (rs6000_vectorize_vec_perm_const): New function.
18029 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18030 an element count and rtx array.
18031 (rs6000_expand_extract_even): Update call accordingly.
18032 (rs6000_expand_interleave): Likewise.
18033 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18034 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18035 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18036 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18037 (rs6000_expand_vec_perm_const): Delete.
18038 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18039 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18040 (altivec_expand_vec_perm_const_le): Take each operand individually.
18041 Operate on constant selectors rather than rtxes.
18042 (altivec_expand_vec_perm_const): Likewise. Update call to
18043 altivec_expand_vec_perm_const_le.
18044 (rs6000_expand_vec_perm_const): Delete.
18045 (rs6000_vectorize_vec_perm_const_ok): Delete.
18046 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18047 reference to the SPE evmerge intructions.
18048 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18049 an element count and rtx array.
18050 (rs6000_expand_extract_even): Update call accordingly.
18051 (rs6000_expand_interleave): Likewise.
18052 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18053 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18055 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18059 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18060 vector mode and that that mode matches the mode of the data
18062 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18063 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18064 directly using expand_vec_perm_1 when forcing selectors into
18066 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18068 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18070 * optabs-query.h (can_vec_perm_p): Delete.
18071 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18072 * optabs-query.c (can_vec_perm_p): Split into...
18073 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18074 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18075 particular selector is valid.
18076 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18077 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18078 (vect_grouped_load_supported): Likewise.
18079 (vect_shift_permute_load_chain): Likewise.
18080 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18081 (vect_transform_slp_perm_load): Likewise.
18082 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18083 (vectorizable_bswap): Likewise.
18084 (vect_gen_perm_mask_checked): Likewise.
18085 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18086 implementations of variable permutation vectors into account
18087 when deciding which selector to use.
18088 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18089 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18090 with a false third argument.
18091 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18092 to test whether the constant selector is valid and can_vec_perm_var_p
18093 to test whether a variable selector is valid.
18095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18097 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18098 * optabs-query.c (can_vec_perm_p): Likewise.
18099 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18100 instead of vec_perm_indices.
18101 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18102 (vect_gen_perm_mask_checked): Likewise,
18103 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18104 (vect_gen_perm_mask_checked): Likewise,
18106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18108 * optabs-query.h (qimode_for_vec_perm): Declare.
18109 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18110 (qimode_for_vec_perm): ...this new function.
18111 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18113 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18115 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18116 does not have a conditional at the top.
18118 2018-01-02 Richard Biener <rguenther@suse.de>
18120 * ipa-inline.c (big_speedup_p): Fix expression.
18122 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18125 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18128 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18132 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18133 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18134 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18135 cond_taken_branch_cost 3->4.
18137 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18139 PR tree-optimization/83581
18140 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18141 TODO_cleanup_cfg if any changes have been made.
18143 PR middle-end/83608
18144 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18145 convert_modes if target mode has the right side, but different mode
18148 PR middle-end/83609
18149 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18150 last argument when extracting from CONCAT. If either from_real or
18151 from_imag is NULL, use expansion through memory. If result is not
18152 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18153 the parts directly to inner mode, if even that fails, use expansion
18156 PR middle-end/83623
18157 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18158 check for bswap in mode rather than HImode and use that in expand_unop
18161 Copyright (C) 2018 Free Software Foundation, Inc.
18163 Copying and distribution of this file, with or without modification,
18164 are permitted in any medium without royalty provided the copyright
18165 notice and this notice are preserved.