1 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
4 statement if id->reset_location is true.
5 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6 id->reset_location is true.
7 (copy_phis_for_bb): Force input_location on the arguments if
8 id->reset_location is true.
9 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10 is set on the function to be inlined.
11 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
12 prevent_decl_creation_for_types fields up and add reset_location field.
14 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
16 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
18 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
20 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
22 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
23 hard registers for the clobbered pseudo.
25 2018-06-27 Paul Koning <ni1d@arrl.net>
27 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
28 mutually exclusive options.
29 * config/pdp11/constraints.md (h): New constraint.
30 (O): Update definition to match shift code generation.
32 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
34 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
36 (output_jump): Change arguments.
37 (pdp11_fixed_cc_regs): New function.
38 (pdp11_cc_mode): Ditto.
39 (pdp11_expand_shift): Ditto.
40 (pdp11_assemble_shift): Ditto.
41 (pdp11_small_shift): Ditto.
42 (pdp11_branch_cost): Remove.
43 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
45 (pdp11_register_move_cost): Update for CC registers.
46 (pdp11_rtx_costs): Add case for LSHIFTRT.
47 (pdp11_output_jump): Add CCNZ mode conditional branches.
48 (notice_update_cc_on_set): Remove.
49 (pdp11_cc_mode): New function.
50 (simple_memory_operand): Correct pre/post decrement case.
51 (no_side_effect_operand): New function.
52 (pdp11_regno_reg_class): Add CC_REGS class.
53 (pdp11_fixed_cc_regs): New function.
54 (pdp11_small_shift): New function.
55 (pdp11_expand_shift): New function to expand shift insns.
56 (pdp11_assemble_shift): New function to output shifts.
57 (pdp11_branch_cost): Remove.
58 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
59 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
61 (PTRDIFF_TYPE): Ditto.
62 (ADJUST_INSN_LENGTH): New macro.
63 (FIXED_REGISTERS): Add CC registers.
64 (CALL_USED_REGISTERS): Ditto.
66 (REG_CLASS_NAMES): Ditto.
67 (REG_CLASS_CONTENTS): Ditto.
68 (SELECT_CC_MODE): Use new function.
69 (TARGET_FLAGS_REGNUM): New macro.
70 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
71 (cc0_reg_rtx): Remove.
72 (CC_STATUS_MDEP): Remove.
73 (CC_STATUS_MDEFP_INIT): Remove.
75 (NOTICE_UPDATE_CC): Remove.
76 (REGISTER_NAMES): Add CC registers.
77 (BRANCH_COST): Change to constant 1.
78 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
80 * config/pdp11/pdp11.opt (mbcopy): Remove.
81 (mbcopy-builtin): Remove.
82 (mbranch-cheap): Remove.
83 (mbranch-expensive): Remove.
84 * config/pdp11/predicates.md (expand_shift_operand): Update to
85 match shift code generation.
86 (ccnz_operator): New predicate.
87 * doc/invoke.texi (PDP-11 Options): Remove deleted options
88 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
89 Remove non-existent option -mabshi, -mno-abshi. Document mutually
91 * doc/md.texi (PDP-11): Document new D and h constraints. Update
92 description of O constraint.
94 2018-06-27 Jeff Law <law@redhat.com>
95 Austin Law <austinklaw@gmail.com>
97 * config/v850/v850.md (addsi3_set_flags): New pattern.
98 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
99 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
100 (zero_extendhisi2_v850_set_flags): Likewise.
101 (zero_extendqisi2_v850_set_flags): Likewise.
102 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
103 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
104 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
106 * config/v850/v850-protos.h (notice_update_cc): Remove.
107 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
108 (v850_print_operand): Handle 'D' and "d".
109 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
110 Add handling of arithmetic/logical operations compared against zero.
111 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
112 Do not look at v850_compare_op, instead get mode from last argument.
113 (v850_gen_compare): Remove
114 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
115 after reload for prologue insns.
116 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
118 (construct_save_jarl): Likewise.
119 (TARGET_FLAGS_REGNUM): Define.
120 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
121 (NOTICE_UPDATE_CC): Remove.
122 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
123 than cc0. Conditionalize on reload_completed.
124 (cmpsi_insn, setfcc_insn): Likewise.
125 (tst1 splitter): Turn into define_and_split which sets the flags
127 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
128 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
129 (cstoresf4, cstoredf4): Clobber the flags.
130 (cmpsi, cmpsf, cmpdf): Remove expanders.
131 (setf_insn): Remove pattern.
132 (addsi3): Turn into define_and_split which clobbers the flags after
133 reload and a suitable pattern (addsi3_clobber_flags) for use after
135 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
136 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
137 (ashrsi3, ashrsi3_v850e2): Likewise.
138 (bins): Clobber the flags.
139 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
140 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
141 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
142 (call_value_internal_short, call_value_internal_long): Likewise.
143 (callt_save_interrupt, callt_return_interrupt): Likewise.
144 (save_interrupt, return_interrupt): Likewise.
145 (callt_save_all_interrupt, save_all_interrupt): Likewise.
146 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
147 (restore_all_interrupt, _restore_all_interrupt): Likewise.
148 (All FP comparisons): Only allow after reload has completed.
150 (divh, divhu): Tweak output template.
151 (branch_z_normal, branch_z_invert): Remove
152 (branch_nz_normal, branch_nz_invert): Likewise.
153 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
155 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
156 * config/v850/v850.c (notice_update_cc): Remove.
157 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
158 (CC_NO_CARRY): Likewise.
159 (NOTICE_UPDATE_CC): Define to nothing.
160 * config/v850/v850.md: Remove block comment on cc0 handling
161 Remove "cc" attribute from all patterns. Remove cc_status handling
162 from all patterns. Minor formatting fixes.
164 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
166 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
167 (cortex-a76.cortex-a55): Likewise.
168 * config/aarch64/aarch64-tune.md: Regenerate.
169 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
170 cortex-a76.cortex-a55.
172 2018-06-27 Jeff Law <law@redhat.com>
174 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
175 (MULTILIB_DIRNAMES): Similarly.
177 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
179 * gimple.h (gimple_return_retbnd): Delete.
180 (gimple_return_set_retbnd): Likewise.
181 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
182 gimple_return_set_retbnd.
183 * gimple-pretty-print.c (dump_gimple_return): Remove call to
184 gimple_return_retbnd and adjust.
185 * tree-inline.h (struct copy_body_data): Remove retbnd field.
186 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
187 Explicitly return NULL in a couple more cases. Move assertion
188 on debug statements and remove unreachable code.
189 (reset_debug_binding): Do not test id->retbnd.
190 (expand_call_inline): Do not set it.
192 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
194 * configure.ac: Add --disable-gcov option.
195 * configure: Regenerate.
196 * Makefile.in: Honour @enable_gcov@.
197 * doc/install.texi: Document --disable-gcov.
199 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
201 * config/arm/arm-cpus.in (cortex-a76): New entry.
202 (cortex-a76.cortex-a55): Likewise.
203 * config/arm/arm-tables.opt: Regenerate.
204 * config/arm/arm-tune.md: Likewise.
205 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
206 * doc/invoke.texi (ARM Options): Document cortex-a76 and
207 cortex-a76.cortex-a55.
209 2018-06-27 Tamar Christina <tamar.christina@arm.com>
212 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
214 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
216 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
218 (EPILOGUE_USES): Likewise.
220 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
222 * tree-inline.c (remap_location): New function extracted from...
223 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
224 (copy_phis_for_bb): ...here. Call remap_location.
225 (copy_cfg_body): Adjust call to copy_edges_for_bb.
227 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
229 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
230 unaligned vsx for 16B memset.
232 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
235 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
236 ieee128_float_type_node to long_double_type_node unless
237 TARGET_LONG_DOUBLE_128 is set.
239 2018-06-26 David Malcolm <dmalcolm@redhat.com>
241 * cfgloop.c (get_loop_location): Convert return type from
242 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
243 by implicit construction from rtx_insn *, and using
244 dump_user_location_t::from_function_decl for the fallback case.
245 * cfgloop.h (get_loop_location): Convert return type from
246 location_t to dump_user_location_t.
247 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
248 dump_printf_loc to pass in a dump_location_t rather than a
249 location_t, via the gimple stmt.
250 * coverage.c (get_coverage_counts): Update calls to
251 dump_printf_loc to pass in dump_location_t rather than a
253 * doc/optinfo.texi (Dump types): Convert example of
254 dump_printf_loc from taking "locus" to taking "insn". Update
255 description of the "_loc" calls to cover dump_location_t.
256 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
258 (dump_user_location_t::dump_user_location_t): New constructors,
259 from gimple *stmt and rtx_insn *.
260 (dump_user_location_t::from_function_decl): New function.
261 (dump_loc): Make static.
262 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
263 const dump_location_t &.
264 (dump_generic_expr_loc): Delete.
265 (dump_printf_loc): Convert param "loc" from location_t to
266 const dump_location_t &.
267 (selftest::test_impl_location): New function.
268 (selftest::dumpfile_c_tests): New function.
269 * dumpfile.h: Include "profile-count.h".
270 (class dump_user_location_t): New class.
271 (struct dump_impl_location_t): New struct.
272 (class dump_location_t): New class.
273 (dump_printf_loc): Convert 2nd param from source_location to
274 const dump_location_t &.
275 (dump_generic_expr_loc): Delete.
276 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
277 const dump_location_t &.
278 * gimple-fold.c (fold_gimple_assign): Update call to
279 dump_printf_loc to pass in a dump_location_t rather than a
280 location_t, via the gimple stmt.
281 (gimple_fold_call): Likewise.
282 * gimple-loop-interchange.cc
283 (loop_cand::analyze_iloop_reduction_var): Update for change to
284 check_reduction_path.
285 (tree_loop_interchange::interchange): Update for change to
287 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
288 change in return-type of find_loop_location.
289 (graphite_regenerate_ast_isl): Likewise.
290 * graphite-optimize-isl.c (optimize_isl): Likewise.
291 * graphite.c (graphite_transform_loops): Likewise.
292 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
293 pass in a dump_location_t rather than a location_t, via the
295 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
296 * ipa.c (walk_polymorphic_call_targets): Likewise.
297 * loop-unroll.c (report_unroll): Convert "locus" param from
298 location_t to dump_location_t.
299 (decide_unrolling): Update for change to get_loop_location's
301 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
302 location_t to dump_user_location_t.
303 (grid_find_single_omp_among_assignments_1): Updates calls to
304 dump_printf_loc to pass in a dump_location_t rather than a
305 location_t, via the gimple stmt.
306 (grid_parallel_clauses_gridifiable): Convert "tloc" from
307 location_t to dump_location_t. Updates calls to dump_printf_loc
308 to pass in a dump_location_t rather than a location_t, via the
310 (grid_inner_loop_gridifiable_p): Likewise.
311 (grid_dist_follows_simple_pattern): Likewise.
312 (grid_gfor_follows_tiling_pattern): Likewise.
313 (grid_target_follows_gridifiable_pattern): Likewise.
314 (grid_attempt_target_gridification): Convert initialization
315 of local "grid" from memset to zero-initialization; FIXME: does
316 this require C++11? Update call to dump_printf_loc to pass in a
317 optinfo_location rather than a location_t, via the gimple stmt.
318 * profile.c (read_profile_edge_counts): Updates call to
319 dump_printf_loc to pass in a dump_location_t rather than a
321 (compute_branch_probabilities): Likewise.
322 * selftest-run-tests.c (selftest::run_tests): Call
324 * selftest.h (dumpfile_c_tests): New decl.
325 * tree-loop-distribution.c (pass_loop_distribution::execute):
326 Update for change in return type of find_loop_location.
327 * tree-parloops.c (parallelize_loops): Likewise.
328 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
329 "locus" from location_t to dump_user_location_t.
330 (canonicalize_loop_induction_variables): Likewise.
331 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
332 for change in return type of find_loop_location.
333 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
334 to dump_printf_loc to pass in a dump_location_t rather than a
335 location_t, via the stmt.
336 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
338 * tree-vect-loop-manip.c (find_loop_location): Convert return
339 type from source_location to dump_user_location_t.
340 (vect_do_peeling): Update for above change.
341 (vect_loop_versioning): Update for change in type of
343 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
344 from location_t to dump_user_location_t.
345 (vect_estimate_min_profitable_iters): Update for change in type
347 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
348 location_t to dump_location_t.
349 (vect_slp_bb): Update for change in type of vect_location.
350 * tree-vectorizer.c (vect_location): Convert from source_location
351 to dump_user_location_t.
352 (try_vectorize_loop_1): Update for change in vect_location's type.
353 (vectorize_loops): Likewise.
354 (increase_alignment): Likewise.
355 * tree-vectorizer.h (vect_location): Convert from source_location
356 to dump_user_location_t.
357 (find_loop_location): Convert return type from source_location to
358 dump_user_location_t.
359 (check_reduction_path): Convert 1st param from location_t to
360 dump_user_location_t.
361 * value-prof.c (check_counter): Update call to dump_printf_loc to
362 pass in a dump_user_location_t rather than a location_t; update
363 call to error_at for change in type of "locus".
364 (check_ic_target): Update call to dump_printf_loc to
365 pass in a dump_user_location_t rather than a location_t, via the
368 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
370 * config/s390/s390.h (enum processor_flags): Do not use
371 default tune parameter when -march was specified.
373 2018-06-26 Jakub Jelinek <jakub@redhat.com>
376 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
377 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
380 2018-06-26 Richard Biener <rguenther@suse.de>
382 PR tree-optimization/86287
384 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
385 (vect_analyze_loop): Initialize n_stmts.
387 2018-06-26 Richard Biener <rguenther@suse.de>
390 * fold-const.c (fold_convertible_p): Pointer extension
393 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
396 * dwarf2out.c (loc_list_has_views): Adjust comments.
397 (dw_loc_list): Split single cross-partition range with
400 2018-06-25 Jeff Law <law@redhat.com>
402 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
403 on -mbig-switch by default.
405 * config/v850/predicates.md (const_float_1_operand): Fix match_code
407 (const_float_0_operand): Remove unused predicate.
408 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
409 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
410 (recipsf2): New expander. Original pattern now called
412 (recipdf2, recipdf2_insn): Similarly.
413 (rsqrtsf2, rsqrtsf2_insn): Similarly
414 (rsqrtdf2, rsqrtdf2_insn): Similarly
416 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
418 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
419 Simplify logic for FreeBSD (twice).
421 2018-06-25 Martin Sebor <msebor@redhat.com>
423 PR tree-optimization/86204
424 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
425 a strnlen result if it's less than the length of the string.
427 2018-06-25 Martin Sebor <msebor@redhat.com>
429 PR tree-optimization/85700
430 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
431 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
432 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
434 2018-06-25 Martin Sebor <msebor@redhat.com>
436 * doc/extend.texi (Zero-length arrays): Update and clarify.
438 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
440 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
441 added IEEE/IBM long double multilib support on PowerPC little
442 endian Linux systems.
443 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
444 (MULTILIB_DEFAULTS): Likewise.
445 * config/rs6000/rs6000.c (rs6000_option_override_internal):
447 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
448 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
449 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
451 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
454 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
455 (REORDER_45): Likewise.
457 2018-06-25 Jeff Law <law@redhat.com>
459 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
460 dividend to 32 bits. Adjust length.
461 (udivmodhi4): Cleanup output template. Fix length.
463 2018-06-25 Carl Love <cel@us.ibm.com>
465 * config/rs6000/vsx.md: Change word selector to prefered location.
467 2018-06-25 Richard Biener <rguenther@suse.de>
469 PR tree-optimization/86304
470 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
471 epilogue-if-converted loops as well.
473 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
475 * lto-section-out.c (lto_begin_section): Do not print section
476 name for noaddr and unnumbered dumps.
478 2018-06-25 Richard Biener <rguenther@suse.de>
480 * tree-vectorizer.h (struct vec_info_shared): New structure
481 with parts split out from struct vec_info and loop_nest from
482 struct _loop_vec_info.
483 (struct vec_info): Adjust accordingly.
484 (struct _loop_vec_info): Likewise.
485 (LOOP_VINFO_LOOP_NEST): Adjust.
486 (LOOP_VINFO_DATAREFS): Likewise.
487 (LOOP_VINFO_DDRS): Likewise.
488 (struct _bb_vec_info): Likewise.
489 (BB_VINFO_DATAREFS): Likewise.
490 (BB_VINFO_DDRS): Likewise.
491 (struct _stmt_vec_info): Add dr_aux member.
492 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
493 (DR_MISALIGNMENT_UNINITIALIZED): New.
494 (set_dr_misalignment): Adjust.
495 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
496 (vect_analyze_loop): Adjust prototype.
497 (vect_analyze_loop_form): Likewise.
498 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
499 Compute dependences lazily.
500 (vect_record_base_alignments): Use shared datarefs/ddrs.
501 (vect_verify_datarefs_alignment): Likewise.
502 (vect_analyze_data_refs_alignment): Likewise.
503 (vect_analyze_data_ref_accesses): Likewise.
504 (vect_analyze_data_refs): Likewise.
505 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
506 constructor parameter for shared part.
507 (vect_analyze_loop_form): Pass in shared part and adjust.
508 (vect_analyze_loop_2): Pass in storage for the number of
509 stmts. Move loop nest finding to the caller. Compute
511 (vect_analyze_loop): Pass in shared part.
512 (vect_transform_loop): Verify shared datarefs are unchanged.
513 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
514 constructor parameter for shared part.
515 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
516 (vect_slp_bb): Verify shared datarefs are unchanged before
518 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
520 (new_stmt_vec_info): Initialize DR_AUX misalignment to
521 DR_MISALIGNMENT_UNINITIALIZED.
522 * tree-vectorizer.c (vec_info::vec_info): Add constructor
523 parameter for shared part.
524 (vec_info::~vec_info): Adjust.
525 (vec_info_shared::vec_info_shared): New.
526 (vec_info_shared::~vec_info_shared): Likewise.
527 (vec_info_shared::save_datarefs): Likewise.
528 (vec_info_shared::check_datarefs): Likewise.
529 (try_vectorize_loop_1): Construct shared part live for analyses
530 of a single loop for multiple vector sizes.
531 * tree-parloops.c (gather_scalar_reductions): Adjust.
533 2018-06-25 Richard Biener <rguenther@suse.de>
535 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
536 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
537 (vect_analyze_data_refs): Remove similar code from here and
538 simplify accordingly.
540 2018-06-25 Richard Biener <rguenther@suse.de>
542 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
543 for reverse storage order accesses rather than asserting
544 they cannot happen here.
546 2018-06-25 Tom de Vries <tdevries@suse.de>
549 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
550 Use data16 instead of .byte for insn prefix.
552 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
555 * parser.c (make_char_string_pack): Pass this literal chars
556 through cpp_interpret_string.
557 (cp_parser_userdef_numeric_literal): Check the result of
558 make_char_string_pack.
560 2018-06-24 Maya Rashish <coypu@sdf.org>
562 * ginclude/stddef.h: Simplify conditions around avoiding
563 re-definition of __size_t.
565 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
567 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
568 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
570 2018-06-22 Maya Rashish <coypu@sdf.org>
572 * doc/invoke.texi (mno-fancy-math-387): Update for changes
573 made to OpenBSD and NetBSD through the years.
575 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
577 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
578 behavior of vec_pack (vector double, vector double) to match
579 behavior of vec_float2 (vector double, vector double).
581 2018-06-22 Olivier Hainque <hainque@adacore.com>
583 * gimplify.c (gimplify_function_tree): Prevent creation
584 of a trampoline for the address of the current function
585 passed to entry/exit instrumentation hooks.
587 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
590 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
593 2018-06-22 Martin Liska <mliska@suse.cz>
595 PR tree-optimization/86263
596 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
597 Bail out if is_enabled is false.
598 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
600 (jump_table_cluster::is_enabled): New function.
602 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
604 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
605 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
606 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
607 (lto_input_ts_binfo_tree_pointers): Likewise.
608 * tree-streamer-out.c (streamer_write_tree_bitfields,
609 write_ts_binfo_tree_pointers): Likewise.
610 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
612 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
614 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
616 2018-06-22 Martin Liska <mliska@suse.cz>
618 * symbol-summary.h (get): Make it pure and inline move
619 functionality from ::get function.
620 (get): Remove and inline into ::get and ::get_create.
621 (get_create): Move code from ::get function.
623 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
626 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
627 -x assembler-with-cpp.
629 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
631 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
632 _FILE_OFFSET_BITS=64 for C++.
634 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
636 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
637 conversion insn that shows up when pr85657-3.c is compiled using
638 IEEE 128-bit long double.
639 (neg<mode>2_internal): Use the correct mode to check whether the
640 mode is IBM extended.
641 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
642 multiply and divide external functions from being created more
645 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
647 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
649 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
650 the edge can be forwarded.
651 (cfg_layout_merge_blocks): Likewise.
653 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
655 * except.c (finish_eh_generation): Commit edge insertions only after
656 the EH edges have been redirected from post-landing to landing pads.
658 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
660 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
661 create_tmp_var_for to create the FRAME decl.
662 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
664 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
666 * tree-inline.c (copy_edges_for_bb): Minor tweak.
667 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
668 debug statement when resetting its value.
669 (expand_call_inline): Copy the locus of the call onto the assignment
670 of the return value, if any. Use local variable in more cases.
672 2018-06-21 Martin Liska <mliska@suse.cz>
674 * ipa-pure-const.c (propagate_nothrow): Use
675 funct_state_summaries->get.
676 (dump_malloc_lattice): Likewise.
677 (propagate_malloc): Likewise.
679 2018-06-21 Richard Biener <rguenther@suse.de>
681 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
682 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
683 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
685 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
686 BLOCK_ABSTRACT_ORIGIN unconditionally.
688 2018-06-21 David Malcolm <dmalcolm@redhat.com>
690 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
692 * ipa-reference.c (ipa_reference_c_finalize): Delete
693 ipa_ref_opt_sum_summaries and set it to NULL.
695 2018-06-21 Tom de Vries <tdevries@suse.de>
697 PR tree-optimization/85859
698 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
699 test with comment from bb_no_side_effects_p.
701 2018-06-21 Richard Biener <rguenther@suse.de>
703 PR tree-optimization/86232
704 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
705 max for constant niter.
707 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
709 * config/aarch64/aarch64-simd.md
710 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
712 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
714 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
715 Make opernads of the unspec commutative.
717 2018-06-21 Richard Biener <rguenther@suse.de>
719 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
720 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
721 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
722 (vect_analyze_data_ref_dependence): Re-order checks to deal with
724 (vect_record_base_alignments): Do not record base alignment
725 for gathers or scatters.
726 (vect_compute_data_ref_alignment): Drop return value that is always
727 true. Bail out early for gathers or scatters.
728 (vect_enhance_data_refs_alignment): Bail out early for gathers
730 (vect_find_same_alignment_drs): Likewise.
731 (vect_analyze_data_refs_alignment): Remove dead code.
732 (vect_slp_analyze_and_verify_node_alignment): Likewise.
733 (vect_analyze_data_refs): For possible gathers or scatters do
734 not create an alternate DR, just check their possible validity
735 and mark them. Adjust DECL_NONALIASED handling to not rely
737 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
738 update inits of gathers or scatters.
739 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
740 Also copy gather/scatter flag to pattern vinfo.
742 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
744 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
745 behavior of vec_packsu (vector unsigned long long, vector unsigned
746 long long) to match behavior of vec_packs with same signature.
748 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
749 Thomas Schwinge <thomas@codesourcery.com>
750 Cesar Philippidis <cesar@codesourcery.com>
752 * gimplify.c (gimplify_scan_omp_clauses): Add support for
753 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
754 (gimplify_adjust_omp_clauses): Likewise.
755 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
756 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
757 (gimplify_omp_target_update): Update handling of acc update and
759 * omp-low.c (install_var_field): Remove unused parameter
760 base_pointers_restrict.
761 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
762 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
764 (omp_target_base_pointers_restrict_p): Delete.
765 (scan_omp_target): Update call to scan_sharing_clauses.
766 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
768 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
769 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
770 (convert_local_omp_clauses): Likewise.
771 * tree-pretty-print.c (dump_omp_clause): Likewise.
772 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
774 (omp_clause_code_name): Likewise.
776 2018-06-20 Jakub Jelinek <jakub@redhat.com>
779 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
782 PR tree-optimization/86231
783 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
784 anti-range don't overwrite *vr0min before using it to compute *vr0max.
786 2018-06-20 Tom de Vries <tdevries@suse.de>
788 PR tree-optimization/86097
789 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
790 iv type if signedness of iv type is not the same as that of *nit.
792 2018-06-20 Jakub Jelinek <jakub@redhat.com>
794 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
795 EDGE_EH edges, verify they are all EDGE_EH.
797 2018-06-20 Maya Rashish <coypu@sdf.org>
799 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
801 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
803 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
804 * config/aarch64/aarch64.c (xgene1_tunings): Add
805 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
806 (aarch64_mode_valid_for_sched_fusion_p):
808 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
809 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
811 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
813 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
814 * config/aarch64/iterators.md (VQ2): New mode iterator.
816 2018-06-20 Martin Liska <mliska@suse.cz>
818 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
819 Change default ratio from 10 to 8.
821 2018-06-20 Martin Liska <mliska@suse.cz>
823 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
825 (bit_test_cluster::find_bit_tests): Likewise.
826 (switch_decision_tree::analyze_switch_statement): Find clusters.
827 * tree-switch-conversion.h (struct jump_table_cluster): Document
830 2018-06-20 Martin Liska <mliska@suse.cz>
832 * tree-switch-conversion.c (switch_conversion::collect):
833 Record m_uniq property.
834 (switch_conversion::expand): Bail out for special conditions.
835 (group_cluster::~group_cluster): New.
836 (group_cluster::group_cluster): Likewise.
837 (group_cluster::dump): Likewise.
838 (jump_table_cluster::emit): New.
839 (switch_decision_tree::fix_phi_operands_for_edges): New.
840 (struct case_node): Remove struct.
841 (jump_table_cluster::can_be_handled): New.
842 (case_values_threshold): Moved to header.
843 (reset_out_edges_aux): Likewise.
844 (jump_table_cluster::is_beneficial): New.
845 (bit_test_cluster::can_be_handled): Likewise.
846 (add_case_node): Remove.
847 (bit_test_cluster::is_beneficial): New.
848 (case_bit_test::cmp): New.
849 (bit_test_cluster::emit): New.
850 (expand_switch_as_decision_tree_p): Remove.
851 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
852 (fix_phi_operands_for_edge): Likewise.
853 (switch_decision_tree::analyze_switch_statement): New.
854 (compute_cases_per_edge): Move ...
855 (switch_decision_tree::compute_cases_per_edge): ... here.
856 (try_switch_expansion): Likewise.
857 (switch_decision_tree::try_switch_expansion): Likewise.
858 (record_phi_operand_mapping): Likewise.
859 (switch_decision_tree::record_phi_operand_mapping): Likewise.
860 (emit_case_decision_tree): Likewise.
861 (switch_decision_tree::emit): Likewise.
862 (balance_case_nodes): Likewise.
863 (switch_decision_tree::balance_case_nodes): Likewise.
864 (dump_case_nodes): Likewise.
865 (switch_decision_tree::dump_case_nodes): Likewise.
866 (emit_jump): Likewise.
867 (switch_decision_tree::emit_jump): Likewise.
868 (emit_cmp_and_jump_insns): Likewise.
869 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
870 (emit_case_nodes): Likewise.
871 (switch_decision_tree::emit_case_nodes): Likewise.
872 (conditional_probability): Remove.
873 * tree-switch-conversion.h (enum cluster_type): New.
875 (struct cluster): Likewise.
876 (cluster::cluster): Likewise.
877 (struct simple_cluster): Likewise.
878 (simple_cluster::simple_cluster): Likewise.
879 (struct group_cluster): Likewise.
880 (struct jump_table_cluster): Likewise.
881 (struct bit_test_cluster): Likewise.
882 (struct min_cluster_item): Likewise.
883 (struct case_tree_node): Likewise.
884 (case_tree_node::case_tree_node): Likewise.
885 (jump_table_cluster::case_values_threshold): Likewise.
886 (struct case_bit_test): Likewise.
887 (struct switch_decision_tree): Likewise.
888 (struct switch_conversion): Likewise.
889 (switch_decision_tree::reset_out_edges_aux): Likewise.
891 2018-06-20 Martin Liska <mliska@suse.cz>
893 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
894 (hoist_edge_and_branch_if_true): Likewise.
895 (expand_switch_using_bit_tests_p): Likewise.
896 (struct case_bit_test): Likewise.
897 (case_bit_test_cmp): Likewise.
898 (emit_case_bit_tests): Likewise.
899 (switch_conversion::switch_conversion): New class.
900 (struct switch_conv_info): Remove old struct.
901 (collect_switch_conv_info): More to ...
902 (switch_conversion::collect): ... this.
903 (check_range): Likewise.
904 (switch_conversion::check_range): Likewise.
905 (check_all_empty_except_final): Likewise.
906 (switch_conversion::check_all_empty_except_final): Likewise.
907 (check_final_bb): Likewise.
908 (switch_conversion::check_final_bb): Likewise.
909 (create_temp_arrays): Likewise.
910 (switch_conversion::create_temp_arrays): Likewise.
911 (free_temp_arrays): Likewise.
912 (gather_default_values): Likewise.
913 (switch_conversion::gather_default_values): Likewise.
914 (build_constructors): Likewise.
915 (switch_conversion::build_constructors): Likewise.
916 (constructor_contains_same_values_p): Likewise.
917 (switch_conversion::contains_same_values_p): Likewise.
918 (array_value_type): Likewise.
919 (switch_conversion::array_value_type): Likewise.
920 (build_one_array): Likewise.
921 (switch_conversion::build_one_array): Likewise.
922 (build_arrays): Likewise.
923 (switch_conversion::build_arrays): Likewise.
924 (gen_def_assigns): Likewise.
925 (switch_conversion::gen_def_assigns): Likewise.
926 (prune_bbs): Likewise.
927 (switch_conversion::prune_bbs): Likewise.
928 (fix_phi_nodes): Likewise.
929 (switch_conversion::fix_phi_nodes): Likewise.
930 (gen_inbound_check): Likewise.
931 (switch_conversion::gen_inbound_check): Likewise.
932 (process_switch): Use the newly created class.
933 (switch_conversion::expand): New.
934 (switch_conversion::~switch_conversion): New.
935 * tree-switch-conversion.h: New file.
937 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
939 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
940 tree-vect-patterns.c.
941 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
942 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
943 (vect_recog_sad_pattern): Likewise.
944 (vect_recog_widen_sum_pattern): Likewise.
945 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
946 (vect_recog_widen_shift_pattern): Remove the type_in argument.
947 (vect_recog_rotate_pattern): Likewise.
948 (vect_recog_mult_pattern): Likewise.
949 (vect_recog_vector_vector_shift_pattern): Likewise.
950 (vect_recog_divmod_pattern): Likewise.
951 (vect_recog_mixed_size_cond_pattern): Likewise.
952 (vect_recog_bool_pattern): Likewise.
953 (vect_recog_mask_conversion_pattern): Likewise.
954 (vect_try_gather_scatter_pattern): Likewise.
955 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
956 (vect_recog_over_widening_pattern): Likewise.
957 (vect_recog_gather_scatter_pattern): Likewise.
958 (vect_recog_func_ptr): Move from tree-vectorizer.h
959 (vect_vect_recog_func_ptrs): Move further down the file.
960 (vect_recog_func): Likewise. Remove the third argument.
961 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
962 (vect_pattern_recog_1): Expect the pattern function to do any
963 necessary target tests. Also expect it to provide a vector type.
964 Remove the type_in handling.
966 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
968 * tree-vect-patterns.c (vect_pattern_detected): New function.
969 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
970 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
971 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
972 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
973 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
974 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
975 (vect_recog_mask_conversion_pattern)
976 (vect_try_gather_scatter_pattern): Likewise.
978 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
980 * tree-vect-patterns.c (vect_get_internal_def): New function.
981 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
982 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
983 (search_type_for_mask_1): Use it.
985 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
987 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
988 redundant WIDEN_SUM_EXPR handling.
989 (vect_recog_sad_pattern): Likewise.
991 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
993 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
994 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
995 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
996 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
997 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
999 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1001 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1002 use the stmt_vec_info of the original bb statement for the
1003 new zero assignment, even if the call is part of a pattern.
1005 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1007 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1008 that the sequence is attached to the original statement rather
1009 than the pattern statement.
1010 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1011 PATTERN_DEF_SEQ from the original statement rather than
1012 the main pattern statement.
1013 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1014 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1015 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1017 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1019 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1020 definition statements before the early exit for statements that aren't
1022 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1024 (vect_transform_loop): ...here. Process pattern definition
1025 statements without first checking whether the main pattern
1026 statement is live or relevant.
1028 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1030 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1031 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1033 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1035 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1036 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1037 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1038 (expand_block_compare): Change select_block_compare_mode call.
1039 (expand_strncmp_align_check): Use new functions, fix comment.
1040 (emit_final_str_compare_gpr): New function.
1041 (expand_strn_compare): Refactor and clean up code.
1042 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1044 2018-06-19 Tony Reix <tony.reix@atos.com>
1045 Damien Bergamini <damien.bergamini@atos.com>
1046 David Edelsohn <dje.gcc@gmail.com>
1048 * collect2.c (static_obj): New variable.
1049 (static_libs): New variable.
1050 (is_in_list): Uncomment declaration.
1051 (main): Track AIX libraries linked statically.
1052 (is_in_list): Uncomment definition.
1053 (scan_prog_file): Don't add AIX shared libraries initializer
1054 to constructor list if linking statically.
1056 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1058 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1060 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1062 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1064 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1067 2018-06-19 Martin Liska <mliska@suse.cz>
1069 * config/i386/i386.c (ix86_can_inline_p): Do not use
1070 ipa_fn_summaries::get_create.
1071 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1073 (devirtualization_time_bonus): Likewise.
1074 (ipcp_propagate_stage): Likewise.
1075 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1076 (edge_set_predicate): Likewise.
1077 (evaluate_conditions_for_known_args): Likewise.
1078 (evaluate_properties_for_edge): Likewise.
1079 (ipa_call_summary::reset): Tranform to ...
1080 (ipa_call_summary::~ipa_call_summary): ... this.
1081 (ipa_fn_summary::reset): Transform to ...
1082 (ipa_fn_summary::~ipa_fn_summary): ... this.
1083 (ipa_fn_summary_t::remove): Rename to ...
1084 (ipa_fn_summary_t::remove_callees): ... this.
1085 (ipa_fn_summary_t::duplicate): Use placement new
1086 instead of memory copy.
1087 (ipa_call_summary_t::duplicate): Likewise.
1088 (ipa_call_summary_t::remove): Remove.
1089 (dump_ipa_call_summary): Change get_create to get.
1090 (ipa_dump_fn_summary): Dump only when summary exists.
1091 (analyze_function_body): Use symbol_summary::get instead
1093 (compute_fn_summary): Likewise.
1094 (estimate_edge_devirt_benefit): Likewise.
1095 (estimate_edge_size_and_time): Likewise.
1096 (inline_update_callee_summaries): Likewise.
1097 (remap_edge_change_prob): Likewise.
1098 (remap_edge_summaries): Likewise.
1099 (ipa_merge_fn_summary_after_inlining): Likewise.
1100 (write_ipa_call_summary): Likewise.
1101 (ipa_fn_summary_write): Likewise.
1102 (ipa_free_fn_summary): Likewise.
1103 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1104 (struct ipa_call_summary): Likewise.
1105 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1107 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1108 (estimate_size_after_inlining): Likewise.
1109 (estimate_growth): Likewise.
1110 (growth_likely_positive): Likewise.
1111 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1112 (inline_call): Likewise.
1113 * ipa-inline.c (caller_growth_limits): Likewise.
1114 (can_inline_edge_p): Likewise.
1115 (can_inline_edge_by_limits_p): Likewise.
1116 (compute_uninlined_call_time): Likewise.
1117 (compute_inlined_call_time): Likewise.
1118 (want_inline_small_function_p): Likewise.
1119 (edge_badness): Likewise.
1120 (update_caller_keys): Likewise.
1121 (update_callee_keys): Likewise.
1122 (inline_small_functions): Likewise.
1123 (inline_to_all_callers_1): Likewise.
1124 (dump_overall_stats): Likewise.
1125 (early_inline_small_functions): Likewise.
1126 (early_inliner): Likewise.
1127 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1128 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1129 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1130 * ipa-split.c (execute_split_functions): Likewise.
1131 * symbol-summary.h: Likewise.
1132 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1134 2018-06-19 Richard Biener <rguenther@suse.de>
1136 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1137 (vectorize_loops): ... here. Fix dbgcnt handling.
1138 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1140 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1143 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1144 ieee128 argument takes up only one (vector) register, not two (floating
1147 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1149 * gimplify.c (gimplify_init_constructor): Really never clear for an
1150 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1152 2018-06-19 Richard Biener <rguenther@suse.de>
1154 PR tree-optimization/86179
1155 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1156 after failed recognition.
1158 2018-06-18 Martin Sebor <msebor@redhat.com>
1161 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1162 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1163 Handle integer subtraction.
1164 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1165 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1167 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1169 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1170 param from rtx to rtx_insn *.
1171 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1173 (frv_ifcvt_modify_insn): Likwise.
1174 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1175 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1176 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1177 as_a <rtx_insn *> cast to local "unprotected_region" once
1178 it's been established that it's not NULL or pc_rtx.
1179 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1180 param "sethi" from rtx to rtx_insn *.
1181 (nds32_group_float_insns): Likewise for param "insn".
1182 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1184 (vax_output_int_subtract): Likewise.
1185 * config/vax/vax.c (vax_output_int_add): Likewise for param
1187 (vax_output_int_subtract): Likewise.
1188 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1189 (emit_pattern_after): Likewise for param "after".
1190 (emit_insn_after): Likewise.
1191 (emit_jump_insn_after): Likewise.
1192 (emit_call_insn_after): Likewise.
1193 (emit_debug_insn_after): Likewise.
1194 (emit_pattern_before): Likewise for param "before".
1195 (emit_insn_before): Likewise.
1196 (emit_jump_insn_before): Likewise.
1197 * final.c (get_insn_template): Likewise for param "insn", removing
1199 * output.h (get_insn_template): Likewise for 2nd param.
1200 * rtl.h (emit_insn_before): Likewise.
1201 (emit_jump_insn_before): Likewise.
1202 (emit_debug_insn_before_noloc): Likewise.
1203 (emit_insn_after): Likewise.
1204 (emit_jump_insn_after): Likewise.
1205 (emit_call_insn_after): Likewise.
1206 (emit_debug_insn_after): Likewise.
1207 (set_insn_deleted): Likewise for param.
1209 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1212 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1213 floating point modes, so that IFmode is numerically greater than
1214 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1215 to declare the ordering. This prevents IFmode from being
1216 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1217 machine. Include rs6000-modes.h to share the fractional values
1218 between genmodes* and the rest of the compiler.
1222 * config/rs6000/rs6000-modes.h: New file.
1223 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1224 meaning of rs6000_long_double_size so that 126..128 selects an
1225 appropriate 128-bit floating point type.
1226 (rs6000_option_override_internal): Likewise.
1227 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1228 (TARGET_LONG_DOUBLE_128): Change the meaning of
1229 rs6000_long_double_size so that 126..128 selects an appropriate
1230 128-bit floating point type.
1231 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1232 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1233 source and destination to match the standard usage.
1234 (truncifkf2): Likewise.
1235 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1236 ISA 2.07 to use an explicit clobber, instead of passing in a
1238 (copysign<mode>3_soft): Likewise.
1240 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1242 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1243 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1244 (vect_slp_analyze_instance_dependence): Likewise.
1245 (vect_enhance_data_refs_alignment): Likewise.
1246 (vect_analyze_data_refs_alignment): Likewise.
1247 (vect_slp_analyze_and_verify_instance_alignment
1248 (vect_analyze_data_ref_accesses): Likewise.
1249 (vect_prune_runtime_alias_test_list): Likewise.
1250 (vect_analyze_data_refs): Likewise.
1251 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1252 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1253 (vect_analyze_scalar_cycles_1): Likewise.
1254 (vect_get_loop_niters): Likewise.
1255 (vect_analyze_loop_form_1): Likewise.
1256 (vect_update_vf_for_slp): Likewise.
1257 (vect_analyze_loop_operations): Likewise.
1258 (vect_analyze_loop): Likewise.
1259 (vectorizable_induction): Likewise.
1260 (vect_transform_loop): Likewise.
1261 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1262 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1263 (vect_make_slp_decision): Likewise.
1264 (vect_detect_hybrid_slp): Likewise.
1265 (vect_slp_analyze_operations): Likewise.
1266 (vect_slp_bb): Likewise.
1267 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1268 (vectorizable_bswap): Likewise.
1269 (vectorizable_call): Likewise.
1270 (vectorizable_simd_clone_call): Likewise.
1271 (vectorizable_conversion): Likewise.
1272 (vectorizable_assignment): Likewise.
1273 (vectorizable_shift): Likewise.
1274 (vectorizable_operation): Likewise.
1275 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1277 2018-06-18 Martin Sebor <msebor@redhat.com>
1279 PR tree-optimization/81384
1280 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1281 * builtins.c (expand_builtin_strnlen): New function.
1282 (expand_builtin): Call it.
1283 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1284 * builtins.def (BUILT_IN_STRNLEN): New.
1285 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1286 Warn for bounds in excess of maximum object size.
1287 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1288 single-value ranges. Handle strnlen.
1289 (handle_builtin_strlen): Handle strnlen.
1290 (strlen_check_and_optimize_stmt): Same.
1291 * doc/extend.texi (Other Builtins): Document strnlen.
1293 2018-06-18 Maya Rashish <coypu@sdf.org>
1295 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1296 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1297 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1299 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1301 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1303 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1305 * tree.c (escaped_string::escape): Replace cast to char * by
1306 const_cast<char *> (unescaped).
1308 2018-06-18 Nick Clifton <nickc@redhat.com>
1311 * tree.c (escaped_string): New class. Converts an unescaped
1312 string into its escaped equivalent.
1313 (warn_deprecated_use): Use the new class to convert the
1314 deprecation message, if present.
1315 (test_escaped_strings): New self test.
1316 (test_c_tests): Add test_escaped_strings.
1317 * doc/extend.texi (deprecated): Add a note that the
1318 deprecation message is affected by the -fmessage-length
1319 option, and that control characters will be escaped.
1320 (#pragma GCC error): Document this pragma.
1321 (#pragma GCC warning): Likewise.
1322 * doc/invoke.texi (-fmessage-length): Document this option's
1323 effect on the #warning and #error preprocessor directives and
1324 the deprecated attribute.
1326 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1328 * tree.c (decl_value_expr_lookup): Revert latest change.
1329 (decl_value_expr_insert): Likewise.
1331 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1333 * gimplify.c (nonlocal_vlas): Delete.
1334 (nonlocal_vla_vars): Likewise.
1335 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1337 (gimplify_body): Do not create and destroy nonlocal_vlas.
1338 * tree-nested.c: Include diagnostic.h.
1339 (use_pointer_in_frame): Tweak.
1340 (lookup_field_for_decl): Add assertion and declare the transformation.
1341 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1342 internal error when the reference is in a wrong context. Do not
1343 create a debug decl by default.
1344 (note_nonlocal_block_vlas): Delete.
1345 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1346 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1347 create a debug decl by default.
1348 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1349 call is in a wrong context.
1350 (fixup_vla_decls): New function.
1351 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1352 debug variables were created.
1353 * tree.c (decl_value_expr_lookup): Add checking assertion.
1354 (decl_value_expr_insert): Likewise.
1356 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1359 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1360 * tree-scalar-evolution.c (interpret_expr): Likewise.
1361 (expression_expensive_p): Likewise.
1362 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1363 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1364 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1365 (ssa_defined_by_minus_one_stmt_p): New.
1367 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1370 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1371 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1372 * dojump.c (do_jump): Likewise.
1373 * expr.c (expand_expr_real_2): Check operand type's sign.
1374 * fold-const.c (const_unop): Handle ABSU_EXPR.
1375 (fold_abs_const): Likewise.
1376 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1377 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1378 (strip_sign_op_1): Likesise.
1379 * match.pd: Add new pattern to generate ABSU_EXPR.
1380 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1381 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1382 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1383 * tree-inline.c (estimate_operator_cost): Likewise.
1384 * tree-pretty-print.c (dump_generic_node): Likewise.
1385 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1386 * tree.def (ABSU_EXPR): New.
1388 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1391 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1392 documented as preserved for backward compatibility only.
1393 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1395 PR rtl-optimization/86108
1396 * bb-reorder.c (create_forwarder_block): Renamed to ...
1397 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1398 jump from new landing pad to the second part.
1399 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1402 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1405 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1406 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1407 Only call store_expr for halves if the mode is the same.
1410 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1411 Fix up comment formatting.
1413 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1415 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1416 (typed_splay_tree::closure,
1417 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1418 (typed_splay_tree::typed_splay_tree,
1419 typed_splay_tree::operator =): Declared private.
1420 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1421 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1422 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1423 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1424 typed_splay_tree::splay_tree_splay,
1425 typed_splay_tree::splay_tree_foreach_helper,
1426 typed_splay_tree::splay_tree_insert,
1427 typed_splay_tree::splay_tree_remove,
1428 typed_splay_tree::splay_tree_lookup,
1429 typed_splay_tree::splay_tree_predecessor,
1430 typed_splay_tree::splay_tree_successor,
1431 typed_splay_tree::splay_tree_min,
1432 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1433 (typed_splay_tree::root, typed_splay_tree::comp,
1434 typed_splay_tree::delete_key,
1435 typed_splay_tree::delete_value): New data members.
1436 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1437 typed_splay_tree::remove.
1439 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1441 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1442 -mginv and -mno-ginv to the assembler.
1443 * config/mips/mips.opt (-mcrc): New option.
1445 * doc/invoke.text (-mcrc): Document.
1448 2018-06-15 Nick Clifton <nickc@redhat.com>
1451 * tree.c (escaped_string): New class. Converts an unescaped
1452 string into its escaped equivalent.
1453 (warn_deprecated_use): Use the new class to convert the
1454 deprecation message, if present.
1455 (test_escaped_strings): New self test.
1456 (test_c_tests): Add test_escaped_strings.
1457 * doc/extend.texi (deprecated): Add a note that the
1458 deprecation message is affected by the -fmessage-length
1459 option, and that control characters will be escaped.
1460 (#pragma GCC error): Document this pragma.
1461 (#pragma GCC warning): Likewise.
1462 * doc/invoke.texi (-fmessage-length): Document this option's
1463 effect on the #warning and #error preprocessor directives and
1464 the deprecated attribute.
1466 2018-06-15 Richard Biener <rguenther@suse.de>
1468 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1469 here, also noting vector size used.
1470 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1471 size used in MSG_OPTIMIZED_LOCATIONS dump.
1472 (pass_slp_vectorize::execute): Adjust.
1474 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1477 * config/arc/arc.c (arc_return_address_register): Fix
1480 2018-06-15 Richard Biener <rguenther@suse.de>
1483 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1484 leave useless conversion stripping to force_gimple_operand_gsi.
1485 (gimplify_build2): Likewise.
1486 (gimplify_build1): Likewise.
1488 2018-06-15 Richard Biener <rguenther@suse.de>
1491 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1492 before adjusting their block.
1494 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1496 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1497 multilibs for *-*-rtems*.
1498 * config/riscv/t-rtems: New file.
1500 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1503 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1504 unsigned_type_for returns NULL.
1507 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1508 subregs of multi-word pseudos unless the float mode has word size.
1510 2018-06-14 Richard Biener <rguenther@suse.de>
1513 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1514 duplicate and harmful type_hash_canon.
1515 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1517 2018-06-14 Richard Biener <rguenther@suse.de>
1520 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1523 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1525 * config/rtems.h (STDINT_LONG32): Define.
1527 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1528 Prachi Godbole <prachi.godbole@imgtec.com>
1530 * config/mips/mips-cpus.def: Define P6600.
1531 * config/mips/mips-tables.opt: Regenerate.
1532 * config/mips/mips.c (mips_ucbranch_type): New enum.
1533 (mips_rtx_cost_data): Add support for P6600.
1534 (mips_issue_rate): Likewise.
1535 (mips_multipass_dfa_lookahead): Likewise.
1536 (mips_avoid_hazard): Likewise.
1537 (mips_reorg_process_insns): Likewise.
1538 (mips_classify_branch_p6600): New function.
1539 * config/mips/mips.h (TUNE_P6600): New define.
1540 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1541 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1542 * config/mips/mips.md: Include p6600.md.
1543 (processor): Add p6600.
1544 * config/mips/p6600.md: New file.
1545 * doc/invoke.texi: Add p6600 to supported architectures.
1547 2018-06-13 Martin Sebor <msebor@redhat.com>
1549 PR tree-optimization/86114
1550 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1552 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1554 2018-06-13 Richard Biener <rguenther@suse.de>
1556 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1557 Properly set vector type of the intermediate stmt.
1558 * tree-vect-stmts.c (vectorizable_operation): The destination
1559 var always has vectype_out type.
1561 2018-06-13 Jeff Law <law@redhat.com>
1563 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1564 integer 0 for argument to print_rtl_with_bb.
1565 (rl78_reorg): Likewise.
1567 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1569 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1570 from rtx to rtx_insn *.
1571 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1573 (add_sched_insns_for_speculation): Likewise for local "target",
1574 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1575 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1576 from rtx_insn ** to rtx_code_label **.
1577 (reorg_emit_nops): Likewise.
1578 (c6x_reorg): Likewise for local "call_labels".
1579 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1581 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1582 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1583 the loops over LABEL_REFS.
1584 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1586 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1587 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1588 (split_branches): Strengthen local "olabel" from rtx to
1589 rtx_insn *, adding a safe_as_a cast.
1590 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1592 (add_insn_after): Likewise for first two params.
1593 (add_insn_before): Likewise.
1594 (remove_insn): Likewise for param.
1595 (emit_pattern_before_noloc): Likewise for second and third params.
1596 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1597 (emit_call_insn_before_noloc): Likewise.
1598 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1600 (emit_barrier_before): Likewise.
1601 (emit_label_before): Strengthen "label" param from "rtx" to
1602 "rtx_code_label *". Strengthen "before" param from "rtx" to
1604 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1606 (emit_pattern_after_noloc): Likewise.
1607 (emit_insn_after_noloc): Likewise.
1608 (emit_jump_insn_after_noloc): Likewise.
1609 (emit_call_insn_after_noloc): Likewise.
1610 (emit_debug_insn_after_noloc): Likewise.
1611 (emit_barrier_after): Likewise.
1612 (emit_label_after): Likewise for both params.
1613 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1614 "loc" param from "int" to "location_t".
1615 (emit_insn_after_setloc): Likewise.
1616 (emit_jump_insn_after_setloc): Likewise.
1617 (emit_call_insn_after_setloc): Likewise.
1618 (emit_debug_insn_after_setloc): Likewise.
1619 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1620 "loc" param from "int" to "location_t".
1621 (emit_pattern_before): Convert NULL_RTX to NULL.
1622 (emit_insn_before_setloc): Convert "loc" param from "int" to
1624 (emit_jump_insn_before_setloc): Likewise.
1625 (emit_call_insn_before_setloc): Likewise.
1626 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1627 rtx_insn *. Convert "loc" param from "int" to "location_t".
1628 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1629 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1630 Convert 3rd param from "int" to "location_t".
1631 (emit_barrier_before, emit_barrier_after, next_real_insn):
1632 Strengthen param from rtx to rtx_insn *.
1633 (emit_label_before): Strengthen 1st param from "rtx" to
1634 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1636 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1637 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1638 Strengthen 2nd param from "rtx" to "rtx_insn *".
1639 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1640 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1641 Likewise. Convert 3rd param from "int" to "location_t".
1642 (emit_label_after): Strengthen 1st param from "rtx" to
1644 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1646 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1647 from "rtx" to "rtx_insn *".
1649 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1651 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1652 bodies streamed in with -Q.
1653 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1654 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1655 * gimple-streamer-out.c: Include gimple-pretty-print.h
1656 (output_bb): Dump stmts streamed.
1657 * lto-section-out.c: Include print-tree.h
1658 (lto_begin_section): Dump sections created.
1659 (lto_output_decl_index): Dump decl encoded.
1660 * lto-streamer-out.c: Include print-tree.h
1661 (create_output_block): Dump output block created.
1662 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1663 (output_function): Dump function output.
1664 (output_constructor): Dump constructor streamed.
1665 (write_global_stream): Output indexes encoded.
1666 (produce_asm_for_decls): Dump streams encoded.
1667 * lto-streamer.c (streamer_dump_file): New global var.
1668 * lto-streamer.h (streamer_dump_file): Declare.
1669 * passes.c (ipa_write_summaries): Initialize streamer dump.
1670 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1673 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1676 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1677 offsets for register save directives. Emit a second batch of save
1678 directives, if need be, when the function accesses prior frames.
1680 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1682 * config/arc/fpu.md (fmasf4): Force operand to register.
1683 (fnmasf4): Likewise.
1685 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1687 * config/arc/arc-protos.h (arc_pad_return): Remove.
1688 * config/arc/arc.c (machine_function): Remove force_short_suffix
1690 (arc_print_operand): Adjust printing of '&'.
1691 (arc_verify_short): Remove conditional printing of short suffix.
1692 (arc_final_prescan_insn): Remove reference to size_reason.
1693 (pad_return): New function.
1694 (arc_reorg): Call pad_return.
1695 (arc_pad_return): Remove.
1696 (arc_init_machine_status): Remove reference to force_short_suffix.
1697 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1698 (attr length): When attribute iscompact is true force to 2
1699 regardless; in the case of maybe check if we want to force the
1700 instruction to have 4 bytes length.
1701 (nopv): Change it to generate 4 byte long nop as well.
1702 (blockage): New pattern.
1703 (simple_return): Remove call to arc_pad_return.
1704 (p_return_i): Likewise.
1706 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1708 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1710 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1712 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1715 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1717 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1718 for ARC700 and ARCv2.
1720 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1723 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1724 operands[2] instead of operands[1].
1727 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1729 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1730 case, check whether the outer register overlaps an unallocatable
1731 register, not just whether it fits the required class.
1733 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1735 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1737 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1738 (memrefs_conflict_p): Likewise.
1739 (init_alias_analysis): Likewise.
1740 * cfgexpand.c (expand_debug_expr): Likewise.
1741 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1742 * cse.c (fold_rtx): Likewise.
1743 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1744 * expr.c (emit_block_move_hints): Likewise.
1745 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1746 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1747 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1748 (emit_group_store): Likewise.
1749 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1750 to read the PRE/POST_MODIFY increment.
1751 * calls.c (store_one_arg): Use strip_offset.
1752 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1754 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1756 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1757 (simplify_binary_operation_1): Extend CONST_INT handling to
1759 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1760 than a HOST_WIDE_INT.
1761 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1763 (adjust_mems, add_stores): Update accodingly.
1764 (vt_canonicalize_addr): Track polynomial offsets.
1765 (emit_note_insn_var_location): Likewise.
1766 (vt_add_function_parameter): Likewise.
1767 (vt_initialize): Likewise.
1769 2018-06-12 Jeff Law <law@redhat.com>
1771 * config.gcc (alpha*-*-freebsd*): Remove.
1772 * config/alpha/freebsd.h: Remove.
1774 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1777 * spellcheck-tree.c (levenshtein_distance): Rename to...
1778 (get_edit_distance): ...this, and update for underlying renaming.
1779 * spellcheck-tree.h (levenshtein_distance): Rename to...
1780 (get_edit_distance): ...this.
1781 * spellcheck.c (levenshtein_distance): Rename to...
1782 (get_edit_distance): ...this. Convert from Levenshtein distance
1783 to Damerau-Levenshtein distance by supporting transpositions of
1784 adjacent characters. Rename "v1" to "v_next" and "v0" to
1786 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1787 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1788 update for underlying renaming.
1789 (selftest::levenshtein_distance_unit_test): Rename to...
1790 (selftest::test_get_edit_distance_unit): ...this, and update for
1791 underlying renaming.
1792 (selftest::test_find_closest_string): Add example from PR 69968
1793 where transposition helps
1794 (selftest::test_metric_conditions): Update for renaming.
1795 (selftest::test_metric_conditions): Likewise.
1796 (selftest::spellcheck_c_tests): Likewise.
1797 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1798 (get_edit_distance): ...this.
1799 (best_match::consider): Update for renaming.
1801 2018-06-12 Martin Sebor <msebor@redhat.com>
1803 PR tree-optimization/85259
1804 * builtins.c (compute_objsize): Handle constant offsets.
1805 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1806 true iff a warning has been issued.
1807 * gimple.h (gimple_nonartificial_location): New function.
1808 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1809 gimple_nonartificial_location and handle -Wno-system-headers.
1810 (handle_builtin_stxncpy): Same.
1812 2018-06-12 Martin Sebor <msebor@redhat.com>
1815 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1817 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1819 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1820 BUILTIN_VEC_XST entries for pointer to double and long long.
1822 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1825 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1827 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1830 2018-06-12 Martin Liska <mliska@suse.cz>
1832 * doc/options.texi: Document IntegerRange.
1834 2018-06-12 Martin Liska <mliska@suse.cz>
1836 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1837 * opt-functions.awk: Handle Deprecated flag.
1838 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1840 (read_cmdline_option): Report warning for a deprecated option.
1841 * opts.h (struct cl_option): Add new field cl_deprecated.
1842 (CL_ERR_DEPRECATED): New.
1844 2018-06-12 Martin Liska <mliska@suse.cz>
1846 * doc/options.texi: Document Deprecated option flag.
1848 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1850 * config/arc/arc-arch.h (arc_extras): New enum.
1851 (arc_cpu_t):Add field extra.
1852 (arc_cpu_types): Consider the extras.
1853 * config/arc/arc-cpus.def: Add extras info.
1854 * config/arc/arc-opts.h (processor_type): Consider extra field.
1855 * config/arc/arc.c (arc_override_options): Handle extra field.
1857 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1859 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1860 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1862 * config/arc/arc.c (arc_init): Update pic warning.
1863 (irq_range): Update irq range parsing warnings.
1864 (arc_override_options): Update various warning messages.
1865 (arc_handle_aux_attribute): Likewise.
1867 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1869 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1871 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1873 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1874 compiler messages directives.
1876 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1878 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1879 * config/mips/mips-tables.opt: Regenerate.
1880 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1882 * doc/invoke.texi: Document -march=i6500.
1884 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1886 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1887 (i6400_gpmul): Add cpu_unit.
1888 (i6400_gpdiv): Likewise.
1889 (i6400_msa_add_d): Update reservations.
1890 (i6400_msa_int_add) Likewise.
1891 (i6400_msa_short_logic3) Likewise.
1892 (i6400_msa_short_logic2) Likewise.
1893 (i6400_msa_short_logic) Likewise.
1894 (i6400_msa_move) Likewise.
1895 (i6400_msa_cmp) Likewise.
1896 (i6400_msa_short_float2) Likewise.
1897 (i6400_msa_div_d) Likewise.
1898 (i6400_msa_long_logic1) Likewise.
1899 (i6400_msa_long_logic2) Likewise.
1900 (i6400_msa_mult) Likewise.
1901 (i6400_msa_long_float2) Likewise.
1902 (i6400_msa_long_float4) Likewise.
1903 (i6400_msa_long_float5) Likewise.
1904 (i6400_msa_long_float8) Likewise.
1905 (i6400_fpu_fadd): Include frint type.
1906 (i6400_fpu_store): New define_insn_reservation.
1907 (i6400_fpu_load): Likewise.
1908 (i6400_fpu_move): Likewise.
1909 (i6400_fpu_fcmp): Likewise.
1910 (i6400_fpu_fmadd): Likewise.
1911 (i6400_int_mult): Include imul3nc type and update reservation.
1912 (i6400_int_div): Include idiv3 type and update reservation.
1913 (i6400_int_load): Update to check type not move_type.
1914 (i6400_int_store): Likewise.
1915 (i6400_int_prefetch): Set zero latency.
1917 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1919 * gcc.c: Document new %@{...} sequence.
1920 (LINK_COMMAND_SPEC): Use it for the -L switches.
1921 (cpp_unique_options): Use it for the -I switches.
1922 (at_file_argbuf): New global variable.
1923 (in_at_file): Likewise.
1924 (alloc_args): Create at_file_argbuf.
1925 (clear_args): Truncate at_file_argbuf.
1926 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1927 (open_at_file): New function.
1928 (close_at_file): Likewise.
1929 (create_at_file): Delete.
1930 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1933 (validate_switches_from_spec): Deal with %@{...} sequence.
1934 (validate_switches): Likewise.
1935 (driver::finalize): Call clear_args.
1937 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1939 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1941 2018-06-11 Martin Sebor <msebor@redhat.com>
1943 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1944 (Wno-class-memaccess): Add @opindex.
1945 (Wno-templates, Wno-multiple-inheritance): Same.
1946 (Wno-virtual-inheritance, Wno-namespaces): Same.
1947 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1948 (Wno-format-overflow, Wno-format-truncation): Same.
1949 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1950 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1951 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1952 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1953 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1954 (Wno-misspelled-isr): Same.
1956 2018-06-11 Martin Sebor <msebor@redhat.com>
1958 * PR tree-optimization/86083
1959 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1961 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1963 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1965 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1968 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1969 on the correct operand.
1970 (*movdi_internal64): Ditto.
1972 2018-06-11 Martin Liska <mliska@suse.cz>
1974 PR tree-optimization/86089
1975 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1977 2018-06-11 Julia Koval <julia.koval@intel.com>
1979 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1980 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1981 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1983 2018-06-11 Olivier Hainque <hainque@adacore.com>
1985 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1986 for Ada with strict dwarf2.
1988 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1991 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1994 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1996 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1998 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2000 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2003 2018-06-08 Martin Liska <mliska@suse.cz>
2005 * tree-cfg.h (debug_function): Fix argument type to match
2008 2018-06-08 Martin Liska <mliska@suse.cz>
2010 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2011 Remove usage of MPX-related (and removed) fields.
2012 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2014 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2016 * cfg.c (debug): Use TDF_NONE rather than 0.
2017 * cfghooks.c (debug): Likewise.
2018 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2019 (struct dump_option_value_info): Convert to...
2020 (struct kv_pair): ...this template type.
2021 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2023 (optinfo_verbosity_options): Likewise.
2024 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2026 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2027 than int for "optgroup_flags" param.
2028 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2030 (dump_dec): Likewise.
2031 (dump_finish): Use TDF_NONE rather than 0.
2032 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2033 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2034 than 0. Update for change to option_ptr.
2035 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2036 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2037 0. Update for changes to optinfo_verbosity_options and
2039 (opt_info_switch_p): Convert optgroup_flags from int to
2041 (dump_basic_block): Use dump_flags_t rather than int
2042 for "dump_kind" param.
2043 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2044 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2045 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2046 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2047 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2048 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2049 TDF_NONE): Convert from macros to...
2050 (enum dump_flag): ...this new enum.
2051 (dump_flags_t): Update to use enum.
2052 (operator|, operator&, operator~, operator|=, operator&=):
2053 Implement for dump_flags_t.
2054 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2055 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2056 Convert from macros to...
2057 (enum optgroup_flag): ...this new enum.
2058 (optgroup_flags_t): New typedef.
2059 (operator|, operator|=): Implement for optgroup_flags_t.
2060 (struct dump_file_info): Convert field "alt_flags" to
2061 dump_flags_t. Convert field "optgroup_flags" to
2063 (dump_basic_block): Use dump_flags_t rather than int for param.
2064 (dump_generic_expr_loc): Likewise.
2065 (dump_dec): Likewise.
2066 (dump_register): Convert param "optgroup_flags" to
2068 (opt_info_enable_passes): Likewise.
2069 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2071 * gimple-pretty-print.c (debug): Likewise.
2072 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2073 (merged_store_group::apply_stores): Likewise.
2074 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2075 * gimple.c (verify_gimple_pp): Likewise.
2076 * graphite-poly.c (print_pbb_body): Likewise.
2077 * passes.c (pass_manager::register_one_dump_file): Convert
2078 local "optgroup_flags" to optgroup_flags_t.
2079 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2081 (debug_body): Likewise.
2082 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2083 to optgroup_flags_t.
2084 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2086 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2087 (convert_mult_to_fma): Likewise.
2088 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2089 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2090 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2091 "dump_kind" to dump_flags_t.
2093 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2095 * config/rs6000/rs6000.c (min, max): Delete.
2097 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2099 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2100 -mabi=spe and -mabi=no-spe.
2102 2018-06-08 Martin Liska <mliska@suse.cz>
2104 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2105 where we expect an existing summary.
2107 2018-06-08 Martin Liska <mliska@suse.cz>
2109 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2110 * ipa-inline.h (estimate_edge_growth): Likewise.
2112 2018-06-08 Martin Liska <mliska@suse.cz>
2114 * cgraph.c (function_version_hasher::hash): Use
2115 cgraph_node::get_uid ().
2116 (function_version_hasher::equal):
2117 * cgraph.h (cgraph_node::get_uid): New method.
2118 * ipa-inline.c (update_caller_keys): Use
2119 cgraph_node::get_uid ().
2120 (update_callee_keys): Likewise.
2121 * ipa-utils.c (searchc): Likewise.
2122 (ipa_reduced_postorder): Likewise.
2123 * lto-cgraph.c (input_node): Likewise.
2124 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2125 * symbol-summary.h (symtab_insertion): Likewise.
2126 (symtab_removal): Likewise.
2127 (symtab_duplication): Likewise.
2128 * tree-pretty-print.c (dump_function_header): Likewise.
2129 * tree-sra.c (convert_callers_for_node): Likewise.
2131 2018-06-08 Martin Liska <mliska@suse.cz>
2133 * cgraph.c (symbol_table::create_edge): Always assign a new
2135 (symbol_table::free_edge): Do not recycle numbers.
2136 * cgraph.h (cgraph_edge::get): New method.
2137 * symbol-summary.h (symtab_removal): Use it.
2138 (symtab_duplication): Likewise.
2139 (call_summary::hashable_uid): Remove.
2141 2018-06-08 Martin Liska <mliska@suse.cz>
2143 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2144 (initialize_growth_caches): Remove.
2145 (free_growth_caches): Likewise.
2146 (do_estimate_edge_time): Use edge_growth_cache.
2147 (do_estimate_edge_size): Likewise.
2148 (do_estimate_edge_hints): Likewise.
2149 * ipa-inline.c (reset_edge_caches): Likewise.
2150 (recursive_inlining): Likewise.
2151 (inline_small_functions): Likewise.
2152 * ipa-inline.h (initialize_growth_caches): Remove.
2153 (estimate_edge_size): Likewise.
2154 (estimate_edge_time): Likewise.
2155 (estimate_edge_hints): Likewise.
2156 (reset_edge_growth_cache): Likewise.
2157 * symbol-summary.h (call_summary::remove): New method.
2159 2018-06-08 Martin Liska <mliska@suse.cz>
2161 * ipa-cp.c (class edge_clone_summary): New summary.
2162 (grow_edge_clone_vectors): Remove.
2163 (ipcp_edge_duplication_hook): Remove.
2164 (class edge_clone_summary_t): New call_summary class.
2165 (ipcp_edge_removal_hook): Remove.
2166 (edge_clone_summary_t::duplicate): New function.
2167 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2168 (create_specialized_node): Likewise.
2169 (ipcp_driver): Initialize edge_clone_summaries and do not
2172 2018-06-08 Martin Liska <mliska@suse.cz>
2174 * symbol-summary.h (get): New function.
2175 (call_summary::m_initialize_when_cloning): New class member.
2177 2018-06-08 Martin Liska <mliska@suse.cz>
2179 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2180 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2181 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2182 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2183 (remove_cgraph_node_from_order): Use the removed_nodes set.
2184 (do_per_function_toporder): Likwise.
2185 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2186 instead of summary_uid.
2187 (symtab_removal): Likewise.
2188 (symtab_duplication): Likewise.
2190 2018-06-08 Martin Liska <mliska@suse.cz>
2192 * ipa-cp.c (ipcp_store_bits_results): Use
2193 ipcp_transformation_sum.
2194 (ipcp_store_vr_results): Likewise.
2195 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2197 (ipcp_transformation_initialize): ... this.
2198 (ipa_set_node_agg_value_chain):
2199 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2200 (write_ipcp_transformation_info): Likewise.
2201 (read_ipcp_transformation_info): Likewise.
2202 (ipcp_update_bits): Likewise.
2203 (ipcp_update_vr): Likewise.
2204 (ipcp_transform_function): Likewise.
2205 * ipa-prop.h: Rename ipcp_transformation_summary to
2206 ipcp_transformation.
2207 (class ipcp_transformation_t): New function summary.
2208 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2209 (ipa_get_agg_replacements_for_node): Likewise.
2211 2018-06-08 Martin Liska <mliska@suse.cz>
2213 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2215 (class funct_state_summary_t): New function_summary class.
2216 (has_function_state): Remove.
2217 (get_function_state): Likewise.
2218 (set_function_state): Likewise.
2219 (add_new_function): Likewise.
2220 (funct_state_summary_t::insert): New function.
2221 (duplicate_node_data): Remove.
2222 (remove_node_data): Remove.
2223 (funct_state_summary_t::duplicate): New function.
2224 (register_hooks): Create new funct_state_summaries.
2225 (pure_const_generate_summary): Use it.
2226 (pure_const_write_summary): Likewise.
2227 (pure_const_read_summary): Likewise.
2228 (propagate_pure_const): Likewise.
2229 (propagate_nothrow): Likewise.
2230 (dump_malloc_lattice): Likewise.
2231 (propagate_malloc): Likewise.
2232 (execute): Do not register hooks, just remove summary
2234 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2237 2018-06-08 Martin Liska <mliska@suse.cz>
2239 * ipa-reference.c (remove_node_data): Remove.
2240 (duplicate_node_data): Likewise.
2241 (class ipa_ref_var_info_summary_t): New class.
2242 (class ipa_ref_opt_summary_t): Likewise.
2243 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2244 (get_reference_optimization_summary): Use
2245 ipa_ref_opt_sum_summaries.
2246 (set_reference_vars_info): Remove.
2247 (set_reference_optimization_summary): Likewise.
2248 (ipa_init): Create summaries.
2249 (init_function_info): Use function summary.
2250 (ipa_ref_opt_summary_t::duplicate): New function.
2251 (ipa_ref_opt_summary_t::remove): New function.
2252 (get_read_write_all_from_node): Fix GNU coding style.
2253 (propagate): Use function summary.
2254 (write_node_summary_p): Fix GNU coding style.
2255 (stream_out_bitmap): Likewise.
2256 (ipa_reference_read_optimization_summary): Use function summary.
2257 (ipa_reference_c_finalize): Do not release hooks.
2259 2018-06-08 Martin Liska <mliska@suse.cz>
2261 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2262 (analyze_function_body): Extract multiple calls of get_create.
2263 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2264 * ipa-inline.c (recursive_inlining): Use ::get method.
2265 * ipa-inline.h (estimate_edge_growth): Likewise.
2267 2018-06-08 Martin Liska <mliska@suse.cz>
2269 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2271 (hsa_function_summary::hsa_function_summary): Use the new enum
2273 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2274 * hsa-gen.c (hsa_get_host_function): Likewise.
2275 (get_brig_function_name): Likewise.
2276 * ipa-hsa.c (process_hsa_functions): Likewise.
2277 (ipa_hsa_write_summary): Likewise.
2278 * symbol-summary.h (symtab_duplication): Use ::get function/
2279 (get): New function.
2281 2018-06-08 Martin Liska <mliska@suse.cz>
2283 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2285 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2286 (hsa_register_kernel): Likewise.
2287 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2288 * hsa-gen.c (hsa_get_host_function): Likewise.
2289 (get_brig_function_name): Likewise.
2290 (generate_hsa): Likewise.
2291 (pass_gen_hsail::execute): Likewise.
2292 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2293 (devirtualization_time_bonus): Likewise.
2294 (ipcp_propagate_stage): Likewise.
2295 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2296 (edge_set_predicate): Likewise.
2297 (evaluate_conditions_for_known_args): Likewise.
2298 (evaluate_properties_for_edge): Likewise.
2299 (ipa_fn_summary::reset): Likewise.
2300 (ipa_fn_summary_t::duplicate): Likewise.
2301 (dump_ipa_call_summary): Likewise.
2302 (ipa_dump_fn_summary): Likewise.
2303 (analyze_function_body): Likewise.
2304 (compute_fn_summary): Likewise.
2305 (estimate_edge_devirt_benefit): Likewise.
2306 (estimate_edge_size_and_time): Likewise.
2307 (estimate_calls_size_and_time): Likewise.
2308 (estimate_node_size_and_time): Likewise.
2309 (inline_update_callee_summaries): Likewise.
2310 (remap_edge_change_prob): Likewise.
2311 (remap_edge_summaries): Likewise.
2312 (ipa_merge_fn_summary_after_inlining): Likewise.
2313 (ipa_update_overall_fn_summary): Likewise.
2314 (read_ipa_call_summary): Likewise.
2315 (inline_read_section): Likewise.
2316 (write_ipa_call_summary): Likewise.
2317 (ipa_fn_summary_write): Likewise.
2318 (ipa_free_fn_summary): Likewise.
2319 * ipa-hsa.c (process_hsa_functions): Likewise.
2320 (ipa_hsa_write_summary): Likewise.
2321 (ipa_hsa_read_section): Likewise.
2322 * ipa-icf.c (sem_function::merge): Likewise.
2323 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2324 (do_estimate_edge_time): Likewise.
2325 (estimate_size_after_inlining): Likewise.
2326 (estimate_growth): Likewise.
2327 (growth_likely_positive): Likewise.
2328 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2329 (inline_call): Likewise.
2330 * ipa-inline.c (caller_growth_limits): Likewise.
2331 (can_inline_edge_p): Likewise.
2332 (can_inline_edge_by_limits_p): Likewise.
2333 (compute_uninlined_call_time): Likewise.
2334 (compute_inlined_call_time): Likewise.
2335 (want_inline_small_function_p): Likewise.
2336 (edge_badness): Likewise.
2337 (update_caller_keys): Likewise.
2338 (update_callee_keys): Likewise.
2339 (recursive_inlining): Likewise.
2340 (inline_small_functions): Likewise.
2341 (inline_to_all_callers_1): Likewise.
2342 (dump_overall_stats): Likewise.
2343 (early_inline_small_functions): Likewise.
2344 (early_inliner): Likewise.
2345 * ipa-inline.h (estimate_edge_growth): Likewise.
2346 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2347 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2348 * ipa-prop.h (IPA_NODE_REF): Likewise.
2349 (IPA_EDGE_REF): Likewise.
2350 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2351 (propagate_malloc): Likewise.
2352 * ipa-split.c (execute_split_functions): Likewise.
2353 * symbol-summary.h: Rename get to get_create.
2355 (get_create): Likewise.
2356 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2358 2018-06-08 Martin Liska <mliska@suse.cz>
2360 * symbol-summary.h (release): Move definition out of class
2362 (symtab_removal): Likewise.
2363 (symtab_duplication): Likewise.
2365 2018-06-08 Martin Liska <mliska@suse.cz>
2367 * symbol-summary.h (function_summary): Move constructor
2368 implementation out of class declaration.
2369 (release): Likewise.
2370 (symtab_insertion): Likewise.
2371 (symtab_removal): Likewise.
2372 (symtab_duplication): Likewise.
2375 2018-06-08 Martin Liska <mliska@suse.cz>
2377 * Makefile.in: Remove support for MPX (macros, related functions,
2378 fields in cgraph_node, ...).
2379 * builtin-types.def (BT_BND): Likewise.
2380 (BT_FN_BND_CONST_PTR): Likewise.
2381 (BT_FN_CONST_PTR_BND): Likewise.
2382 (BT_FN_VOID_PTR_BND): Likewise.
2383 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2384 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2385 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2386 (expand_builtin_mempcpy_with_bounds): Likewise.
2387 (expand_builtin_memset_with_bounds): Likewise.
2388 (expand_builtin_memset_args): Likewise.
2389 (std_expand_builtin_va_start): Likewise.
2390 (expand_builtin): Likewise.
2391 (expand_builtin_with_bounds): Likewise.
2392 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2393 (DEF_LIB_BUILTIN_CHKP): Likewise.
2394 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2395 (DEF_CHKP_BUILTIN): Likewise.
2396 (BUILT_IN_MEMCPY): Likewise.
2397 (BUILT_IN_MEMMOVE): Likewise.
2398 (BUILT_IN_MEMPCPY): Likewise.
2399 (BUILT_IN_MEMSET): Likewise.
2400 (BUILT_IN_STPCPY): Likewise.
2401 (BUILT_IN_STRCAT): Likewise.
2402 (BUILT_IN_STRCHR): Likewise.
2403 (BUILT_IN_STRCPY): Likewise.
2404 (BUILT_IN_STRLEN): Likewise.
2405 (BUILT_IN_MEMCPY_CHK): Likewise.
2406 (BUILT_IN_MEMMOVE_CHK): Likewise.
2407 (BUILT_IN_MEMPCPY_CHK): Likewise.
2408 (BUILT_IN_MEMSET_CHK): Likewise.
2409 (BUILT_IN_STPCPY_CHK): Likewise.
2410 (BUILT_IN_STRCAT_CHK): Likewise.
2411 (BUILT_IN_STRCPY_CHK): Likewise.
2412 * calls.c (store_bounds): Likewise.
2413 (emit_call_1): Likewise.
2414 (special_function_p): Likewise.
2415 (maybe_warn_nonstring_arg): Likewise.
2416 (initialize_argument_information): Likewise.
2417 (finalize_must_preallocate): Likewise.
2418 (compute_argument_addresses): Likewise.
2419 (expand_call): Likewise.
2420 * cfgexpand.c (expand_call_stmt): Likewise.
2421 (expand_return): Likewise.
2422 (expand_gimple_stmt_1): Likewise.
2423 (pass_expand::execute): Likewise.
2424 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2425 (cgraph_node::remove): Likewise.
2426 (cgraph_node::dump): Likewise.
2427 (cgraph_node::verify_node): Likewise.
2428 * cgraph.h (chkp_function_instrumented_p): Likewise.
2429 (symtab_node::get_alias_target): Likewise.
2430 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2431 (cgraph_local_p): Likewise.
2432 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2433 (cgraph_edge::rebuild_references): Likewise.
2434 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2435 (walk_polymorphic_call_targets): Likewise.
2436 (cgraph_node::expand_thunk): Likewise.
2437 (symbol_table::output_weakrefs): Likewise.
2438 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2439 (ix86_handle_option): Likewise.
2440 * config/i386/constraints.md: Likewise.
2441 * config/i386/i386-builtin-types.def (BND): Likewise.
2445 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2446 (BDESC_FIRST): Likewise.
2448 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2449 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2450 * config/i386/i386.c (enum reg_class): Likewise.
2451 (ix86_target_string): Likewise.
2452 (ix86_option_override_internal): Likewise.
2453 (ix86_conditional_register_usage): Likewise.
2454 (ix86_valid_target_attribute_inner_p): Likewise.
2455 (ix86_set_indirect_branch_type): Likewise.
2456 (ix86_set_current_function): Likewise.
2457 (ix86_function_arg_regno_p): Likewise.
2458 (init_cumulative_args): Likewise.
2459 (ix86_function_arg_advance): Likewise.
2460 (ix86_function_arg): Likewise.
2461 (ix86_pass_by_reference): Likewise.
2462 (ix86_function_value_regno_p): Likewise.
2463 (ix86_function_value_1): Likewise.
2464 (ix86_function_value_bounds): Likewise.
2465 (ix86_return_in_memory): Likewise.
2466 (ix86_setup_incoming_vararg_bounds): Likewise.
2467 (ix86_va_start): Likewise.
2468 (indirect_thunk_need_prefix): Likewise.
2469 (print_reg): Likewise.
2470 (ix86_print_operand): Likewise.
2471 (ix86_expand_call): Likewise.
2472 (ix86_output_function_return): Likewise.
2473 (reg_encoded_number): Likewise.
2474 (BDESC_VERIFYS): Likewise.
2475 (ix86_init_mpx_builtins): Likewise.
2476 (ix86_init_builtins): Likewise.
2477 (ix86_emit_cmove): Likewise.
2478 (ix86_emit_move_max): Likewise.
2479 (ix86_expand_builtin): Likewise.
2480 (ix86_builtin_mpx_function): Likewise.
2481 (ix86_get_arg_address_for_bt): Likewise.
2482 (ix86_load_bounds): Likewise.
2483 (ix86_store_bounds): Likewise.
2484 (ix86_load_returned_bounds): Likewise.
2485 (ix86_store_returned_bounds): Likewise.
2486 (ix86_class_likely_spilled_p): Likewise.
2487 (ix86_hard_regno_mode_ok): Likewise.
2488 (x86_order_regs_for_local_alloc): Likewise.
2489 (ix86_mitigate_rop): Likewise.
2490 (ix86_bnd_prefixed_insn_p): Likewise.
2491 (ix86_mpx_bound_mode): Likewise.
2492 (ix86_make_bounds_constant): Likewise.
2493 (ix86_initialize_bounds): Likewise.
2494 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2495 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2496 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2497 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2498 (TARGET_CHKP_BOUND_MODE): Likewise.
2499 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2500 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2501 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2502 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2503 * config/i386/i386.h (TARGET_MPX): Likewise.
2504 (TARGET_MPX_P): Likewise.
2505 (VALID_BND_REG_MODE): Likewise.
2506 (FIRST_BND_REG): Likewise.
2507 (LAST_BND_REG): Likewise.
2508 (enum reg_class): Likewise.
2509 (BND_REG_P): Likewise.
2510 (BND_REGNO_P): Likewise.
2511 (BNDmode): Likewise.
2512 (ADJUST_INSN_LENGTH): Likewise.
2513 * config/i386/i386.md: Likewise.
2514 * config/i386/i386.opt: Likewise.
2515 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2516 (defined): Likewise.
2517 (LINK_MPX): Likewise.
2518 (MPX_SPEC): Likewise.
2519 (LIBMPX_SPEC): Likewise.
2520 (LIBMPXWRAPPERS_SPEC): Likewise.
2521 (CHKP_SPEC): Likewise.
2522 * config/i386/predicates.md: Likewise.
2523 * dbxout.c (dbxout_type): Likewise.
2524 * doc/extend.texi: Likewise.
2525 * doc/invoke.texi: Likewise.
2526 * doc/md.texi: Likewise.
2527 * doc/tm.texi: Likewise.
2528 * doc/tm.texi.in: Likewise.
2529 * dwarf2out.c (is_base_type): Likewise.
2530 (gen_formal_types_die): Likewise.
2531 (gen_subprogram_die): Likewise.
2532 (gen_type_die_with_usage): Likewise.
2533 (gen_decl_die): Likewise.
2534 (dwarf2out_late_global_decl): Likewise.
2535 * expr.c (expand_assignment): Likewise.
2536 (emit_storent_insn): Likewise.
2537 (store_expr_with_bounds): Likewise.
2538 (store_expr): Likewise.
2539 (expand_expr_real_1): Likewise.
2540 * expr.h (store_expr_with_bounds): Likewise.
2541 * function.c (use_register_for_decl): Likewise.
2542 (struct bounds_parm_data): Likewise.
2543 (assign_parms_augmented_arg_list): Likewise.
2544 (assign_parm_find_entry_rtl): Likewise.
2545 (assign_parm_is_stack_parm): Likewise.
2546 (assign_parm_load_bounds): Likewise.
2547 (assign_bounds): Likewise.
2548 (assign_parms): Likewise.
2549 (expand_function_start): Likewise.
2550 * gcc.c (CHKP_SPEC): Likewise.
2551 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2552 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2553 (wrestrict_dom_walker::check_call): Likewise.
2554 * gimple.c (gimple_build_call_from_tree): Likewise.
2555 * gimple.h (enum gf_mask): Likewise.
2556 (gimple_call_with_bounds_p): Likewise.
2557 (gimple_call_set_with_bounds): Likewise.
2558 * gimplify.c (gimplify_init_constructor): Likewise.
2559 * ipa-cp.c (initialize_node_lattices): Likewise.
2560 (propagate_constants_across_call): Likewise.
2561 (find_more_scalar_values_for_callers_subset): Likewise.
2562 * ipa-hsa.c (process_hsa_functions): Likewise.
2563 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2564 * ipa-icf.c (sem_function::merge): Likewise.
2565 * ipa-inline.c (early_inliner): Likewise.
2566 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2567 (warn_function_cold): Likewise.
2568 (propagate_pure_const): Likewise.
2569 * ipa-ref.h (enum GTY): Likewise.
2570 * ipa-split.c (find_retbnd): Likewise.
2571 (consider_split): Likewise.
2572 (split_function): Likewise.
2573 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2574 * ipa.c (walk_polymorphic_call_targets): Likewise.
2575 (symbol_table::remove_unreachable_nodes): Likewise.
2576 (process_references): Likewise.
2577 (cgraph_build_static_cdtor_1): Likewise.
2578 * lto-cgraph.c (lto_output_node): Likewise.
2579 (output_refs): Likewise.
2580 (compute_ltrans_boundary): Likewise.
2581 (input_overwrite_node): Likewise.
2582 (input_node): Likewise.
2583 (input_cgraph_1): Likewise.
2584 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2585 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2586 (class pass_chkp_instrumentation_passes): Likewise.
2587 (make_pass_chkp_instrumentation_passes): Likewise.
2588 * passes.def: Likewise.
2589 * rtl.h (struct GTY): Likewise.
2590 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2591 * stor-layout.c (layout_type): Likewise.
2592 * symtab.c: Likewise.
2593 * target.def: Likewise.
2594 * targhooks.c (default_chkp_bound_type): Likewise.
2595 (default_chkp_bound_mode): Likewise.
2596 (default_builtin_chkp_function): Likewise.
2597 (default_chkp_function_value_bounds): Likewise.
2598 (default_chkp_make_bounds_constant): Likewise.
2599 (default_chkp_initialize_bounds): Likewise.
2600 * targhooks.h (default_chkp_bound_type): Likewise.
2601 (default_chkp_bound_mode): Likewise.
2602 (default_builtin_chkp_function): Likewise.
2603 (default_chkp_function_value_bounds): Likewise.
2604 (default_chkp_make_bounds_constant): Likewise.
2605 (default_chkp_initialize_bounds): Likewise.
2606 * toplev.c (compile_file): Likewise.
2607 (process_options): Likewise.
2608 * tree-core.h (DEF_BUILTIN): Likewise.
2609 (DEF_BUILTIN_CHKP): Likewise.
2610 * tree-inline.c (declare_return_variable): Likewise.
2611 (remap_gimple_stmt): Likewise.
2612 (copy_bb): Likewise.
2613 (initialize_inlined_parameters): Likewise.
2614 (expand_call_inline): Likewise.
2615 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2616 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2617 (make_pass_ipa_chkp_produce_thunks): Likewise.
2618 (make_pass_chkp): Likewise.
2619 (make_pass_chkp_opt): Likewise.
2620 (make_pass_chkp_instrumentation_passes): Likewise.
2621 * tree-pretty-print.c (dump_generic_node): Likewise.
2622 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2623 * tree-ssa-dce.c (propagate_necessity): Likewise.
2624 (eliminate_unnecessary_stmts): Likewise.
2625 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2626 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2627 * tree-ssa-sccvn.h: Likewise.
2628 * tree-ssa-strlen.c (get_string_length): Likewise.
2629 (valid_builtin_call): Likewise.
2630 (adjust_last_stmt): Likewise.
2631 (handle_builtin_strchr): Likewise.
2632 (handle_builtin_strcpy): Likewise.
2633 (handle_builtin_stxncpy): Likewise.
2634 (handle_builtin_memcpy): Likewise.
2635 (handle_builtin_strcat): Likewise.
2636 (strlen_check_and_optimize_stmt): Likewise.
2637 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2638 * tree-streamer-in.c: Likewise.
2639 * tree-streamer.c (record_common_node): Likewise.
2640 * tree.c (tree_code_size): Likewise.
2641 (wide_int_to_tree_1): Likewise.
2642 (type_contains_placeholder_1): Likewise.
2643 (build_common_tree_nodes): Likewise.
2644 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2645 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2646 (POINTER_BOUNDS_P): Likewise.
2647 (BOUNDED_TYPE_P): Likewise.
2648 (BOUNDED_P): Likewise.
2649 (CALL_WITH_BOUNDS_P): Likewise.
2650 (pointer_bounds_type_node): Likewise.
2651 * value-prof.c (gimple_ic): Likewise.
2652 * var-tracking.c (vt_add_function_parameters): Likewise.
2653 * varasm.c (make_decl_rtl): Likewise.
2654 (assemble_start_function): Likewise.
2655 (output_constant): Likewise.
2656 (maybe_assemble_visibility): Likewise.
2657 * varpool.c (ctor_for_folding): Likewise.
2658 * chkp-builtins.def: Remove.
2659 * ipa-chkp.c: Remove.
2660 * ipa-chkp.h: Remove.
2661 * rtl-chkp.c: Remove.
2662 * rtl-chkp.h: Remove.
2663 * tree-chkp-opt.c: Remove.
2664 * tree-chkp.c: Remove.
2665 * tree-chkp.h: Remove.
2667 2018-06-07 Carl Love <cel@us.ibm.com>
2669 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2670 vextract_fp_from_shortl): Add BE support.
2672 2018-06-07 Paul Koning <ni1d@arrl.net>
2674 * compare-elim.c (try_merge_compare): Don't merge compare if
2675 address contains a side effect.
2676 (try_eliminate_compare): Likewise.
2678 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2680 * config.gcc: Support "tremont".
2681 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2682 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2684 * config/i386/i386.c (m_TREMONT): Define.
2685 (processor_target_table): Add "tremont".
2686 (PTA_TREMONT): Define.
2687 (ix86_lea_outperforms): Add TARGET_TREMONT.
2688 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2689 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2690 and M_INTEL_GOLDMONT_PLUS.
2691 (fold_builtin_cpu): Add "tremont".
2692 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2693 (ix86_option_override_internal): Add "tremont".
2694 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2695 (processor_type): Add PROCESSOR_TREMONT.
2696 * config/i386/x86-tune.def: Add m_TREMONT.
2697 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2699 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2701 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2702 symbol defined for msp430i* devices to be lower case.
2704 gcc/testsuite/gcc.target/msp430/
2705 * msp430i-device-symbol.c: New test.
2706 * msp430f-device-symbol.c: New test.
2707 * msp430.h: New test header file.
2709 2018-06-07 Richard Biener <rguenther@suse.de>
2711 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2712 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2713 Properly wrap signed arithmetic if overflow wraps.
2715 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2717 PR tree-optimization/69615
2718 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2719 of a cast from a same precision integral SSA_NAME in a bb dominated
2720 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2721 cast to utype if rhs2 has already a compatible type.
2723 2018-06-07 Richard Biener <rguenther@suse.de>
2725 PR tree-optimization/85935
2726 * graphite-scop-detection.c (find_params_in_bb): Analyze
2727 condition operands with respect to the correct loop. Assert
2728 the analysis doesn't fail.
2730 2018-06-04 Carl Love <cel@us.ibm.com>
2732 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2733 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2734 as it is slightly cheaper.
2735 (first_match_or_eos_index_<mode>):
2736 Calculate index using natural element order.
2737 (first_match_index_<mode>):
2738 Calculate index using natural element order.
2739 (first_match_or_eos_index_<mode>):
2740 Calculate index using natural order.
2741 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2742 for BE and LE modes.
2743 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2744 P9V_BUILTIN_VCLZLSBB_V16QI.
2745 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2748 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2750 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2751 indentation and line wrap for many prototypes. Add missing
2752 @smallexample directives around block of prototypes for vec_xl and
2755 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2757 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2758 track if we pass or return IEEE 128-bit floating point.
2759 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2760 C++ mangling that is compatible with GCC 8.1.
2761 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2762 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2763 floating point types.
2764 (rs6000_function_arg_advance_1): Likewise.
2765 (rs6000_mangle_type): Optionally generate mangled names that match
2766 what GCC 8.1 generated for IEEE 128-bit floating point types.
2767 (rs6000_globalize_decl_name): If we have an external function that
2768 passes or returns IEEE 128-bit types, generate a weak reference
2769 from the mangled name used in GCC 8.1 to the current mangled
2771 (rs6000_init_builtins): Make __ibm128 use the long double type if
2772 long double is IBM extended double. Make __float128 use the long
2773 double type if long double is IEEE 128-bit.
2776 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2777 macro for __ibm128 built-in functions.
2778 (PACK_IF): Add __ibm128 pack/unpack functions.
2779 (UNPACK_IF): Likewise.
2780 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2781 enable long double built-in functions if long double is IEEE
2782 128-bit floating point.
2783 (rs6000_invalid_builtin): Update long double built-in function
2785 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2786 functions, adjust the built-in function to use the long double
2787 built-in function if __ibm128 and long double are the same type.
2788 * doc/extend.texi (PowerPC builtins): Update documention for
2789 __builtin_{,un}pack_longdouble. Add documentation for
2790 __builtin_{,un}pack_ibm128.
2792 2018-06-06 Jim Wilson <jimw@sifive.com>
2794 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2795 (struct machine_function): New field interrupt_mode.
2796 (riscv_handle_type_attribute): New function. Add forward declaration.
2797 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2798 (riscv_expand_epilogue): Check interrupt_mode field.
2799 (riscv_set_current_function): Check interrupt attribute args and
2800 set interrupt_mode field.
2801 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2802 (riscv_sret, riscv_uret): New.
2803 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2804 new arguments to interrupt attribute.
2806 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2809 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2810 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2812 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2814 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2815 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2816 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2817 several redundant entries.
2819 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2821 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2822 type from "rtx" to "rtx_insn *".
2823 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2824 for local "call_insn", removing cast.
2825 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2827 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2829 PR tree-optimization/86066
2830 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2831 for BIT_INSERT_EXPR stores.
2833 2018-06-06 Richard Biener <rguenther@suse.de>
2835 PR tree-optimization/86062
2836 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2837 component refs ontop
2838 of to be offsetted base.
2840 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2842 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2843 to be static and remove check on interrupt attribute name.
2845 gcc/testsuite/gcc.target/msp430/
2846 * function-attributes-4.c: New test.
2847 * static-interrupts.c: New test.
2849 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2851 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2852 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2854 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2857 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2859 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2860 Remove second argument, change how error is called.
2861 (aarch64_layout_arg): Remove second argument from
2862 aarch64_err_no_fpadvsimd call.
2863 (aarch64_init_cumulative_args): Ditto.
2864 (aarch64_gimplify_va_arg_expr): Ditto.
2865 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2867 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2869 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2870 (*simple_return_indirect_internal<mode>): Rename from
2871 simple_return_indirect_internal. Use W mode iterator.
2872 (rstorssp): New expander.
2873 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2874 (clrssbsy): New expander.
2875 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2877 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2879 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2881 (cmse_check_pointed_object): Likewise.
2883 2018-06-05 Martin Liska <mliska@suse.cz>
2885 PR gcov-profile/47618
2886 * doc/invoke.texi: Document how -fprofile-dir format
2889 2018-06-05 Richard Biener <rguenther@suse.de>
2891 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2892 removal pretend DOM info isn't available so we do not update
2893 it and only remove edges, not dominated blocks. Actually free
2894 DOM info in case we removed something. Remove unreachable blocks.
2895 (mfb_keep_latches): Work with either DOM info or marked backedges.
2896 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2897 first. Mark backedges if DOM info isn't available.
2898 (Re-)compute DOM info after cleanup_control_flow_pre.
2900 2018-06-05 Richard Biener <rguenther@suse.de>
2902 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2903 (locus_discrim_hasher::hash): Adjust.
2904 (locus_discrim_hasher::equal): Likewise.
2905 (next_discriminator_for_locus): Work on line directly.
2906 (same_line_p): Pass in expanded locus1 as well.
2907 (assign_discriminators): Avoid redundant location expansions.
2909 2018-06-05 Richard Biener <rguenther@suse.de>
2911 PR tree-optimization/86046
2912 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2913 if required after clearing TREE_ADDRESSABLE.
2915 2018-06-05 Richard Biener <rguenther@suse.de>
2917 PR tree-optimization/86047
2918 * tree-ssa-loop.c (for_each_index): Glob handling of all
2919 decls and constants and really handle all of them.
2921 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2924 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2925 qualifier_void_pointer and qualifier_const_void_pointer.
2926 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2927 (arm_init_builtins): Handle the above.
2928 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2929 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2930 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2933 2018-06-05 Martin Liska <mliska@suse.cz>
2935 * auto-profile.c (read_autofdo_file): Do not use
2936 gcov_ctr_summary struct.
2937 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2938 * coverage.c (struct counts_entry): Likewise.
2939 (read_counts_file): Read just single summary entry.
2940 (get_coverage_counts): Use gcov_summary struct.
2941 * coverage.h (get_coverage_counts): Likewise.
2942 * gcov-dump.c (dump_working_sets): Likewise.
2943 (tag_summary): Dump just single summary.
2944 * gcov-io.c (gcov_write_summary): Write just histogram
2946 (gcov_read_summary): Read just single summary.
2947 (compute_working_sets): Use gcov_summary struct.
2948 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2949 of GCOV_COUNTERS_SUMMABLE.
2950 (GCOV_COUNTERS_SUMMABLE): Remove.
2951 (GCOV_FIRST_VALUE_COUNTER): Replace with
2952 GCOV_COUNTER_V_INTERVAL.
2953 (struct gcov_ctr_summary): Remove.
2954 (struct gcov_summary): Directly use fields of former
2956 (compute_working_sets): Use gcov_summary struct.
2957 * gcov.c (read_count_file): Do not use ctrs fields.
2958 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2960 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2962 * profile.c: Likewise.
2963 * profile.h: Likewise.
2965 2018-06-05 Martin Liska <mliska@suse.cz>
2967 PR gcov-profile/84846
2968 * gcov.c (output_lines): Print working directory only
2969 in intermediate format.
2971 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2973 * config/s390/s390-builtin-types.def: Add void function type.
2974 * config/s390/s390-builtins.def: Use the function type for the
2977 2018-06-04 Jim Wilson <jimw@sifive.com>
2979 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2981 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2982 and maybe_eh_return. Change regno to unsigned int. Use new args to
2983 handle EH_RETURN_DATA_REGNO registers properly.
2984 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2985 (riscv_expand_epilogue): Update comment. Change argument name and
2986 type. Update code to use new name and type. Pass new args to
2987 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2989 * config/riscv/riscv.md (NORMAL_RETURN): New.
2990 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2991 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2992 (eh_return): Call gen_eh_return_internal and emit barrier.
2993 (eh_return_internal): Call riscv_expand_epilogue.
2995 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2997 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2998 bit_insertion field and declare can_be_merged_into method.
2999 (merged_store_group::can_be_merged_into): New method.
3000 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3001 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3002 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3004 2018-06-04 Richard Biener <rguenther@suse.de>
3006 PR tree-optimization/85955
3007 * builtins.c (fold_builtin_sincos): Convert pointers to
3008 destination to appropriate type before dereferencing.
3010 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3012 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3014 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3016 * expr.c (expand_expr_real_1): Force the operand into memory if
3017 its TYPE_MODE is BLKmode and if there is no integer mode for
3018 the number of bits being extracted.
3020 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3024 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3025 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3027 2018-06-04 Richard Biener <rguenther@suse.de>
3029 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3030 (cleanup_tree_cfg_noloop): ... single caller. Do
3031 start_recording_case_labels later.
3033 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3035 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3036 to _IMMINTRIN_H_INCLUDED.
3037 * config/i386/pconfigintrin.h: Ditto.
3038 * config/i386/waitpkgintrin.h: Ditto.
3039 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3040 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3041 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3042 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3043 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3044 waitpkgintrin.h and cldemoteintrin.h.
3046 2018-06-04 Richard Biener <rguenther@suse.de>
3048 PR tree-optimization/86038
3049 * tracer.c (find_best_successor): Check probability for
3050 being initialized, bail out if not.
3052 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3055 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3056 of bits to ignore when comparing architectures.
3058 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3060 PR tree-optimization/69615
3061 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3062 maximum or minimum of the type, try to merge it also as if
3063 range1 is + [-, x - 1] or + [x + 1, -].
3066 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3068 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3070 PR tree-optimization/86034
3071 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3072 the unsigned bitfield type in a bit insertion sequence if it does not
3073 have a larger precision than the bitfield size.
3074 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3076 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3078 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3080 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3082 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3083 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3084 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3085 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3087 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3089 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3090 Disable -fdelete-null-pointer-checks for ELF toolchain.
3092 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3093 Kito Cheng <kito.cheng@gmail.com>
3095 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3096 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3097 (nds32*-*-*): Add glibc and uclibc conditions.
3098 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3099 (TARGET_EXCEPT_UNWIND_INFO): Define.
3100 * config/nds32/elf.h: New file.
3101 * config/nds32/linux.h: New file.
3102 * config/nds32/nds32-elf.opt: New file.
3103 * config/nds32/nds32-linux.opt: New file.
3104 * config/nds32/nds32-fp-as-gp.c
3105 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3106 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3108 (nds32_asm_file_end): Ditto.
3109 (nds32_print_operand): Ditto.
3110 (nds32_insert_attributes): Ditto.
3111 (nds32_init_libfuncs): New function.
3112 (TARGET_HAVE_TLS): Define.
3113 (TARGET_INIT_LIBFUNCS): Define.
3114 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3116 (TARGET_ELF): Apply different mcmodel setting.
3117 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3118 been migrated into elf.h and linux.h files.
3119 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3120 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3121 (mcmodel): The content has been migrated into nds32-elf.opt and
3122 nds32-linux.opt files.
3123 * config/nds32/t-elf: New file.
3124 * config/nds32/t-linux: New file.
3126 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3127 Shiva Chen <shiva0217@gmail.com>
3129 * config/nds32/constants.md (unspec_volatile_element): Add
3130 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3131 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3133 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3134 (make_pass_nds32_fp_as_gp): Declare.
3135 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3137 (nds32_asm_function_end_prologue): Remove unused asm output.
3138 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3139 (nds32_asm_file_start): Output necessary fp_as_gp information.
3140 (nds32_option_override): Adjust register usage.
3141 (nds32_expand_prologue): Consider fp_as_gp situation.
3142 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3143 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3146 (simple_return): Ditto.
3147 (omit_fp_begin): Output special directive for fp_as_gp.
3148 (omit_fp_end): Output special directive for fp_as_gp.
3149 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3150 mforbid-fp-as-gp): New options.
3152 2018-06-01 Mark Wielaard <mark@klomp.org>
3154 * dwarf2out.c (dwarf2out_finish): Remove generation of
3155 DW_AT_loclists_base.
3157 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3159 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3160 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3161 (struct merged_store_group): Add bit_insertion field.
3162 (dump_char_array): Use standard hexadecimal format.
3163 (merged_store_group::merged_store_group): Set bit_insertion to false.
3164 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3165 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3166 also print the mask in the dump file.
3167 (pass_store_merging::gate): Minor tweak.
3168 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3169 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3170 stores with INTEGER_CST stores.
3171 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3172 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3173 and use it throughout. Generate bit insertion sequences if need be.
3174 (pass_store_merging::process_store): Remove redundant condition.
3175 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3177 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3179 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3180 the 128-bit floating point types. Fix function comment.
3182 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3184 * config/aarch64/aarch64-simd.md
3185 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3187 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3190 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3192 PR tree-optimization/85989
3193 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3195 (backprop::intersect_uses): Check it when deciding whether this
3196 is a backedge reference.
3197 (backprop::process_block): Add each phi to m_visited_phis
3198 after visiting it, then clear it at the end.
3200 2018-06-01 Richard Biener <rguenther@suse.de>
3202 * tree-vectorizer.h (vect_dr_stmt): New function.
3203 (vect_get_load_cost): Adjust.
3204 (vect_get_store_cost): Likewise.
3205 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3206 Use vect_dr_stmt instead of DR_SMTT.
3207 (vect_record_base_alignments): Likewise.
3208 (vect_calculate_target_alignment): Likewise.
3209 (vect_compute_data_ref_alignment): Likewise and make static.
3210 (vect_update_misalignment_for_peel): Likewise.
3211 (vect_verify_datarefs_alignment): Likewise.
3212 (vector_alignment_reachable_p): Likewise.
3213 (vect_get_data_access_cost): Likewise. Pass down
3214 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3215 (vect_get_peeling_costs_all_drs): Likewise.
3216 (vect_peeling_hash_get_lowest_cost): Likewise.
3217 (vect_enhance_data_refs_alignment): Likewise.
3218 (vect_find_same_alignment_drs): Likewise.
3219 (vect_analyze_data_refs_alignment): Likewise.
3220 (vect_analyze_group_access_1): Likewise.
3221 (vect_analyze_group_access): Likewise.
3222 (vect_analyze_data_ref_access): Likewise.
3223 (vect_analyze_data_ref_accesses): Likewise.
3224 (vect_vfa_segment_size): Likewise.
3225 (vect_small_gap_p): Likewise.
3226 (vectorizable_with_step_bound_p): Likewise.
3227 (vect_prune_runtime_alias_test_list): Likewise.
3228 (vect_analyze_data_refs): Likewise.
3229 (vect_supportable_dr_alignment): Likewise.
3230 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3231 (vect_gen_prolog_loop_niters): Likewise.
3232 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3233 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3235 (vect_recog_mask_conversion_pattern): Likewise.
3236 (vect_try_gather_scatter_pattern): Likewise.
3237 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3238 to vect_get_store_cost.
3239 (vect_get_store_cost): Get stmt_info instead of DR.
3240 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3241 (vect_get_load_cost): Get stmt_info instead of DR.
3243 2018-06-01 Richard Biener <rguenther@suse.de>
3246 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3247 vars wrapped in MEM_REFs.
3249 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3251 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3252 Fix subreg tests so that we only return a choice between
3253 GENERAL_REGS and FP_REGS if the original classes included both.
3255 2018-06-01 Richard Biener <rguenther@suse.de>
3258 * tree-ssa-structalias.c (get_function_part_constraint):
3259 Handle NULL fi->decl.
3260 (find_func_aliases_for_call): Properly handle indirect
3261 fi from direct call.
3262 (find_func_clobbers): Likewise.
3263 (ipa_pta_execute): Likewise.
3264 (create_variable_info_for): For functions that are ifunc_resolver
3265 resolve to a varinfo that contains the result of the resolver call.
3266 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3269 2018-05-31 Michael Collison <michael.collison@arm.com>
3271 * config/aarch64/aarch64.md:
3272 (*fix_to_zero_extenddfdi2): New pattern.
3273 * gcc.target/aarch64/fix_extend1.c: New testcase.
3275 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3279 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3280 and BUILT_IN_STRNCMP_EQ.
3281 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3282 BUILT_IN_STRNCMP_EQ.
3283 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3284 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3285 (gimple_fold_builtin): Likewise.
3286 * tree-ssa-strlen.c (compute_string_length): New function.
3287 (determine_min_obsize): New function.
3288 (handle_builtin_string_cmp): New function to handle calls to
3289 string compare functions.
3290 (strlen_optimize_stmt): Add handling to builtin string compare
3292 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3293 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3294 * tree.c (build_common_builtin_nodes): Add new defines of
3295 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3297 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3300 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3301 functions with naked attribute.
3303 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3305 * config/i386/sse.md (avx_vec_concat<mode>):
3306 Substitute concat_tg_mode mode attribute with xtg_mode.
3307 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3308 (concat_tg_mode): Remove mode attribute.
3310 2018-05-31 Martin Sebor <msebor@redhat.com>
3313 * calls.c (alloc_max_size): Correct a logic error/typo.
3314 Treat excessive arguments as infinite. Warn for invalid arguments.
3315 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3317 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3320 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3321 and movx for Haswell.
3323 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3324 Cesar Philippidis <cesar@codesourcery.com>
3327 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3328 when emitting error on private/firstprivate reductions.
3329 * omp-low.c (lower_omp_target): Avoid reference-type processing
3330 on pointers for firstprivate clause.
3332 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3334 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3337 * config/aarch64/aarch64-simd.md
3338 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3339 (aarch64_ld1_x3_<mode>): Likewise
3340 (aarch64_st1x2<VALLDIF:mode>): Likewise
3341 (aarch64_st1_x2_<mode>): Likewise
3342 (aarch64_st1x3<VALLDIF:mode>): Likewise
3343 (aarch64_st1_x3_<mode>): Likewise
3344 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3345 (vld1_s8_x3): Likewise.
3346 (vld1_u16_x3): Likewise.
3347 (vld1_s16_x3): Likewise.
3348 (vld1_u32_x3): Likewise.
3349 (vld1_s32_x3): Likewise.
3350 (vld1_u64_x3): Likewise.
3351 (vld1_s64_x3): Likewise.
3352 (vld1_f16_x3): Likewise.
3353 (vld1_f32_x3): Likewise.
3354 (vld1_f64_x3): Likewise.
3355 (vld1_p8_x3): Likewise.
3356 (vld1_p16_x3): Likewise.
3357 (vld1_p64_x3): Likewise.
3358 (vld1q_u8_x3): Likewise.
3359 (vld1q_s8_x3): Likewise.
3360 (vld1q_u16_x3): Likewise.
3361 (vld1q_s16_x3): Likewise.
3362 (vld1q_u32_x3): Likewise.
3363 (vld1q_s32_x3): Likewise.
3364 (vld1q_u64_x3): Likewise.
3365 (vld1q_s64_x3): Likewise.
3366 (vld1q_f16_x3): Likewise.
3367 (vld1q_f32_x3): Likewise.
3368 (vld1q_f64_x3): Likewise.
3369 (vld1q_p8_x3): Likewise.
3370 (vld1q_p16_x3): Likewise.
3371 (vld1q_p64_x3): Likewise.
3372 (vst1_s64_x2): Likewise.
3373 (vst1_u64_x2): Likewise.
3374 (vst1_f64_x2): Likewise.
3375 (vst1_s8_x2): Likewise.
3376 (vst1_p8_x2): Likewise.
3377 (vst1_s16_x2): Likewise.
3378 (vst1_p16_x2): Likewise.
3379 (vst1_s32_x2): Likewise.
3380 (vst1_u8_x2): Likewise.
3381 (vst1_u16_x2): Likewise.
3382 (vst1_u32_x2): Likewise.
3383 (vst1_f16_x2): Likewise.
3384 (vst1_f32_x2): Likewise.
3385 (vst1_p64_x2): Likewise.
3386 (vst1q_s8_x2): Likewise.
3387 (vst1q_p8_x2): Likewise.
3388 (vst1q_s16_x2): Likewise.
3389 (vst1q_p16_x2): Likewise.
3390 (vst1q_s32_x2): Likewise.
3391 (vst1q_s64_x2): Likewise.
3392 (vst1q_u8_x2): Likewise.
3393 (vst1q_u16_x2): Likewise.
3394 (vst1q_u32_x2): Likewise.
3395 (vst1q_u64_x2): Likewise.
3396 (vst1q_f16_x2): Likewise.
3397 (vst1q_f32_x2): Likewise.
3398 (vst1q_f64_x2): Likewise.
3399 (vst1q_p64_x2): Likewise.
3400 (vst1_s64_x3): Likewise.
3401 (vst1_u64_x3): Likewise.
3402 (vst1_f64_x3): Likewise.
3403 (vst1_s8_x3): Likewise.
3404 (vst1_p8_x3): Likewise.
3405 (vst1_s16_x3): Likewise.
3406 (vst1_p16_x3): Likewise.
3407 (vst1_s32_x3): Likewise.
3408 (vst1_u8_x3): Likewise.
3409 (vst1_u16_x3): Likewise.
3410 (vst1_u32_x3): Likewise.
3411 (vst1_f16_x3): Likewise.
3412 (vst1_f32_x3): Likewise.
3413 (vst1_p64_x3): Likewise.
3414 (vst1q_s8_x3): Likewise.
3415 (vst1q_p8_x3): Likewise.
3416 (vst1q_s16_x3): Likewise.
3417 (vst1q_p16_x3): Likewise.
3418 (vst1q_s32_x3): Likewise.
3419 (vst1q_s64_x3): Likewise.
3420 (vst1q_u8_x3): Likewise.
3421 (vst1q_u16_x3): Likewise.
3422 (vst1q_u32_x3): Likewise.
3423 (vst1q_u64_x3): Likewise.
3424 (vst1q_f16_x3): Likewise.
3425 (vst1q_f32_x3): Likewise.
3426 (vst1q_f64_x3): Likewise.
3427 (vst1q_p64_x3): Likewise.
3429 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3431 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3432 user_label_prefix to name.
3434 * tree-core.h: Update comment about the format of NAME string
3435 passed to handler in attribute_spec.
3437 * config/msp430/msp430.md: Remove erroneous subreg expression from
3438 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3439 zero_extend{q,h}isi2.
3441 2018-05-30 Borislav Petkov <bp@suse.de>
3443 * doc/extend.texi: Document some architecture specific
3444 constraints and sort entries.
3446 2018-05-30 Martin Sebor <msebor@redhat.com>
3449 * builtins.c (expand_builtin_stpcpy_1): New function.
3450 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3451 only if the former succeeds.
3453 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3455 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3458 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3460 * doc/invoke.texi (-flinker-output): Document
3462 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3464 * passes.c (ipa_write_summaries): Only modify statements if body
3466 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3467 incrementally linking.
3468 (ipa_passes): Likewise.
3469 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3470 pass down resolution info.
3471 * common.opt (flag_incremental_link): Update info.
3472 * gcc.c (plugin specs): Turn flinker-output=* to
3473 -plugin-opt=-linker-output-known
3474 * toplev.c (compile_file): Also cut compilation when doing incremental
3476 * flag-types. (enum lto_partition_model): Add
3477 LTO_LINKER_OUTPUT_NOLTOREL.
3478 (invoke.texi): Add -flinker-output docs.
3479 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3480 link same way as WPA; do not stream in dead initializers.
3482 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3483 darf2out_register_external_decl): Support incremental link.
3485 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3487 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3489 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3491 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3492 it down to simple_object_copy_lto_debug_sections.
3493 (run_gcc): Determine incremental LTO link time and configure
3494 lto1 into non-wpa mode, disable renaming of debug sections.
3496 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3498 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3499 descriptions of various incorrectly documented functions.
3501 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3504 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3505 address check not strict.
3507 2018-05-30 Richard Biener <rguenther@suse.de>
3509 PR tree-optimization/85964
3510 * tracer.c (better_p): Drop initialized count check, we only
3511 call the function with initialized counts now.
3512 (find_best_successor): Do find a best edge if one
3513 has uninitialized count.
3514 (find_best_predecessor): Likewise. Do BB frequency check only
3515 if count is initialized.
3517 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3519 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3520 (aarch64_ldrstr_offset_compare): New.
3521 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3522 load/store orderings.
3523 (aarch64_gen_adjusted_ldpstp): Likewise.
3525 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3527 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3528 Check for subset of GENERAL_REGS and FP_REGS.
3529 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3532 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3534 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3535 and wi::to_poly_offset. Add the current offset and then check
3536 whether the sum fits, rather than using an unchecked addition of
3537 a checked term. Check for a shwi rather than a uhwi.
3538 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3539 (store_constructor): Use poly_int_tree_p.
3540 (expand_expr_real_1): Likewise.
3541 * function.c (assign_temp): Likewise.
3542 * fold-const.c (const_binop): Use poly_int_tree_p and
3544 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3546 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3547 to_poly_offset for MEM offsets.
3548 * ipa-icf.c (sem_variable::equals): Likewise.
3549 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3550 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3551 wi::to_poly_offset for BIT_FIELD_REF offsets.
3552 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3554 * var-tracking.c (emit_note_insn_var_location): Use
3555 tree_to_poly_uint64.
3557 2018-05-29 Jim Wilson <jimw@sifive.com>
3559 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3561 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3564 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3565 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3567 (sse4_1_round<mode>2): Use nonimmediate_operand
3568 for operand 1 predicate.
3570 2018-05-29 Martin Sebor <msebor@redhat.com>
3571 Richard Biener <rguenther@suse.de>
3574 * calls.c (get_size_range): Call determine_value_range instead
3575 of get_value_range..
3576 * tree-vrp.h (determine_value_range): Declared new function.
3577 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3579 2018-05-29 Richard Biener <rguenther@suse.de>
3581 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3582 sure to use non-pattern stmts for get_earlier_stmt arguments.
3583 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3584 called on pattern stmts.
3585 (get_later_stmt): Likewise.
3587 2018-05-29 Martin Liska <mliska@suse.cz>
3589 PR gcov-profile/85759
3590 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3593 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3595 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3597 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3601 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3602 VEC_PACK_FLOAT_EXPR): New tree codes.
3603 * tree-pretty-print.c (op_code_prio): Handle
3604 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3605 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3606 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3607 * tree-inline.c (estimate_operator_cost): Likewise.
3608 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3609 * fold-const.c (const_binop): Likewise.
3610 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3611 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3612 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3613 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3614 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3615 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3616 * expr.c (expand_expr_real_2): Likewise.
3617 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3618 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3619 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3621 * optabs.c (expand_widen_pattern_expr): For
3622 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3623 sign from result type rather than operand's type.
3624 (expand_binop_directly): For vec_packu_float_optab and
3625 vec_packs_float_optab allow result type to be different from operand's
3627 * optabs-tree.c (optab_for_tree_code): Handle
3628 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3629 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3630 * tree-vect-generic.c (expand_vector_operations_1): Handle
3631 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3632 VEC_PACK_FLOAT_EXPR.
3633 * tree-vect-stmts.c (supportable_widening_operation): Handle
3635 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3636 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3637 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3638 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3639 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3641 (vec_pack<floatprefix>_float_<mode>): New expander.
3642 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3644 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3645 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3646 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3647 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3648 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3650 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3651 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3652 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3653 VEC_PACK_FLOAT_EXPR): Document.
3655 2018-05-29 Richard Biener <rguenther@suse.de>
3657 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3659 (stmt_vec_info_vec): Make pointer.
3660 (init_stmt_vec_info_vec): Remove.
3661 (free_stmt_vec_info_vec): Likewise.
3662 (set_stmt_vec_info_vec): New function.
3663 (free_stmt_vec_infos): Likewise.
3664 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3665 (set_vinfo_for_stmt): Likewise.
3666 (get_earlier_stmt): Likewise.
3667 (get_later_stmt): Likewise.
3668 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3669 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3670 (vec_info::~vec_info): Free stmt_vec_infos.
3671 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3672 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3673 (pass_slp_vectorize::execute): Likewise.
3674 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3675 (free_stmt_vec_info_vec): Likewise.
3676 (set_stmt_vec_info_vec): New function.
3677 (free_stmt_vec_infos): Likewise.
3678 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3679 the global stmt_vec_info_vec.
3680 * tree-parloops.c (gather_scalar_reductions): Use
3681 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3684 2018-05-29 Richard Biener <rguenther@suse.de>
3686 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3688 2018-05-29 Martin Liska <mliska@suse.cz>
3689 David Malcolm <dmalcolm@redhat.com>
3691 * vec.c (test_reverse): New.
3692 (vec_c_tests): Add new test.
3693 * vec.h (vl_ptr>::reverse): New function.
3695 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3697 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3699 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3702 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3704 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3706 2018-05-28 Richard Biener <rguenther@suse.de>
3708 PR tree-optimization/85933
3709 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3710 look at stmts marked as vectorizable.
3712 2018-05-28 Richard Biener <rguenther@suse.de>
3714 PR tree-optimization/85934
3715 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3716 vector boolean check before scalar optimization.
3718 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3720 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3723 2018-05-28 Mark Wielaard <mark@klomp.org>
3725 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3726 if it is an expression containing a minus sign.
3728 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3730 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3732 2018-05-27 Paul Koning <ni1d@arrl.net>
3734 * config/pdp11/pdp11.md (truncsihi2): Remove.
3736 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3737 Chung-Ju Wu <jasonwucj@gmail.com>
3739 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3741 (unaligned_store_dw): Ditto.
3742 * config/nds32/nds32-memory-manipulation.c
3743 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3744 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3745 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3746 (emit_setmem_word_loop): Rename to ...
3747 (emit_setmem_doubleword_loop): ... this.
3748 (nds32_gen_dup_4_byte_to_word_value): New function.
3749 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3750 (nds32_expand_setmem_loop): Refine implementation.
3751 (nds32_expand_setmem_loop_v3m): Ditto.
3752 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3755 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3757 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3759 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3761 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3762 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3763 machine->attr_no_prologue_p.
3764 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3765 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3766 (nds32_expand_epilogue): Consider attr_naked_p.
3767 (nds32_expand_epilogue_v3pop): Likewise.
3768 (nds32_can_use_return_insn): Likewise.
3769 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3770 attr_no_prologue_p fields.
3771 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3773 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3776 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3778 * config/i386/sse.md
3779 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3781 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3783 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3785 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3787 (*<floatsuffix>floatv2div2sf2): Rename to ...
3788 (*float<floatunssuffix>v2div2sf2): ... this.
3789 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3790 (float<floatunssuffix>v2div2sf2_mask): ... this.
3791 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3792 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3793 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3795 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3797 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3799 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3801 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3803 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3805 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3806 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3807 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3808 gen_ufix_truncv8dfv8si2.
3809 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3810 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3811 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3812 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3813 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3814 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3815 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3816 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3818 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3822 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3824 2018-05-25 Jim Wilson <jimw@sifive.com>
3826 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3827 * config/riscv/riscv.c (struct machine_function): Add
3828 interrupt_handler_p and attribute_checked_p fields.
3829 (riscv_attribute_table): Add interrupt.
3830 (riscv_interrupt_type_p): New.
3831 (riscv_save_reg_p): Save extra regs for interrupt handler.
3832 (riscv_use_save_libcall): Return false for interrupt handler.
3833 (riscv_first_stack_step): Add forward declaration.
3834 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3835 for interrupt handler with large frame. Use it for saved reg list.
3836 (riscv_expand_prologue): Move flag_stack_usage_info support to
3837 eliminate duplication.
3838 (riscv_expand_epilogue): Generate mret for interrupt handler.
3839 (riscv_epilogue_uses): New.
3840 (riscv_can_use_return_insn): Return false for interrupt handler.
3841 (riscv_function_ok_for_sibcall): Likewise.
3842 (riscv_set_current_function): Add interrupt handler support.
3843 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3844 * config/riscv/riscv.md (UNSPECV_MRET): New.
3846 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3848 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3850 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3852 PR tree-optimization/85712
3853 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3854 this candidate has already been replaced in-situ by a copy.
3856 2018-05-25 Jason Merrill <jason@redhat.com>
3858 PR c++/80485 - inline function non-zero address.
3859 * symtab.c (nonzero_address): Check DECL_COMDAT.
3861 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3864 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3865 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3866 extension of SImode operation. Use const123_operand predicate.
3867 (*saddsi_1): Remove.
3868 (*saddl_se_1): Ditto.
3870 (*ssubl_se_1): Ditto.
3871 * config/alpha/predicates.md (const123_operand): New predicate.
3872 * config/alpha/constraints.md (P): Use IN_RANGE.
3874 2018-05-25 Richard Biener <rguenther@suse.de>
3876 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3878 (ref_maybe_used_by_stmt_p): Likewise.
3879 (stmt_may_clobber_ref_p): Likewise.
3880 (stmt_may_clobber_ref_p_1): Likewise.
3881 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3883 (ref_maybe_used_by_stmt_p): Likewise.
3884 (stmt_may_clobber_ref_p): Likewise.
3885 (stmt_may_clobber_ref_p_1): Likewise.
3886 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3887 the alias oracle to disambiguate DRs with stmts DR analysis
3889 (vect_analyze_data_refs): Do not give up on not analyzable
3890 DRs for BB vectorization. Remove code truncating the dataref
3893 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3896 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3897 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3898 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3900 2018-05-25 Richard Biener <rguenther@suse.de>
3902 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3903 function, combining stmt data ref gathering and fatal analysis
3905 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3906 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3907 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3908 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3909 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3911 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3913 PR tree-optimization/85720
3914 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3915 SCC if all partitions are builtins.
3916 (version_loop_by_alias_check): New parameter. Generate cancelable
3917 runtime alias check if all partitions are builtins.
3918 (distribute_loop): Update call to above function.
3920 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3922 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3923 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3924 (parm_default_def_partition_arg): Ditto.
3925 (set_parm_default_def_partition): Ditto.
3926 (get_parm_default_def_partitions): Ditto and make it static.
3927 (get_undefined_value_partitions): Ditto and make it static.
3928 (remove_ssa_form): Refactor call to init_var_map here.
3929 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3930 computation for loop region.
3931 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3932 (register_default_def): Delete.
3933 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3934 (parm_default_def_partition_arg): Ditto.
3935 (set_parm_default_def_partition): Ditto.
3936 (get_parm_default_def_partitions): Ditto and make it static.
3937 (get_undefined_value_partitions): Ditto and make it static.
3938 (coalesce_with_default, coalesce_with_default): Update comment.
3939 (create_coalesce_list_for_region): New func factored out from
3940 create_outofssa_var_map.
3941 (populate_coalesce_list_for_outofssa): New func factored out from
3942 create_outofssa_var_map and coalesce_ssa_name.
3943 (create_outofssa_var_map): Delete.
3944 (coalesce_ssa_name): Refactor to support live range computation.
3945 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3946 (get_parm_default_def_partitions): Delete.
3947 (get_undefined_value_partitions): Ditto.
3948 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3949 computation for loop region.
3950 (new_tree_live_info, loe_visit_block): Ditto.
3951 (live_worklist, set_var_live_on_entry): Ditto.
3952 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3953 * tree-ssa-live.h (struct _var_map): New fields.
3954 (init_var_map): Change decl.
3955 (region_contains_p): New.
3957 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3959 * tree-ssa-live.h (live_merge_and_clear): Delete.
3961 2018-05-25 Richard Biener <rguenther@suse.de>
3964 * tree-dump.c (dequeue_and_dump): Remove access to removed
3965 operand 2 of a SWITCH_EXPR.
3967 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3969 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3970 multiplication and division.
3971 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3972 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3973 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3974 (cond_udiv_optab, cond_umod_optab): New optabs.
3975 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3976 (IFN_COND_RDIV): New internal functions.
3977 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3978 TRUNC_MOD_EXPR and RDIV_EXPR.
3979 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3980 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3982 (SVE_INT_BINARY): Include mult.
3983 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3984 (optab, sve_int_op): Handle mult.
3985 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3987 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3988 for SVE_INT_BINARY_SD.
3990 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3992 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3993 (optab, sve_int_op): Handle div and udiv.
3994 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3995 for SVE_INT_BINARY_SD.
3996 (*<optab><mode>3): New insn for the same.
3998 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4000 * tree-vect-patterns.c: Include predict.h.
4001 (vect_recog_divmod_pattern): Restrict check for division support
4002 to when optimizing for size.
4004 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4006 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4007 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4008 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4009 (gimple_match_op::set_op): Likewise.
4010 (gimple_resimplify4): Declare.
4011 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4012 (expr::gen_transform): Likewise.
4013 (decision_tree::gen): Generate a simplification routine for 4 operands.
4014 * gimple-match-head.c (gimple_simplify): Add an overload for
4015 4 operands. In the top-level function, handle up to 4 call
4016 arguments and call gimple_resimplify4.
4017 (gimple_resimplify4): New function.
4018 (build_call_internal): Pass a fourth operand.
4019 (maybe_push_to_seq): Likewise.
4020 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4021 Fold VEC_COND_EXPRs of an operation and a default value into
4022 an IFN_COND_* function if possible.
4023 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4025 (SVE_COND_FP_BINARY): Include them.
4026 (optab, sve_fp_op): Handle them.
4027 (SVE_INT_BINARY_REV): New code iterator.
4028 (SVE_COND_FP_BINARY_REV): New int iterator.
4029 (commutative): New int attribute.
4030 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4032 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4034 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4035 (*cond_<optab><mode>): New patterns for reversed operands.
4037 2018-05-25 Richard Biener <rguenther@suse.de>
4039 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4040 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4041 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4042 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4043 * tree-vect-data-refs.c (everywhere): Adjust users.
4044 * tree-vect-loop.c (everywhere): Likewise.
4045 * tree-vect-slp.c (everywhere): Likewise.
4046 * tree-vect-stmts.c (everywhere): Likewise.
4047 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4049 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4051 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4053 (gcc_cv_as_section_exclude): ... this.
4054 Try Solaris as #exclude syntax.
4055 * configure: Regenerate.
4056 * config.in: Regenerate.
4057 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4059 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4060 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4062 * varasm.c (default_elf_asm_named_section): Don't check if
4063 HAVE_GAS_SECTION_EXCLUDE is defined.
4065 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4067 * doc/md.texi: Update the documentation of the cond_* optabs
4068 to mention the new final operand. Fix GET_MODE_NUNITS call.
4069 Describe the scalar case too.
4070 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4071 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4073 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4074 (get_conditional_internal_fn): Update comment.
4075 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4076 accumulator value as a final argument to conditional functions.
4077 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4078 a define_expand and add an "else" operand. Assert for now that
4079 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4080 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4081 (*cond_<optab><mode>): New patterns.
4082 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4083 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4084 (UNSPEC_COND_EOR): Delete.
4085 (optab): Remove associated mappings.
4086 (SVE_INT_BINARY): New code iterator.
4087 (sve_int_op): Remove int attribute and add "minus" to the code
4089 (SVE_COND_INT_OP): Delete.
4090 (SVE_COND_FP_OP): Rename to...
4091 (SVE_COND_FP_BINARY): ...this.
4093 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4095 * optabs.c (can_reuse_operands_p): New function.
4096 (maybe_legitimize_operands): Try to reuse the results for
4099 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4101 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4102 Add {q} suffix to insn mnemonic.
4104 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4106 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4107 (msp430_warn_func_return): New.
4109 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4111 * fold-const.c (tree_nonzero_bits): New function.
4112 * fold-const.h (tree_nonzero_bits): Likewise.
4113 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4114 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4116 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4120 * varasm.c (assemble_alias): Check ifunc_resolver only on
4123 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4126 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4127 when memory input operand is handled.
4129 2018-05-24 Luis Machado <luis.machado@linaro.org>
4131 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4133 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4135 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4137 * match.pd: Delay FMA folds until after vectorization.
4139 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4142 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4143 address check not strict.
4145 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4147 * gimple-match.h (gimple_match_op): New class.
4148 (mprts_hook): Replace parameters with a gimple_match_op *.
4149 (maybe_build_generic_op): Likewise.
4150 (gimple_simplified_result_is_gimple_val): Replace parameters with
4151 a const gimple_match_op *.
4152 (gimple_simplify): Replace code_helper * and tree * parameters with
4153 a gimple_match_op * parameter.
4154 (gimple_resimplify1): Replace code_helper *, tree and tree *
4155 parameters with a gimple_match_op * parameter.
4156 (gimple_resimplify2): Likewise.
4157 (gimple_resimplify3): Likewise.
4158 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4159 parameters with a gimple_match_op * parameter.
4160 * gimple-match-head.c (gimple_simplify): Change prototypes of
4161 auto-generated functions to take a gimple_match_op * instead of
4162 separate code_helper * and tree * parameters. Make the same
4163 change in the top-level overload and update calls to the
4164 gimple_resimplify routines. Update calls to the auto-generated
4165 functions and to maybe_push_res_to_seq in the publicly-facing
4166 operation-specific gimple_simplify overloads.
4167 (gimple_match_op::MAX_NUM_OPS): Define.
4168 (gimple_resimplify1): Replace rcode and ops with a single res_op
4169 parameter. Update call to gimple_simplify.
4170 (gimple_resimplify2): Likewise.
4171 (gimple_resimplify3): Likewise.
4172 (mprts_hook): Replace parameters with a gimple_match_op *.
4173 (maybe_build_generic_op): Likewise.
4174 (build_call_internal): Replace type, nargs and ops with
4175 a gimple_match_op *.
4176 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4177 with a single gimple_match_op *. Update calls to mprts_hook,
4178 build_call_internal and gimple_simplified_result_is_gimple_val.
4179 Factor out code that is common to the tree_code and combined_fn cases.
4180 * genmatch.c (expr::gen_transform): Replace tem_code and
4181 tem_ops with a gimple_match_op called tem_op. Update calls
4182 to the gimple_resimplify functions and maybe_push_res_to_seq.
4183 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4184 res_ops. Update call to the gimple_resimplify functions.
4185 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4186 (decision_tree::gen): Make the functions take a gimple_match_op *
4187 called res_op instead of separate res_code and res_ops parameters.
4188 Update call accordingly.
4189 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4190 and ops with a single res_op parameter. Update calls to
4191 maybe_build_generic_op and maybe_push_res_to_seq.
4192 (fold_stmt_1): Update calls to gimple_simplify and
4193 replace_stmt_with_simplification.
4194 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4195 and gimple_simplified_result_is_gimple_val.
4196 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4198 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4199 with a gimple_match_op *.
4200 (vn_nary_build_or_lookup): Likewise. Update call to
4201 vn_nary_build_or_lookup_1.
4202 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4203 gimple_match_op *. Update calls to the gimple_resimplify routines
4204 and to gimple_simplified_result_is_gimple_val.
4205 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4206 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4207 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4208 (visit_nary_op): Likewise.
4209 (visit_reference_op_load): Likewise.
4211 2018-05-23 Luis Machado <luis.machado@linaro.org>
4213 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4214 modifier for printing the step amount.
4216 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4219 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4222 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4224 * doc/sourcebuild.texi (Endianness): New subsubsection.
4226 2018-05-23 Luis Machado <luis.machado@linaro.org>
4228 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4229 <prefetch_dynamic_strides>: New const bool field.
4230 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4231 prefetch_dynamic_strides.
4232 (exynosm1_prefetch_tune): Likewise.
4233 (thunderxt88_prefetch_tune): Likewise.
4234 (thunderx_prefetch_tune): Likewise.
4235 (thunderx2t99_prefetch_tune): Likewise.
4236 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4238 (aarch64_override_options_internal): Update to set
4239 PARAM_PREFETCH_DYNAMIC_STRIDES.
4240 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4241 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4242 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4243 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4244 prefetch-dynamic-strides setting.
4246 2018-05-23 Luis Machado <luis.machado@linaro.org>
4248 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4249 <minimum_stride>: New const int field.
4250 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4251 minimum_stride field defaulting to -1.
4252 (exynosm1_prefetch_tune): Likewise.
4253 (thunderxt88_prefetch_tune): Likewise.
4254 (thunderx_prefetch_tune): Likewise.
4255 (thunderx2t99_prefetch_tune): Likewise.
4256 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4257 <default_opt_level>: Set to 3.
4258 (aarch64_override_options_internal): Update to set
4259 PARAM_PREFETCH_MINIMUM_STRIDE.
4260 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4261 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4262 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4263 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4264 stride is constant and is below the minimum stride threshold.
4266 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4268 * config/arm/arm-cpus.in (mode26): Delete.
4269 (armv4): Delete mode26 reference.
4270 * config/arm/arm.c (arm_configure_build_target): Delete use of
4273 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4275 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4277 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4278 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4280 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4281 (floatunsdidf2): Ditto.
4283 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4285 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4286 (fixuns_trunc<mode>si2_avx512f): Ditto.
4287 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4288 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4289 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4291 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4293 PR rtl-optimization/79985
4294 * df-scan.c (df_insn_refs_collect): Remove special case for
4295 global registers and asm statements.
4297 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4299 * extend.texi (Global Register Variables): Rewrite the bullet list.
4300 Note that the register is available for allocation. Note that access
4301 via inline asm must use constraints. Add note about async-signal
4302 handlers. Remove paragraph about automagic register selection.
4304 2018-05-23 Richard Biener <rguenther@suse.de>
4306 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4307 of fixed offset from memset VN.
4309 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4311 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4313 (alloc_cand_and_find_basis): Initialize first_interp field.
4314 (slsr_process_mul): Modify first_interp field.
4315 (slsr_process_add): Likewise.
4316 (slsr_process_cast): Modify first_interp field for each new
4318 (slsr_process_copy): Likewise.
4319 (dump_candidate): Dump first_interp field.
4320 (replace_mult_candidate): Process all interpretations, not just
4322 (replace_rhs_if_not_dup): Likewise.
4323 (replace_one_candidate): Likewise.
4325 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4327 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4329 (aarch64_needs_frame_chain): New function.
4330 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4332 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4335 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4336 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4337 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4338 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4339 as true for strict-align.
4340 (aarch64_can_inline_p): Perform checks even when callee has no
4341 attributes to check for strict alignment.
4342 * doc/extend.texi (AArch64 Function Attributes): Document
4344 * doc/invoke.texi: (AArch64 Options): Likewise.
4346 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4348 PR tree-optimization/85853
4349 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4350 the handling of the root of the node to...
4351 (vect_slp_analyze_node_operations_1): ...this new function,
4352 and run the whole thing with the child nodes' def types
4353 set according to their SLP node's def type.
4355 2018-05-23 Richard Biener <rguenther@suse.de>
4358 * tree-data-ref.c (create_runtime_alias_checks): Defer
4359 and ignore overflow warnings.
4361 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4363 PR tree-optimization/85822
4364 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4367 2018-05-23 Richard Biener <rguenther@suse.de>
4369 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4370 memset constants via native_interpret_expr.
4372 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4375 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4377 (cgraph_node::create_alias): Likewise.
4378 (cgraph_node::get_availability): Check ifunc_resolver instead
4379 of looking up ifunc attribute.
4380 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4381 * varasm.c (do_assemble_alias): Likewise.
4382 (assemble_alias): Likewise.
4383 (default_binds_local_p_3): Likewise.
4384 * cgraph.h (cgraph_node): Add ifunc_resolver.
4385 (cgraph_node::only_called_directly_or_aliased_p): Return false
4387 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4389 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4390 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4391 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4392 instead of looking up ifunc attribute.
4394 2018-05-22 Luis Machado <luis.machado@linaro.org>
4396 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4398 2018-05-22 Martin Sebor <msebor@redhat.com>
4401 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4402 only when expasion succeeds.
4403 (expand_builtin_strcmp): Same.
4404 (expand_builtin_strncmp): Same.
4406 2018-05-22 Martin Sebor <msebor@redhat.com>
4408 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4410 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4411 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4413 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4414 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4416 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4417 Add check for legitimate_address.
4418 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4419 (aarch64_swap_ldrstr_operands): New.
4420 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4423 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4424 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4426 * config/aarch64/aarch64.md: New patterns to generate stp
4428 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4429 single words and double words.
4430 (load_pair_sw, load_pair_dw): Likewise.
4431 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4433 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4435 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4436 for different mode ldpstp and add peephole for merged zero stores.
4438 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4440 (aarch64_gen_store_pair): Rename calls to match new patterns.
4441 (aarch64_gen_load_pair): Rename calls to match new patterns.
4442 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4443 (load_pair<DREG:mode><DREG2:mode>): ... This.
4444 (store_pair<mode>): Rename to...
4445 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4446 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4448 (V_INT_EQUIV): Handle SImode.
4449 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4452 2018-05-22 Martin Sebor <msebor@redhat.com>
4455 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4456 or ajust the presumed bound on an operation to avoid unnecessary
4459 2018-05-22 Martin Sebor <msebor@redhat.com>
4461 PR tree-optimization/85826
4462 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4463 assuming that a DECL necesarily has a constant size.
4465 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4468 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4470 2018-05-22 Richard Biener <rguenther@suse.de>
4472 PR tree-optimization/85834
4473 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4474 non-constant and non-zero memset arguments.
4476 2018-05-22 Martin Liska <mliska@suse.cz>
4479 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4481 2018-05-22 Richard Biener <rguenther@suse.de>
4483 PR tree-optimization/85863
4484 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4485 comparisons when vectype is specified.
4486 (vectorizable_condition): Do not specify vectype for
4487 vect_is_simple_cond when SLP vectorizing.
4489 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4492 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4493 define __ibm128 as long double.
4494 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4495 as a distinct type when IEEE 128-bit support is enabled.
4496 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4497 128-bit types to use the correct functions.
4498 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4499 convert between 128-bit floating point types that have different
4500 modes but the same representation, instead of using gen_lowpart to
4502 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4504 (IFKF_reg): New attributes to give the register constraints for
4506 (extend<mode>tf2_internal): New insns to mark an explicit
4507 conversion between 128-bit floating point types that have a
4508 different mode but share the same representation.
4510 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4512 PR tree-optimization/85814
4513 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4514 a null return from get_strinfo when unsharing the next
4515 strinfo in the chain.
4517 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4520 * varasm.c (weak_finish): Clean up weak_decls.
4522 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4524 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4525 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4526 UNSPEC_UADALP values.
4527 * config/aarch64/iterators.md (ABAL): New int iterator.
4530 (sur): Add mappings for the above.
4531 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4533 (aarch64_<sur>abal<mode>_4): Likewise.
4534 (aarch64_<sur>adalp<mode>_3): Likewise.
4535 (<sur>sadv16qi): New define_expand.
4537 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4539 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4540 (*movdf_internal): Ditto.
4541 (*rcpsf2_sse): Ditto.
4542 (*rsqrtsf2_sse): Ditto.
4543 (*sqrt<mode>2_sse): Ditto.
4545 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4547 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4549 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4550 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4551 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4552 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4554 * config/aarch64/arm_neon.h: Likewise.
4555 * config/aarch64/iterators.md (VQ_I): New.
4557 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4559 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4561 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4564 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4566 * config/nds32/constraints.md (S): New constraint.
4567 * config/nds32/nds32.md (call_internal): Use constraint S.
4568 (call_value_internal): Likewise.
4569 (sibcall_internal): Likewise.
4570 (sibcall_value_internal): Likewise.
4572 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4573 Chung-Ju Wu <jasonwucj@gmail.com>
4575 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4578 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4579 Chung-Ju Wu <jasonwucj@gmail.com>
4581 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4582 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4583 (nds32_rtx_costs_impl): Simplify.
4584 (nds32_address_cost_impl): Simplify.
4585 (nds32_init_rtx_costs): New function.
4586 (nds32_rtx_costs_speed_prefer): Likewise.
4587 (nds32_rtx_costs_size_prefer): Likewise.
4588 (nds32_address_cost_speed_prefer): Likewise.
4589 (nds32_address_cost_speed_fwprop): Likewise.
4590 (nds32_address_cost_size_prefer): Likewise.
4591 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4592 * config/nds32/nds32.c (nds32_option_override): Use
4593 nds32_init_rtx_costs function.
4595 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4597 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4598 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4599 (TARGET_PIPELINE_N8): Likewise.
4600 (TARGET_PIPELINE_N10): Likewise.
4601 (TARGET_PIPELINE_N13): Likewise.
4602 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4604 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4606 * config/nds32/nds32-fpu.md: Update copyright year.
4608 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4610 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4612 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4614 * config/nds32/nds32.c
4615 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4616 * config/nds32/nds32.opt (minline-asm-r15): New option.
4618 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4620 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4622 * config/nds32/nds32.md (abssi2): New pattern.
4624 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4626 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4627 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4628 Merge insn pattern from sse_cvtsi2ss<round_name> and
4629 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4630 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4631 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4632 using SWI48 mode iterator.
4633 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4634 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4635 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4636 pattern from sse_cvttss2si<round_saeonly_name>
4637 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4638 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4639 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4640 using SWI48 mode iterator.
4641 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4642 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4643 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4644 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4645 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4646 using SWI48 mode iterator.
4647 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4648 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4649 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4650 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4651 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4652 SWI48 mode iterator.
4653 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4654 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4655 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4656 pattern from sse_cvttsd2si<round_saeonly_name>
4657 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4659 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4661 * config/nds32/nds32-md-auxiliary.c
4662 (nds32_valid_smw_lwm_base_p): Refine.
4663 (nds32_output_smw_single_word): Refine.
4664 (nds32_output_smw_double_word): New.
4665 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4667 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4669 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4670 (nds32_output_stack_pop): Refine.
4671 (nds32_expand_unaligned_load): Refine.
4672 (nds32_expand_unaligned_store): Refine.
4674 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4675 Chung-Ju Wu <jasonwucj@gmail.com>
4677 * config/nds32/constants.md: Add TP_REGNUM constant.
4678 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4679 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4681 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4682 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4683 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4684 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4685 and PIC code generation.
4686 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4688 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4690 * config/nds32/nds32.md: Support TLS and PIC.
4691 * config/nds32/nds32.c: Support TLS and PIC.
4692 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4693 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4696 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4698 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4699 mode with E_ prefix.
4701 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4702 Chung-Ju Wu <jasonwucj@gmail.com>
4704 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4705 * config/nds32/nds32-md-auxiliary.c
4706 (symbolic_reference_mentioned_p): New.
4707 (nds32_legitimize_ict_address): New.
4708 (nds32_expand_ict_move): New.
4709 (nds32_indirect_call_referenced_p): New.
4710 (nds32_symbol_binds_local_p): Delete.
4711 (nds32_long_call_p): Modify.
4712 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4713 * config/nds32/nds32-protos.h
4714 (symbolic_reference_mentioned_p): Declare.
4715 (nds32_legitimize_ict_address): Declare.
4716 (nds32_expand_ict_move): Declare.
4717 (nds32_indirect_call_referenced_p): Declare.
4718 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4719 (nds32_relax_group): Use nds32_ict_const_p as condition.
4720 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4721 (nds32_asm_file_start): Output ict_model directive in asm code.
4722 (nds32_legitimate_address_p): Consider indirect call.
4723 (nds32_print_operand): Consider indirect call.
4724 (nds32_print_operand_address): Consider indirect call.
4725 (nds32_insert_attributes): Handle "indirect_call" attribute.
4726 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4727 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4728 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4729 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4730 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4731 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4732 (TARGET_ICT_MODEL_SMALL): Define.
4733 (TARGET_ICT_MODEL_LARGE): Define.
4734 * config/nds32/nds32.md (movsi): Consider ict model.
4735 (call, call_value): Consider ict model.
4736 (sibcall, sibcall_value): Consider ict model.
4737 * config/nds32/nds32.opt (mict-model): New option.
4738 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4741 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4742 Monk Chiang <sh.chiang04@gmail.com>
4743 Jim Wilson <jimw@sifive.com>
4745 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4746 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4747 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4748 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4749 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4750 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4751 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4752 compute save_libcall_adjustment properly.
4753 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4754 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4755 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4756 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4757 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4758 (ABI_SPEC): Handle mabi=ilp32e.
4759 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4760 (RVE): Add RVE mask.
4761 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4762 <-march>: Add rv32e as an example.
4764 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4767 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4768 (intra_create_variable_infos): Handle C++ constructors.
4770 2018-05-18 Martin Liska <mliska@suse.cz>
4772 * passes.def: Remove a redundant pass.
4774 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4777 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4779 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4781 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4783 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4785 (armv2, armv2a, armv3, armv3m): Delete architectures.
4786 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4787 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4788 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4790 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4791 (*mulsidi3adddi): Likewise.
4792 (mulsidi3): Likewise.
4793 (*mulsidi3_nov6): Likewise.
4794 (umulsidi3): Likewise.
4795 (umulsidi3_nov6): Likewise.
4796 (umaddsidi4): Likewise.
4797 (*umulsidi3adddi): Likewise.
4798 (smulsi3_highpart): Likewise.
4799 (*smulsi3_highpart_nov6): Likewise.
4800 (umulsi3_highpart): Likewise.
4801 (*umulsi3_highpart_nov6): Likewise.
4802 * config/arm/arm.h (arm_arch3m): Delete.
4803 * config/arm/arm.c (arm_arch3m): Delete.
4804 (arm_option_override_internal): Update armv3-related comment.
4805 (arm_configure_build_target): Delete use of isa_bit_mode32.
4806 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4807 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4808 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4810 (mulusa3): Likewise.
4811 * config/arm/arm-protos.h (arm_arch3m): Delete.
4812 * config/arm/arm-tables.opt: Regenerate.
4813 * config/arm/arm-tune.md: Likewise.
4814 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4815 deleted architectures.
4817 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4819 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4820 (armv5t, armv5te): New features.
4821 (ARMv5, ARMv5e): Delete fgroups.
4822 (ARMv5t, ARMv5te): Adjust for above changes.
4824 (armv5, armv5e): Delete arches.
4825 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4827 (*call_reg_arm): Likewise.
4828 (*call_value_reg_armv5): Likewise.
4829 (*call_value_reg_arm): Likewise.
4830 (*call_symbol): Likewise.
4831 (*call_value_symbol): Likewise.
4832 (*sibcall_insn): Likewise.
4833 (*sibcall_value_insn): Likewise.
4835 (prefetch): Likewise.
4836 (define_split and define_peephole2 dependent on arm_arch5):
4838 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4840 (TARGET_ARM_QBIT): Likewise.
4841 (TARGET_DSP_MULTIPLY): Likewise.
4842 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4843 (arm_arch5, arm_arch5e): Delete.
4844 (arm_arch5t, arm_arch5te): Declare.
4845 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4846 (arm_arch5t): Declare.
4847 (arm_option_reconfigure_globals): Update for the above.
4848 (arm_options_perform_arch_sanity_checks): Update comment, replace
4849 use of arm_arch5 with arm_arch5t.
4850 (use_return_insn): Likewise.
4851 (arm_emit_call_insn): Likewise.
4852 (output_return_instruction): Likewise.
4853 (arm_final_prescan_insn): Likewise.
4854 (arm_coproc_builtin_available): Likewise.
4855 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4856 arm_arch5e with arm_arch5t and arm_arch5te.
4857 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4858 (arm_arch5t, arm_arch5te): Declare.
4859 * config/arm/arm-tables.opt: Regenerate.
4860 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4861 * config/arm/t-multilib: Likewise.
4862 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4863 instead of arm_arch5.
4864 (*call_reg_thumb1): Likewise.
4865 (*call_value_reg_thumb1_v5): Likewise.
4866 (*call_value_reg_thumb1): Likewise.
4867 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4869 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4871 2018-05-18 Martin Liska <mliska@suse.cz>
4873 PR gcov-profile/84846
4874 * doc/gcov.texi: Document -t option of gcov tool.
4876 2018-05-18 Martin Liska <mliska@suse.cz>
4878 PR gcov-profile/84846
4879 * gcov.c (print_usage): Add new -t option.
4880 (process_args): Handle the option.
4881 (generate_results): Use stdout as output when requested by
4884 2018-05-18 Martin Liska <mliska@suse.cz>
4886 PR gcov-profile/84846
4887 * coverage.c (coverage_init): Write PWD to .gcno file.
4888 * doc/gcov.texi: Document how working directory is printed.
4889 * gcov-dump.c (dump_gcov_file): Print PWD.
4890 * gcov.c (output_intermediate_file): Likewise.
4891 (read_graph_file): Read PWD string.
4892 (output_lines): Print PWD.
4894 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4897 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4898 for retval and return false if all args to phi are zero.
4900 2018-05-18 Richard Biener <rguenther@suse.de>
4902 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4904 (evrp_dom_walker::before_dom_children): Call it.
4906 2018-05-18 Richard Biener <rguenther@suse.de>
4908 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4909 results when processing array refs with variable index.
4911 2018-05-18 Toon Moene <toon@moene.org>
4913 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4914 directly after that of -floop-interchange. Indicate that both
4915 options are enabled by default when specifying -O3.
4917 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4919 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4920 iterator. Delete separate integer-mode vec_set<mode> expander.
4921 (aarch64_simd_vec_setv2di): Delete.
4922 (vec_setv2di): Delete.
4923 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4924 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4925 the "w, r" alternative.
4927 2018-05-18 Martin Liska <mliska@suse.cz>
4929 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4930 * tree-pass.h (make_pass_lower_switch_O0): New function.
4931 * tree-switch-conversion.c (node_has_low_bound): Remove.
4932 (node_has_high_bound): Likewise.
4933 (node_is_bounded): Likewise.
4934 (class pass_lower_switch): Make it a template type and create
4936 (pass_lower_switch::execute): Add template argument.
4937 (make_pass_lower_switch): New function.
4938 (make_pass_lower_switch_O0): New function.
4939 (do_jump_if_equal): Remove.
4940 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4941 all the hard work to tree optimization passes.
4943 2018-05-18 Martin Liska <mliska@suse.cz>
4945 * dbgcnt.c (limit_low): Renamed from limit.
4946 (limit_high): New variable.
4947 (dbg_cnt_is_enabled): Check for upper limit.
4948 (dbg_cnt): Adjust dumping.
4949 (dbg_cnt_set_limit_by_index): Add new argument for high
4951 (dbg_cnt_set_limit_by_name): Likewise.
4952 (dbg_cnt_process_single_pair): Parse new format.
4953 (dbg_cnt_process_opt): Use strtok.
4954 (dbg_cnt_list_all_counters): Remove 'value' and add
4956 * doc/invoke.texi: Document changes.
4958 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4960 * doc/sourcebuild.texi (scalar_all_fma): Document.
4961 * tree.def (FMA_EXPR): Delete.
4962 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4963 * internal-fn.c (ternary_direct): New macro.
4964 (expand_ternary_optab_fn): Likewise.
4965 (direct_ternary_optab_supported_p): Likewise.
4966 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4967 * builtins.c (fold_builtin_fma): Delete.
4968 (fold_builtin_3): Don't call it.
4969 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4970 * expr.c (expand_expr_real_2): Likewise.
4971 * fold-const.c (operand_equal_p): Likewise.
4972 (fold_ternary_loc): Likewise.
4973 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4974 * gimple.c (DEFTREECODE): Likewise.
4975 * gimplify.c (gimplify_expr): Likewise.
4976 * optabs-tree.c (optab_for_tree_code): Likewise.
4977 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4978 * tree-eh.c (operation_could_trap_p): Likewise.
4979 (stmt_could_throw_1_p): Likewise.
4980 * tree-inline.c (estimate_operator_cost): Likewise.
4981 * tree-pretty-print.c (dump_generic_node): Likewise.
4982 (op_code_prio): Likewise.
4983 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4984 * tree-ssa-operands.c (get_expr_operands): Likewise.
4985 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4986 * fold-const-call.h (fold_fma): Delete.
4987 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4988 CFN_FNMA and CFN_FNMS.
4990 * genmatch.c (combined_fn): New enum.
4991 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4992 (commutative_op): New function.
4993 (commutate): Use it. Handle more than 2 operands.
4994 (dt_operand::gen_gimple_expr): Use commutative_op.
4995 (parser::parse_expr): Allow :c to be used with non-binary
4996 operators if the commutative operand is known.
4997 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4998 CFN_FMS, CFN_FNMA and CFN_FNMS.
4999 (backprop::process_assign_use): Remove FMA_EXPR handling.
5000 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5001 (gen_hsa_fma): New function.
5002 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5003 IFN_FNMA and IFN_FNMS.
5004 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5005 * gimple-fold.h (follow_all_ssa_edges): Declare.
5006 * gimple-fold.c (follow_all_ssa_edges): New function.
5007 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5008 gimple_build interface and use follow_all_ssa_edges to fold the result.
5009 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5010 instead of checking for optabs directly.
5011 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5012 rather than FMA_EXPRs.
5013 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5014 call to IFN_FMA instead of an FMA_EXPR.
5016 2018-05-17 Jim Wilson <jimw@sifive.com>
5018 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5019 sign extended promoted subreg, and the range does not have the sign bit
5020 set, then do a sign extend.
5022 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5023 test, check for sign extended subreg and/or constant operands, and
5024 do a sign extend in that case.
5026 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5028 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5029 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5031 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5032 Change logics_shift_reg to logics_shift_imm.
5033 (thunderx2t99_fp_loadpair_basic): Delete.
5034 (thunderx2t99_fp_storepair_basic): Delete.
5035 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5036 (thunderx2t99_asimd_polynomial): Delete.
5037 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5038 and neon_fp_mul_d_scalar_q.
5039 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5040 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5041 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5042 (thunderx2t99_asimd_lut): Add missing tbl types.
5043 (thunderx2t99_asimd_ext): Delete.
5044 (thunderx2t99_asimd_load1_1_mult): Delete.
5045 (thunderx2t99_asimd_load1_2_mult): Delete.
5046 (thunderx2t99_asimd_load1_ldp): New.
5047 (thunderx2t99_asimd_load1): New.
5048 (thunderx2t99_asimd_load2): Add missing *load2* types.
5049 (thunderx2t99_asimd_load3): New.
5050 (thunderx2t99_asimd_load4): New.
5051 (thunderx2t99_asimd_store1_1_mult): Delete.
5052 (thunderx2t99_asimd_store1_2_mult): Delete.
5053 (thunderx2t99_asimd_store2_mult): Delete.
5054 (thunderx2t99_asimd_store2_onelane): Delete.
5055 (thunderx2t99_asimd_store_stp): New.
5056 (thunderx2t99_asimd_store1): New.
5057 (thunderx2t99_asimd_store2): New.
5058 (thunderx2t99_asimd_store3): New.
5059 (thunderx2t99_asimd_store4): New.
5061 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5063 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5064 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5066 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5067 Segher Boessenkool <segher@kernel.crashing.org>
5070 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5073 2018-05-17 Richard Biener <rguenther@suse.de>
5075 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5076 for pruning loop and prune defs feeding only already visited PHIs.
5078 2018-05-17 Richard Biener <rguenther@suse.de>
5080 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5082 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5083 Richard Biener <rguenther@suse.de>
5085 PR tree-optimization/85793
5086 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5087 for VMAT_ELEMENTWISE.
5089 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5091 * internal-fn.h (lookup_internal_fn): Declare
5092 * internal-fn.c (lookup_internal_fn): New function.
5093 * gimple.c (gimple_build_call_from_tree): Handle calls to
5095 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5096 internal function names.
5097 * tree-pretty-print.c (dump_generic_node): Likewise.
5098 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5100 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5102 * gimple-fold.h (gimple_build): Make the function forms take
5103 combined_fn rather than built_in_function.
5104 (gimple_simplify): Likewise.
5105 * gimple-match-head.c (gimple_simplify): Likewise.
5106 * gimple-fold.c (gimple_build): Likewise.
5107 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5108 rather than gimple_build_call_internal.
5109 (get_initial_defs_for_reduction): Likewise.
5110 (vect_create_epilog_for_reduction): Likewise.
5111 (vectorizable_live_operation): Likewise.
5113 2018-05-17 Martin Liska <mliska@suse.cz>
5115 * gimple-ssa-sprintf.c (format_directive): Do not use
5116 space in between 'G_' and '('.
5118 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5121 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5122 even if the mask is not all ones.
5125 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5127 (ix86_gimple_fold_builtin): Likewise.
5130 * config/i386/i386.c: Include tree-vector-builder.h.
5131 (ix86_vector_shift_count): New function.
5132 (ix86_fold_builtin): Fold shift builtins by scalar count.
5133 (ix86_gimple_fold_builtin): Likewise.
5135 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5136 _mm512_setzero): New intrinsics.
5138 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5139 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5141 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5142 code generation for cases where splatting a value is not useful.
5143 * simplify-rtx.c (simplify_ternary_operation): Simplify
5144 vec_merge across a vec_duplicate and a paradoxical subreg forming
5145 a vector mode to a vec_concat.
5147 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5149 * config.gcc: Support "goldmont-plus".
5150 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5152 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5153 PROCESSOR_GOLDMONT_PLUS.
5154 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5155 (processor_target_table): Add "goldmont-plus".
5156 (PTA_GOLDMONT_PLUS): Define.
5157 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5158 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5159 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5160 (fold_builtin_cpu): Add "goldmont-plus".
5161 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5162 (ix86_option_override_internal): Add "goldmont-plus".
5163 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5164 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5165 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5166 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5168 2018-05-17 Richard Biener <rguenther@suse.de>
5170 PR tree-optimization/85757
5171 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5172 remove defs that only feed that PHI from further processing.
5174 2018-05-16 Jim Wilson <jimw@sifive.com>
5176 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5178 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5180 2018-05-16 Mark Wielaard <mark@klomp.org>
5182 * dwarf2out.c (count_index_strings): New function.
5183 (output_indirect_strings): Call count_index_strings and generate
5184 header for dwarf_version >= 5.
5186 2018-05-16 Mark Wielaard <mark@klomp.org>
5188 * dwarf2out.c (dwarf_FORM): New function.
5189 (set_indirect_string): Use dwarf_FORM.
5190 (reset_indirect_string): Likewise.
5191 (size_of_die): Likewise.
5192 (value_format): Likewise.
5193 (output_die): Likewise.
5194 (add_skeleton_AT_string): Likewise.
5195 (output_macinfo_op): Likewise.
5196 (index_string): Likewise.
5197 (output_index_string_offset): Likewise.
5198 (output_index_string): Likewise.
5199 (count_index_strings): Likewise.
5201 2018-05-16 Carl Love <cel@us.ibm.com>
5203 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5204 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5206 2018-05-16 Martin Jambor <mjambor@suse.cz>
5208 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5209 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5211 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5213 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5214 (fnma<mode>4): Likewise.
5215 (fms<mode>4): Likewise.
5216 (fnms<mode>4): Likewise.
5217 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5218 (aarch64_fnma<mode>4): Likewise.
5219 (aarch64_fms<mode>4): Likewise.
5220 (aarch64_fnms<mode>4): Likewise.
5221 (aarch64_fnmadd<mode>4): Likewise.
5223 2018-05-16 Jason Merrill <jason@redhat.com>
5225 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5227 2018-05-16 Richard Biener <rguenther@suse.de>
5229 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5230 (dump_stmt_cost): Declare.
5231 (add_stmt_cost): Dump cost we add.
5232 (add_stmt_costs): New function.
5233 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5235 (vect_analyze_stmt): Adjust prototype.
5236 (vectorizable_condition): Likewise.
5237 (vectorizable_live_operation): Likewise.
5238 (vectorizable_reduction): Likewise.
5239 (vectorizable_induction): Likewise.
5240 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5241 cost vector to pass to vectorizable_ and record afterwards.
5242 (vect_model_reduction_cost): Take cost vector argument and adjust.
5243 (vect_model_induction_cost): Likewise.
5244 (vectorizable_reduction): Likewise.
5245 (vectorizable_induction): Likewise.
5246 (vectorizable_live_operation): Likewise.
5247 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5248 SLP_TREE_NUMBER_OF_VEC_STMTS.
5249 (vect_analyze_slp_cost_1): Remove.
5250 (vect_analyze_slp_cost): Likewise.
5251 (vect_slp_analyze_node_operations): Take visited args and
5252 a target cost vector. Avoid processing already visited stmt sets.
5253 (vect_slp_analyze_operations): Use a local cost vector to gather
5254 costs and register those of non-discarded instances.
5255 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5256 (vect_schedule_slp_instance): Remove copying of
5257 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5259 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5260 adding cost. Record cost entry location.
5261 (vect_prologue_cost_for_slp_op): Function to compute cost of
5262 a constant or invariant generated for SLP vect in the prologue,
5263 split out from vect_analyze_slp_cost_1.
5264 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5265 (vect_model_promotion_demotion_cost): Likewise.
5266 (vect_model_store_cost): Likewise, make static.
5267 (vect_model_load_cost): Likewise.
5268 (vectorizable_bswap): Add cost vector arg and adjust.
5269 (vectorizable_call): Likewise.
5270 (vectorizable_simd_clone_call): Likewise.
5271 (vectorizable_conversion): Likewise.
5272 (vectorizable_assignment): Likewise.
5273 (vectorizable_shift): Likewise.
5274 (vectorizable_operation): Likewise.
5275 (vectorizable_store): Likewise.
5276 (vectorizable_load): Likewise.
5277 (vectorizable_condition): Likewise.
5278 (vectorizable_comparison): Likewise.
5279 (can_vectorize_live_stmts): Likewise.
5280 (vect_analyze_stmt): Likewise.
5281 (vect_transform_stmt): Adjust calls to vectorizable_*.
5282 * tree-vectorizer.c: Include gimple-pretty-print.h.
5283 (dump_stmt_cost): New function.
5285 2018-05-16 Richard Biener <rguenther@suse.de>
5287 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5288 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5289 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5290 (check_name): New callback.
5291 (dse_classify_store): Track cycles via a visited bitmap of PHI
5292 defs and simplify handling of in-loop and across loop dead stores
5293 and properly fail for loop-variant refs. Handle byte-tracking with
5294 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5297 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5299 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5300 (vect_get_mask_type_for_stmt): Likewise.
5301 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5303 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5304 to determine the statement's vector type and the vector type that
5305 should be used for calculating nunits. Deal with cases in which
5306 the type has to be deferred.
5307 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5308 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5309 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5310 (vect_determine_vf_for_stmt): New functions, split out from...
5311 (vect_determine_vectorization_factor): ...here.
5312 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5313 (vect_get_mask_type_for_stmt): New functions, split out from
5314 vect_determine_vectorization_factor.
5316 2018-05-16 Richard Biener <rguenther@suse.de>
5318 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5319 verify the [VEC_]COND_EXPR embedded comparison.
5321 2018-05-15 Martin Sebor <msebor@redhat.com>
5323 PR tree-optimization/85753
5324 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5325 RECORD_TYPE in addition to ARRAY_TYPE.
5327 2018-05-15 Martin Sebor <msebor@redhat.com>
5330 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5332 2018-05-15 Richard Biener <rguenther@suse.de>
5334 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5335 add by_clobber_p one. Change algorithm to collect all defs
5336 representing uses we need to walk and try reducing them to
5337 a single one before failing.
5338 (dse_dom_walker::dse_optimize_stmt): Adjust.
5340 2018-05-13 Mark Wielaard <mark@klomp.org>
5342 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5343 (size_of_loc_descr): Likewise.
5344 (output_loc_operands): Likewise.
5345 (output_loc_operands_raw): Likewise.
5346 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5347 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5348 (hash_loc_operands): Likewise.
5349 (compare_loc_operands): Likewise.
5351 2018-05-14 Mark Wielaard <mark@klomp.org>
5353 * dwarf2out.c (count_index_addrs): New function.
5354 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5356 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5358 PR tree-optimization/83648
5359 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5360 return value as malloc candidate.
5362 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5365 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5366 param as true in call to suggest_attribute.
5368 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5370 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5371 -mreadonly-in-sdata.
5373 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5375 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5377 (aarch64_crypto_aesd_fused): Likewise.
5379 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5381 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5382 (movsi_aarch64): Likewise.
5383 (load_pairsi): Likewise.
5384 (load_pairdi): Likewise.
5385 (store_pairsi): Likewise.
5386 (store_pairdi): Likewise.
5387 (load_pairsf): Likewise.
5388 (load_pairdf): Likewise.
5389 (store_pairsf): Likewise.
5390 (store_pairdf): Likewise.
5391 (zero_extend): Likewise.
5392 (trunc): Swap alternatives.
5393 (fcvt_target): Add '?' to prefer w over r.
5395 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5398 * config/i386/i386.md: Disallow non-commutative arithmetics in
5399 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5400 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5401 in the peephole2 before it.
5403 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5405 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5406 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5407 (ix86_handle_option): Handle -mcldemote.
5408 * config.gcc: New header.
5409 * config/i386/cldemoteintrin.h: New file.
5410 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5411 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5413 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5414 OPTION_MASK_ISA_CLDEMOTE.
5415 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5416 (ix86_valid_target_attribute_inner_p): Ditto.
5417 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5418 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5419 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5420 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5421 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5423 * config/i386/i386.opt: Add -mcldemote.
5424 * config/i386/x86intrin.h: New header.
5425 * doc/invoke.texi: Add -mcldemote.
5427 2018-05-14 Richard Biener <rguenther@suse.de>
5429 * doc/match-and-simplify.texi: Adjust :s documentation.
5431 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5433 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5434 intended memcpy size.
5435 (REORDER_45): Likewise.
5437 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5439 * sort.cc: New file.
5440 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5441 * vec.c (qsort_chk): Use gcc_qsort.
5442 * Makefile.in (OBJS-libcommon): Add sort.o.
5443 (build/sort.o): New target. Use it...
5444 (BUILD_RTL): ... here, and...
5445 (build/gencfn-macros): ... here, and...
5446 (build/genmatch): ... here.
5448 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5449 Chung-Ju Wu <jasonwucj@gmail.com>
5451 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5452 * config/nds32/nds32-graywolf.md: New file.
5453 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5454 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5456 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5457 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5458 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5459 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5460 * config/nds32/pipelines.md: Include n15 settings.
5462 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5463 Chung-Ju Wu <jasonwucj@gmail.com>
5465 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5466 * config/nds32/nds32-n13.md: New file.
5467 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5468 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5470 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5471 * config/nds32/nds32.md (pipeline_model): Add n13.
5472 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5473 * config/nds32/pipelines.md: Include n13 settings.
5475 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5476 Chung-Ju Wu <jasonwucj@gmail.com>
5478 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5479 * config/nds32/nds32-n10.md: New file.
5480 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5481 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5483 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5484 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5485 * config/nds32/nds32.md (pipeline_model): Add n10.
5486 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5487 * config/nds32/pipelines.md: Include n10 settings.
5489 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5490 Kito Cheng <kito.cheng@gmail.com>
5491 Chung-Ju Wu <jasonwucj@gmail.com>
5493 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5494 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5495 Add enum values for DSP extension instructions.
5496 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5498 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5499 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5501 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5502 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5503 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5504 * config/nds32/nds32-intrinsic.md: Likewise.
5505 * config/nds32/nds32_intrinsic.h: Likewise.
5506 * config/nds32/nds32-md-auxiliary.c: Likewise.
5507 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5508 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5509 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5510 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5511 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5512 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5513 TYPE_DMAC in switch statement.
5514 * config/nds32/nds32.c: New checking and implementation for DSP
5515 extension instructions.
5516 * config/nds32/nds32.h: Likewise.
5517 * config/nds32/nds32.md: Likewise.
5518 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5519 * config/nds32/predicates.md: Implement new predicates for DSP
5522 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5524 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5525 Reformat alternatives and attributes so it is easier to identify
5526 which constraints/attributes go with which instruction.
5527 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5528 (mov<mode>_softfloat32, FMOVE64): Likewise.
5529 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5530 (mov<mode>_softfloat64, FMOVE64): Likewise.
5532 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5534 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5536 (Basic PowerPC Built-in Functions): The new name of the
5537 subsection previously known as "PowerPC Built-in Functions".
5538 (Basic PowerPC Built-in Functions Available on all Configurations):
5540 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5541 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5542 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5543 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5545 2018-05-11 Martin Jambor <mjambor@suse.cz>
5548 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5551 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5554 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5556 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5558 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5559 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5560 (ix86_handle_option): Handle -mwaitpkg.
5561 * config.gcc: New header.
5562 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5563 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5564 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5566 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5567 OPTION_MASK_ISA_WAITPKG.
5568 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5569 (ix86_option_override_internal): Add PTA_WAITPKG.
5570 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5571 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5572 IX86_BUILTIN_TPAUSE.
5573 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5574 __builtin_ia32_umwait and __builtin_ia32_tpause.
5575 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5576 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5577 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5578 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5579 UNSPECV_TPAUSE): New.
5580 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5581 * config/i386/i386.opt: Add -mwaitpkg.
5582 * config/i386/waitpkgintrin.h: New file.
5583 * config/i386/x86intrin.h: New header.
5584 * doc/invoke.texi: Add -mwaitpkg.
5586 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5589 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5591 (cortex-m0): Use armv6s-m isa.
5592 (cortex-m0plus): Likewise.
5593 (cortex-m1): Likewise.
5594 (cortex-m0.small-multiply): Likewise.
5595 (cortex-m0plus.small-multiply): Likewise.
5596 (cortex-m1.small-multiply): Likewise.
5598 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5599 Jakub Jelinek <jakub@redhat.com>
5601 PR tree-optimization/85692
5602 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5603 source permute as well.
5605 2018-05-11 Martin Liska <mliska@suse.cz>
5608 * doc/extend.texi: Document LLVM style format for no_sanitize
5611 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5613 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5614 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5615 (mode_supports_vsx_dform_quad): Likewise.
5616 (mode_supports_vmx_dform): Move these functions to be next to the
5617 other mode_supports functions.
5618 (mode_supports_dq_form): Likewise.
5619 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5620 mode_supports_dq_form.
5621 (reg_offset_addressing_ok_p): Likewise.
5622 (offsettable_ok_by_alignment): Likewise.
5623 (rs6000_legitimate_offset_address_p): Likewise.
5624 (legitimate_lo_sum_address_p): Likewise.
5625 (rs6000_legitimize_address): Likewise.
5626 (rs6000_legitimize_reload_address): Likewise.
5627 (rs6000_secondary_reload_inner): Likewise.
5628 (rs6000_preferred_reload_class): Likewise.
5629 (rs6000_output_move_128bit): Likewise.
5631 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5633 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5634 Generate SImode target register for null target.
5635 <case IX86_BUILTIN_XGETBV>: Ditto.
5636 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5637 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5639 2018-05-10 Carl Love <cel@us.ibm.com>
5641 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5642 dcbtt and dcbtstt if operands[2] is 0.
5644 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5647 * config/i386/sse.md (usadv64qi): New expander.
5649 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5651 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5652 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5653 -maltivec=be support.
5654 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5655 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5656 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5657 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5658 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5659 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5660 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5661 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5662 altivec_vsumsws): Adjust.
5663 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5664 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5665 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5667 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5668 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5669 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5670 (altivec_lve<VI_char>x): Delete expand.
5671 (*altivec_lve<VI_char>x_internal): Rename to...
5672 (altivec_lve<VI_char>x): ... this.
5673 (altivec_lvxl_<mode>): Delete expand.
5674 (*altivec_lvxl_<mode>_internal): Rename to ...
5675 (altivec_lvxl_<mode>): ... this.
5676 (altivec_stvxl_<mode>): Delete expand.
5677 (*altivec_stvxl_<mode>_internal): Rename to ...
5678 (altivec_stvxl_<mode>): ... this.
5679 (altivec_stve<VI_char>x): Delete expand.
5680 (*altivec_stve<VI_char>x_internal): Rename to ...
5681 (altivec_stve<VI_char>x): ... this.
5682 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5683 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5684 reduc_plus_scal_<mode>): Adjust.
5685 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5687 (rs6000_cpu_cpp_builtins): Adjust.
5688 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5689 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5690 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5691 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5692 -maltivec=be support.
5693 (rs6000_split_vec_extract_var): Adjust.
5694 (rs6000_split_v4si_init): Adjust.
5695 (swap_selector_for_mode): Delete.
5696 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5697 altivec_expand_stvex_be): Delete.
5698 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5699 -maltivec=be support.
5700 (rs6000_gimple_fold_builtin): Ditto.
5701 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5703 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5704 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5705 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5706 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5707 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5708 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5709 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5710 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5711 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5712 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5713 anonymous split): Adjust.
5714 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5715 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5717 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5719 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5720 when --with-gxx-include-dir is also specified.
5721 * configure: Regenerate.
5723 2018-05-09 Jim Wilson <jimw@sifive.com>
5726 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5727 * config/riscv/t-withmultilib: New.
5728 * config/riscv/withmultilib.h: New.
5729 * doc/install.texi: Document RISC-V --with-multilib-list support.
5731 2018-05-09 Richard Biener <rguenther@suse.de>
5733 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5735 (vect_bb_vectorization_profitable_p): Adjust. Compute
5736 actual scalar cost using the cost vector and the add_stmt_cost
5739 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5741 PR rtl-optimization/85645
5742 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5743 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5745 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5747 PR rtl-optimization/85645
5748 * shrink-wrap.c (spread_components): Return a boolean saying if
5749 anything was changed.
5750 (try_shrink_wrapping_separate): Iterate spread_components until
5751 nothing changes anymore.
5753 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5755 PR rtl-optimization/85645
5756 * regrename.c (build_def_use): Also kill the chains that include the
5757 destination of a REG_CFA_REGISTER note.
5759 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5761 PR rtl-optimization/85645
5762 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5763 insn that has a REG_CFA_REGISTER note.
5765 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5767 * cfgexpand.c (expand_clobber): New function.
5768 (expand_gimple_stmt_1): Use it.
5769 * tree-vect-stmts.c (vect_clobber_variable): New function,
5771 (vectorizable_simd_clone_call): ...here.
5772 (vectorizable_store): Emit a clobber either side of an
5773 IFN_STORE_LANES sequence.
5774 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5776 2018-05-09 Tom de Vries <tom@codesourcery.com>
5779 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5780 (define_insn "trap_if_false"): Add exit after trap.
5782 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5784 PR rtl-optimization/85638
5785 * bb-reorder.c: Include common/common-target.h.
5786 (create_forwarder_block): New function extracted from...
5787 (fix_up_crossing_landing_pad): ...here. Rename into...
5788 (dw2_fix_up_crossing_landing_pad): ...this.
5789 (sjlj_fix_up_crossing_landing_pad): New function.
5790 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5791 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5792 from both partitions and exit the loop after one iteration.
5794 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5797 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5799 (Basic PowerPC Built-in Functions): The new name of the
5800 subsection previously known as "PowerPC Built-in Functions".
5801 (Basic PowerPC Built-in Functions Available on all Configurations):
5803 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5805 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5807 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5809 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5812 2018-05-08 Jim Wilson <jimw@sifive.com>
5814 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5815 (LD_EMUL_SUFFIX): New.
5816 (LINK_SPEC): Use it.
5818 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5820 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5822 (Basic PowerPC Built-in Functions): The new name of the
5823 subsection previously known as "PowerPC Built-in Functions".
5824 (Basic PowerPC Built-in Functions Available on all Configurations):
5826 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5828 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5830 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5832 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5835 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5838 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5839 after cmpelim optimization.
5841 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5843 * config.gcc: Support "goldmont".
5844 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5845 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5847 * config/i386/i386.c (m_GOLDMONT): Define.
5848 (processor_target_table): Add "goldmont".
5849 (PTA_GOLDMONT): Define.
5850 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5851 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5852 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5853 (fold_builtin_cpu): Add "goldmont".
5854 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5855 (ix86_option_override_internal): Add "goldmont".
5856 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5857 (processor_type): Add PROCESSOR_GOLDMONT.
5858 * config/i386/i386.md: Add CPU "glm".
5859 * config/i386/glm.md: New file.
5860 * config/i386/x86-tune.def: Add m_GOLDMONT.
5861 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5863 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5866 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5868 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5869 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5870 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5873 * config/i386/i386.c (ix86_fold_builtin): Handle
5874 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5877 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5878 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5880 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5883 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5884 (check_arch): Likewise.
5885 (check_fpu): Return the result rather than printing it.
5886 (end arch): Fix operator precedence.
5887 (end cpu): Likewise.
5888 (END): Print the result from check_fpu.
5890 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5891 Alan Hayward <alan.hayward@arm.com>
5892 David Sherwood <david.sherwood@arm.com>
5894 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5895 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5896 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5897 (*fcmuo<mode>_and): New patterns.
5899 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5901 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5902 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5903 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5904 (cmp_op, sve_imm_con): New code attributes.
5905 (SVE_COND_INT_CMP, imm_con): Delete.
5906 (cmp_op): Remove above unspecs from int attribute.
5907 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5909 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5910 comparison-specific unspecs.
5911 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5912 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5913 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5914 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5915 (*vec_fcm<cmp_op><mode>): Rename to...
5916 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5917 (*vec_fcmuo<mode>): Rename to...
5918 (*fcmuo<mode>): ...this and adjust likewise.
5919 (*pred_fcm<cmp_op><mode>): New pattern.
5920 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5921 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5923 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5925 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5926 (aarch64_emit_sve_predicated_cond): New function.
5927 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5928 (aarch64_emit_unspec_cond_or): Replace with...
5929 (aarch64_emit_sve_or_conds): ...this new function. Use
5930 aarch64_emit_sve_ptrue_op for the individual comparisons and
5931 aarch64_emit_binop to OR them together.
5932 (aarch64_emit_inverted_unspec_cond): Replace with...
5933 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5934 aarch64_emit_sve_ptrue_op for the comparison and
5935 aarch64_emit_unop to invert the result.
5936 (aarch64_expand_sve_vec_cmp_float): Update after the above
5937 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5939 2018-05-07 Nathan Sidwell <nathan@acm.org>
5941 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5942 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5943 (Backwards Compatibility): Likewise.
5945 2018-05-07 Luis Machado <luis.machado@linaro.org>
5949 2018-05-07 Luis Machado <luis.machado@linaro.org>
5951 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5952 <prefetch_dynamic_strides>: New const bool field.
5953 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5954 prefetch_dynamic_strides.
5955 (exynosm1_prefetch_tune): Likewise.
5956 (thunderxt88_prefetch_tune): Likewise.
5957 (thunderx_prefetch_tune): Likewise.
5958 (thunderx2t99_prefetch_tune): Likewise.
5959 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5961 (aarch64_override_options_internal): Update to set
5962 PARAM_PREFETCH_DYNAMIC_STRIDES.
5963 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5964 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5965 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5966 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5967 prefetch-dynamic-strides setting.
5969 2018-05-07 Luis Machado <luis.machado@linaro.org>
5971 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5972 <minimum_stride>: New const int field.
5973 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5974 minimum_stride field.
5975 (exynosm1_prefetch_tune): Likewise.
5976 (thunderxt88_prefetch_tune): Likewise.
5977 (thunderx_prefetch_tune): Likewise.
5978 (thunderx2t99_prefetch_tune): Likewise.
5979 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5980 (aarch64_override_options_internal): Update to set
5981 PARAM_PREFETCH_MINIMUM_STRIDE.
5982 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5983 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5984 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5985 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5986 stride is constant and is below the minimum stride threshold.
5988 2018-05-07 Luis Machado <luis.machado@linaro.org>
5990 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5993 2018-05-07 Luis Machado <luis.machado@linaro.org>
5995 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5996 <prefetch_dynamic_strides>: New const bool field.
5997 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5998 prefetch_dynamic_strides.
5999 (exynosm1_prefetch_tune): Likewise.
6000 (thunderxt88_prefetch_tune): Likewise.
6001 (thunderx_prefetch_tune): Likewise.
6002 (thunderx2t99_prefetch_tune): Likewise.
6003 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6005 (aarch64_override_options_internal): Update to set
6006 PARAM_PREFETCH_DYNAMIC_STRIDES.
6007 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6008 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6009 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6010 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6011 prefetch-dynamic-strides setting.
6013 2018-05-07 Luis Machado <luis.machado@linaro.org>
6015 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6016 <minimum_stride>: New const int field.
6017 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6018 minimum_stride field.
6019 (exynosm1_prefetch_tune): Likewise.
6020 (thunderxt88_prefetch_tune): Likewise.
6021 (thunderx_prefetch_tune): Likewise.
6022 (thunderx2t99_prefetch_tune): Likewise.
6023 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6024 (aarch64_override_options_internal): Update to set
6025 PARAM_PREFETCH_MINIMUM_STRIDE.
6026 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6027 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6028 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6029 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6030 stride is constant and is below the minimum stride threshold.
6032 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6035 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6036 the type is addressable. Don't force op into register if it has
6039 2018-05-05 Roland McGrath <mcgrathr@google.com>
6042 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6043 any section for which we don't know a specific type it should have,
6044 regardless of name. Previously this was done only for the exact
6045 names ".init_array", ".fini_array", and ".preinit_array".
6046 (default_elf_asm_named_section): Add comment about
6047 relationship with default_section_type_flags and SECTION_NOTYPE.
6048 (get_section): Don't consider it a type conflict if one side has
6049 SECTION_NOTYPE and the other doesn't, as long as neither has the
6050 SECTION_BSS et al used in the default_section_type_flags logic.
6052 2018-05-05 Tom de Vries <tom@codesourcery.com>
6055 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6056 (workaround_barsyncs): New function.
6057 (nvptx_reorg): Use workaround_barsyncs.
6058 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6059 (define_expand "nvptx_membar_cta"): New define_expand.
6060 (define_insn "*nvptx_membar_cta"): New insn.
6062 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6064 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6065 To improve optimization opportunities.
6066 * builtin-types.def: The new needed builtin types for the above.
6068 2018-05-04 Richard Biener <rguenther@suse.de>
6070 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6071 * gimple-ssa-store-merging.c
6072 (imm_store_chain_info::output_merged_store): Remove redundant create,
6073 release split_store vector contents on failure.
6074 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6075 scalar stmt vector on cache hit.
6077 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6079 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6081 * config.gcc (powerpc-xilinx-eabi*): Remove.
6082 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6084 (fusion_addis_mem_combo_load): Ditto.
6085 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6087 (rs6000_cpu_cpp_builtins): Ditto.
6088 * config/rs6000/rs6000-linux.c
6089 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6090 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6091 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6093 (rs6000_setup_reg_addr_masks): Ditto.
6094 (rs6000_init_hard_regno_mode_ok): Ditto.
6095 (rs6000_option_override_internal): Ditto.
6096 (legitimate_lo_sum_address_p): Ditto.
6097 (rs6000_legitimize_address): Ditto.
6098 (rs6000_legitimize_reload_address): Ditto.
6099 (rs6000_legitimate_address_p): Ditto.
6100 (abi_v4_pass_in_fpr): Ditto.
6101 (setup_incoming_varargs): Ditto.
6102 (rs6000_gimplify_va_arg): Ditto.
6103 (rs6000_split_multireg_move): Ditto.
6104 (rs6000_savres_strategy): Ditto.
6105 (rs6000_emit_prologue_components): Ditto.
6106 (rs6000_emit_epilogue_components): Ditto.
6107 (rs6000_emit_prologue): Ditto.
6108 (rs6000_emit_epilogue): Ditto.
6109 (rs6000_elf_file_end): Ditto.
6110 (rs6000_function_value): Ditto.
6111 (rs6000_libcall_value): Ditto.
6112 * config/rs6000/rs6000.h: Ditto.
6113 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6114 (TARGET_MINMAX): ... this. New.
6115 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6116 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6117 (*movsi_internal1_single): Delete.
6118 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6119 mfpu=, mxilinx-fpu): Delete.
6120 * config/rs6000/singlefp.h: Delete.
6121 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6122 * config/rs6000/t-rs6000: Ditto.
6123 * config/rs6000/t-xilinx: Delete.
6124 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6125 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6126 (VStype_simple): Delete.
6127 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6128 * config/rs6000/xfpu.h: Delete.
6129 * config/rs6000/xfpu.md: Delete.
6130 * config/rs6000/xilinx.h: Delete.
6131 * config/rs6000/xilinx.opt: Delete.
6132 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6133 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6135 2018-05-04 Tom de Vries <tom@codesourcery.com>
6138 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6141 2018-05-04 Richard Biener <rguenther@suse.de>
6144 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6145 (expand_complex_div_wide): Likewise.
6146 (expand_complex_operations_1): Likewise.
6147 (expand_complex_libcall): Preserve EH info of the original stmt.
6148 (tree_lower_complex): Handle removed blocks.
6149 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6150 on complex multiplication and division libcall builtins.
6152 2018-05-04 Richard Biener <rguenther@suse.de>
6155 * fold-const.c (negate_expr_p): Restrict negation of operand
6156 zero of a division to when we know that can happen without
6158 (fold_negate_expr_1): Likewise.
6160 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6163 * real.h (real_nextafter): Declare.
6164 * real.c (real_nextafter): New function.
6165 * fold-const-call.c (fold_const_nextafter): New function.
6166 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6167 CASE_CFN_NEXTTOWARD.
6168 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6169 even when arg1_mode is different from arg0_mode.
6171 2018-05-03 Nathan Sidwell <nathan@acm.org>
6173 * doc/extend.texi (Deprecated Features): Remove
6175 (Backwards Compatibility): Likewise.
6176 * doc/invoke.texi (C++ Language Options): Likewise.
6177 (C++ Dialect Options): Likewise.
6179 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6182 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6183 _mm512_mask_mullox_epi64): New intrinsics.
6185 2018-05-03 Tom de Vries <tom@codesourcery.com>
6188 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6189 dump files): Add offload-tree.
6191 2018-05-03 Richard Biener <rguenther@suse.de>
6193 PR tree-optimization/85615
6194 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6195 to loops not nested in BBs loop father to avoid creating multi-entry
6198 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6200 PR tree-optimization/70291
6201 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6202 arguments. Change return type to tree. Emit libcall as a new
6203 statement rather than replacing existing one when inplace_p is true.
6204 (expand_complex_multiplication_components): New function.
6205 (expand_complex_multiplication): Expand floating-point complex
6206 multiplication using the above.
6207 (expand_complex_division): Rename inner_type parameter to type.
6208 Update expand_complex_libcall call-site.
6209 (expand_complex_operations_1): Update expand_complex_multiplication
6210 and expand_complex_division call-sites.
6212 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6215 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6216 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6217 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6218 the highest significant bit of the shift count mask is clear. In
6219 check whether and[sq]i3 is needed verify that all significant bits
6220 of the shift count other than the highest are set.
6222 2018-05-02 Tom de Vries <tom@codesourcery.com>
6225 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6226 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6227 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6228 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6229 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6230 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6231 __builtin_goacc_parlevel_size.
6233 2018-05-02 Richard Biener <rguenther@suse.de>
6235 PR tree-optimization/85597
6236 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6237 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6240 2018-05-02 Tom de Vries <tom@codesourcery.com>
6243 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6244 dump files): Add ltrans-tree.
6246 2018-05-02 Tom de Vries <tom@codesourcery.com>
6249 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6250 dump files): Add wpa-ipa.
6252 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6254 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6255 powerpc*-*-linux*paired* target.
6256 * config/rs6000/750cl.h: Delete.
6257 * config/rs6000/paired.h: Delete.
6258 * config/rs6000/paired.md: Delete.
6259 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6261 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6262 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6263 comment. Remove paired float support.
6264 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6265 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6267 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6268 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6270 * config/rs6000/rs6000.c: Remove paired float support.
6271 (paired_expand_vector_init, paired_expand_vector_move,
6272 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6273 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6274 paired_expand_builtin, paired_expand_predicate_builtin,
6275 paired_init_builtins): Delete.
6276 * config/rs6000/rs6000.h: Remove paired float support.
6277 * config/rs6000/rs6000.md: Remove paired float support.
6278 (move_from_CR_ov_bit): Delete.
6279 * config/rs6000/rs6000.opt (mpaired): Delete.
6280 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6281 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6283 2018-05-02 Richard Biener <rguenther@suse.de>
6286 * gimplify.c (gimplify_save_expr): When in SSA form allow
6287 SAVE_EXPRs to compute to SSA vars.
6289 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6292 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6293 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6294 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6295 clobber operands[2], instead use a new pseudo. Formatting fixes.
6297 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6299 PR tree-optimization/85586
6300 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6301 exit early for statements in the same group if the accesses are
6304 2018-05-02 Tom de Vries <tom@codesourcery.com>
6307 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6310 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6312 PR tree-optimization/85143
6313 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6315 2018-05-01 Tom de Vries <tom@codesourcery.com>
6318 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6319 not found" error message.
6321 2018-05-01 Tom de Vries <tom@codesourcery.com>
6324 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6325 * vec.c (test_ordered_remove_if): New function.
6326 (vec_c_tests): Call test_ordered_remove_if.
6327 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6328 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6329 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6330 VEC_ORDERED_REMOVE_IF.
6332 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6334 PR tree-optimization/82665
6335 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6336 pointer subtraction where arguments come from a memchr call.
6338 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6340 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6341 --push-state --as-needed and --pop-state instead of --as-needed and
6342 --no-as-needed if ld supports it.
6343 * configure: Regenerated.
6346 * doc/install.texi2html: Replace _002d with - and _002a with * in
6347 generated html files using sed.
6349 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6352 * gcc-rich-location.c (blank_line_before_p): New function.
6353 (use_new_line): New function.
6354 (gcc_rich_location::add_fixit_insert_formatted): New function.
6355 * gcc-rich-location.h
6356 (gcc_rich_location::add_fixit_insert_formatted): New function.
6358 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6360 * selftest.c (assert_streq): Rename "expected" and "actual" to
6361 "val1" and "val2". Extend NULL-handling to cover both inputs
6362 symmetrically, while still requiring both to be non-NULL for a pass.
6363 * selftest.h (assert_streq): Rename "expected" and "actual" to
6365 (ASSERT_EQ): Likewise.
6366 (ASSERT_EQ_AT): Likewise.
6367 (ASSERT_KNOWN_EQ): Likewise.
6368 (ASSERT_KNOWN_EQ_AT): Likewise.
6369 (ASSERT_NE): Likewise.
6370 (ASSERT_MAYBE_NE): Likewise.
6371 (ASSERT_MAYBE_NE_AT): Likewise.
6372 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6373 the assertion to pass.
6374 (ASSERT_STREQ_AT): Likewise.
6376 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6378 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6379 interaction with -pie.
6381 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6383 * selftest.h: Fix alphabetization of per-source-file selftest
6386 2018-04-30 Jason Merrill <jason@redhat.com>
6388 PR c++/61982 - dead stores to destroyed objects.
6389 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6392 2018-04-30 Jason Merrill <jason@redhat.com>
6394 * tree.c (build_clobber): New.
6395 * tree.h: Declare it.
6396 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6398 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6400 * diagnostic-show-locus.c (layout::layout): Update for
6401 location_get_source_line returning a char_span.
6402 (struct char_span): Move to input.h.
6403 (struct correction): Update for fields in char_span becoming
6405 (struct source_line): Update for location_get_source_line
6406 returning a char_span.
6407 (layout::print_line): Likewise.
6408 * edit-context.c (edited_file::print_content): Likewise.
6409 (edited_file::print_diff_hunk): Likewise.
6410 (edited_file::print_run_of_changed_lines): Likewise.
6411 (edited_file::get_num_lines): Likewise.
6412 (edited_line::edited_line): Likewise.
6413 * final.c (asm_show_source): Likewise.
6414 * input.c (location_get_source_line): Convert return type
6415 from const char * to char_span, losing the final "line_len"
6417 (dump_location_info): Update for the above.
6418 (get_substring_ranges_for_loc): Likewise. Use a char_span
6419 when handling the literal within the line.
6420 (test_reading_source_line): Update for location_get_source_line
6421 returning a char_span.
6422 * input.h (class char_span): Move here from
6423 diagnostic-show-locus.c, converting from a struct to a class.
6424 Make data members private.
6425 (char_span::operator bool): New.
6426 (char_span::length): New.
6427 (char_span::get_buffer): New.
6428 (char_span::operator[]): New.
6429 (char_span::subspan): Make const.
6430 (char_span::xstrdup): New.
6431 (location_get_source_line): Convert return type from const char *
6432 to char_span, losing the final "line_size" param.
6434 2018-04-30 Jan Hubicka <jh@suse.cz>
6436 * lto-wrapper.c (ltrans_priorities): New static var.
6437 (cmp_priority): New.
6438 (run_gcc): Read priorities and if doing parallel build order
6439 the Makefile by them.
6441 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6443 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6445 2018-04-30 Richard Biener <rguenther@suse.de>
6447 * tree-cfg.c (verify_address): Remove base argument, add
6448 flag whether to check TREE_ADDRESSABLE and do that.
6449 (verify_expr): Remove.
6450 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6451 (verify_gimple_assign_single): Likewise.
6452 (verify_gimple_switch): Likewise.
6453 (verify_expr_location_1): Dereference tp once. Add (disabled)
6454 piece from verify_expr.
6455 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6457 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6459 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6461 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6463 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6464 (small_data_pattern): Likewise.
6465 (arc_rewrite_small_data): Likewise.
6466 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6467 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6468 (get_symbol_alignment): New function.
6469 (legitimate_small_data_address_p): Likewise.
6470 (legitimate_scaled_address): Update, call
6471 legitimate_small_data_address_p.
6472 (output_sdata): New static variable.
6473 (arc_print_operand): Update how we handle small data operands.
6474 (arc_print_operand_address): Likewise.
6475 (arc_legitimate_address_p): Update, use
6476 legitimate_small_data_address_p.
6477 (arc_rewrite_small_data_p): Remove.
6478 (arc_rewrite_small_data_1): Likewise.
6479 (arc_rewrite_small_data): Likewise.
6480 (small_data_pattern): Likewise.
6481 (compact_sda_memory_operand): Update to use
6482 legitimate_small_data_address_p and get_symbol_alignment.
6483 (prepare_move_operands): Don't rewite sdata pattern.
6484 (prepare_extend_operands): Remove.
6485 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6487 (zero_extendqisi2): Likewise.
6488 (zero_extendhisi2): Likewise.
6489 (extendqihi2): Likewise.
6490 (extendqisi2): Likewise.
6491 (extendhisi2): Likewise.
6495 * config/arc/constraints.md (Usd): Change it to memory constraint.
6497 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6499 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6500 as source of std instructions.
6501 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6502 allow 6-bit constants as source for store instructions.
6503 (movdi_insn): Update instruction pattern to allow 6-bit constants
6504 as source for store instructions.
6506 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6508 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6510 2018-04-30 Nathan Sidwell <nathan@acm.org>
6511 Sandra Loosemore <sandra@codesourcery.com>
6513 * dumpfile.c (dump_open): Allow '-' for stdout.
6514 * doc/invoke.texi (Developer Options): Document dump filename
6515 determination early. Document stdin/stdout selection.
6517 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6519 Microblaze Target: PIC data text relative
6521 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6522 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6524 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6525 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6526 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6527 New addressing mode for data-text relative position indepenedent code.
6528 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6529 'ADDRESS_SYMBOLIC_TXT_REL'.
6530 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6531 (microblaze_legitimate_pic_operand): Exclude function calls from
6532 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6533 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6535 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6536 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6537 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6538 for 'address + offset'.
6539 (microblaze_expand_prologue): Add new function prologue call for
6541 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6542 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6543 table in case of TARGET_PIC_DATA_TEXT_REL.
6544 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6545 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6546 Add new macros 'UNSPEC_TEXT',
6547 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6548 + exclude function calls from 'UNSPEC_PLT' in case of data text
6550 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6551 new target hook for generating address diff vector tables in case of
6553 * doc/tm.texi : Regenerate.
6554 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6555 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6556 of addr diff vector generation.
6557 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6558 target hook definition.
6559 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6560 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6561 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6562 option for data text relative.
6564 2018-04-30 Richard Biener <rguenther@suse.de>
6566 * tree-chrec.h (evolution_function_is_constant_p): Remove
6568 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6570 2018-04-30 Richard Biener <rguenther@suse.de>
6573 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6575 2018-04-30 Richard Biener <rguenther@suse.de>
6577 PR tree-optimization/28364
6578 PR tree-optimization/85275
6579 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6580 copying first exit test.
6582 2018-04-28 Mark Wielaard <mark@klomp.org>
6584 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6586 (dwarf_AT): Handle DW_AT_addr_base.
6587 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6589 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6592 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6593 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6594 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6595 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6597 2018-04-28 Richard Biener <rguenther@suse.de>
6599 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6600 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6601 to reflect use. Only add interesting stmts.
6603 2018-04-27 Martin Jambor <mjambor@suse.cz>
6606 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6607 the jump function allows for passing through aggregate values.
6609 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6611 * input.h (in_system_header_at): Convert from macro to inline
6613 (from_macro_expansion_at): Likewise.
6614 (from_macro_definition_at): Likewise.
6616 2018-04-27 Jeff Law <law@redhat.com>
6618 * config.gcc: Mark tile* targets as deprecated/obsolete.
6620 2018-04-27 Richard Biener <rguenther@suse.de>
6622 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6625 2018-04-27 Richard Biener <rguenther@suse.de>
6627 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6629 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6631 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6632 with Yd constraint. Set "preferred_for_speed" attribute from
6633 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6635 (*movdi_internal): Ditto.
6636 (movti_interunit splitters): Remove
6637 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6638 (movdi_interunit splitters): Ditto.
6639 * config/i386/constraints.md (Ye): Remove.
6640 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6642 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6645 * config/aarch64/constraints.md (Usg): Limit to 31.
6648 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6650 PR tree-optimization/85529
6651 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6652 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6653 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6654 zero extension or masking of the MSB bit.
6655 (optimize_range_tests): Add FIRST_BB argument, pass it through
6656 to optimize_range_tests_var_bound.
6657 (maybe_optimize_range_tests, reassociate_bb): Adjust
6658 optimize_range_tests callers.
6660 2018-04-26 Richard Biener <rguenther@suse.de>
6661 Jakub Jelinek <jakub@redhat.com>
6663 * cgraph.h (symbol_table): Just declare debug method here.
6664 * symtab.c (symbol_table::debug): Define.
6666 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6668 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6670 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6672 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6673 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6674 (*movdi_internal): Substitute Yi and Yj constraint with x
6675 and Ym and Yn constraint with y constraint. Update "isa"
6676 attribute and set "preferred_for_speed" attribute from
6677 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6678 (*movsi_internal): Ditto.
6679 (*movdf_internal): Ditto.
6680 (*movsf_internal): Ditto.
6681 (*zero_extendsidi2): Ditto.
6682 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6683 (sse2_loadld): Ditto.
6684 (*vec_extract<ssevecmodelower>_0): Ditto.
6685 (*vec_extractv4si_0_zext_sse4): Ditto.
6686 (vec_concatv2di): Ditto.
6687 (*vec_dup<mode>): Ditto.
6688 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6689 * config/i386/constraints.md (Yi): Remove.
6694 2018-04-26 Nathan Sidwell <nathan@acm.org>
6696 * dumpfile.c (dump_open): New.
6697 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6698 (dump_finish): Detect stdio/stderr by value not name.
6700 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6702 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6704 2018-04-26 Tom de Vries <tom@codesourcery.com>
6707 * config/nvptx/nvptx.c (verify_neutering_jumps)
6708 (verify_neutering_labels): New function
6709 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6711 2018-04-26 Tom de Vries <tom@codesourcery.com>
6714 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6715 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6718 2018-04-26 Richard Biener <rguenther@suse.de>
6719 Tom de Vries <tom@codesourcery.com>
6722 * lto-streamer-out.c (output_function): Fixup loops if required to match
6723 discovery done in the reader.
6725 2018-04-26 Richard Biener <rguenther@suse.de>
6727 PR tree-optimization/85116
6728 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6729 have a loop exit from the single latch predecessor. Remove
6730 case of header with just condition.
6731 (ch_base::copy_headers): Exclude infinite loops from any
6733 (pass_ch::execute): Record exits.
6735 2018-04-26 Richard Biener <rguenther@suse.de>
6737 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6738 prologue cost vector and pass it to vect_get_load_cost.
6739 (vect_get_peeling_costs_all_drs): Likewise.
6740 (vect_peeling_hash_get_lowest_cost): Likewise.
6741 (vect_enhance_data_refs_alignment): Likewise.
6743 2018-04-26 Richard Biener <rguenther@suse.de>
6746 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6747 checking of integer<->pointer conversions.
6748 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6749 sign-/zero-extending pointer types.
6750 (expand_omp_for_static_chunk): Likewise.
6752 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6753 Jean Lee <xiaoyur347@gmail.com>
6755 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6756 (TARGET_ASAN_SHADOW_OFFSET): Define.
6757 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6758 true for -fsanitize=address.
6760 2018-04-25 Mark Wielaard <mark@klomp.org>
6762 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6765 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6767 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6768 than "alu", remove explicit "memory" and "imm_disp" attributes.
6769 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6772 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6773 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6774 gen_lowpart_no_emit.
6776 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6779 * config/i386/i386.c (ix86_expand_builtin): Change memory
6780 operand to XI, extend p0 to Pmode.
6781 * config/i386/i386.md: Change unspec volatile and operand
6782 1 mode to XI, change operand 0 mode to P.
6784 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6786 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6787 GET_MODE_MASK before any checking.
6788 (nds32_can_use_bset_p): Likewise.
6789 (nds32_can_use_btgl_p): Likewise.
6791 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6793 * config/nds32/nds32-doubleword.md: New define_split pattern for
6794 illegal register number.
6796 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6798 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6800 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6802 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6804 2018-04-25 Richard Biener <rguenther@suse.de>
6806 * lto-streamer.h (LTO_major_version): Bump to 8.
6808 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6810 * BASE-VER: Set to 9.0.0.
6812 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6814 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6815 in __abskf2 and __powikf2.
6817 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6820 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6821 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6822 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6823 Use the above on operand 2. Reindent.
6824 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6826 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6829 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6831 * config/i386/i386.opt (mcet): Removed.
6832 * doc/install.texi: Remove -mcet documentation.
6833 * doc/invoke.texi: Likewise.
6835 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6838 * doc/install.texi: Remove -mcet from bootstrap-cet.
6840 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6843 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6844 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6848 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6849 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6850 containing a CONST_VECTOR.
6852 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6854 * doc/install.texi: Update newlib dependency for nvptx.
6856 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6859 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6860 instead of INTVAL when shifting x left.
6862 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6864 PR tree-optimization/85478
6865 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6866 vect_grouped_store_supported for single element vectors.
6868 2018-04-24 Richard Biener <rguenther@suse.de>
6871 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6872 load cost increase to the case of non-constant step.
6874 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6877 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6878 destination if any_malformed_asm.
6880 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6883 * expr.c (store_field): In the bitfield case, if the value comes from
6884 a function call and is returned in registers by means of a PARALLEL,
6885 do not change the mode of the temporary unless BLKmode and VOIDmode.
6887 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6889 PR rtl-optimization/85423
6890 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6891 dependencies to debug insns when the previous insn is non-debug.
6893 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6895 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6896 enums into a single definition.
6897 (fls): Fix predicates and printing.
6900 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6902 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6903 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6904 and short u6 immediate.
6905 (check_if_valid_sleep_operand): Remove.
6906 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6908 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6910 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6911 flag_always_save_lp condition.
6912 * config/nds32/nds32.opt (malways-save-lp): New option.
6914 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6916 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6917 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6918 * config/nds32/nds32.h
6919 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6920 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6922 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6924 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6925 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6927 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6928 Chung-Ju Wu <jasonwucj@gmail.com>
6930 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6932 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6933 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6935 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6937 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6939 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6941 * config/nds32/nds32-protos.h (nds32_data_alignment,
6942 nds32_local_alignment): Declare.
6943 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6944 nds32_local_alignment): New functions.
6945 (TARGET_CONSTANT_ALIGNMENT): Define.
6946 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6948 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6950 * config/nds32/nds32.c
6951 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6952 (TARGET_MODES_TIEABLE_P): Likewise.
6954 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6956 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6959 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6960 Chung-Ju Wu <jasonwucj@gmail.com>
6962 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6963 for unaligned access.
6964 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6966 * config/nds32/nds32-intrinsic.md: Likewise.
6967 * config/nds32/nds32_intrinsic.h: Likewise.
6968 * config/nds32/nds32.h (nds32_builtins): Likewise.
6969 * config/nds32/nds32.opt (munaligned-access): New option.
6970 * config/nds32/nds32.c (nds32_asm_file_start): Display
6971 flag_unaligned_access status.
6973 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6975 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6976 -mno-relax is present.
6977 * config/riscv/linux.h (LINK_SPEC): Ditto.
6979 2018-04-20 Martin Sebor <msebor@redhat.com>
6982 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6984 (gimple_fold_builtin_stxcpy_chk): Same.
6985 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6987 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6990 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6991 __powikf2 when long double is IEEE 128-bit.
6993 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6995 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6996 step to make sure stack always aligned.
6998 2018-04-20 Carl Love <cel@us.ibm.com>
7001 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7002 size check for arg0.
7004 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7005 Tom de Vries <tom@codesourcery.com>
7008 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7009 Emit insns for calls too.
7010 (nvptx_find_par): Always look for worker-level predecessor insn.
7011 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7013 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7014 (nvptx_process_pars): Propagate frames for calls.
7016 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7019 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7021 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7022 (ix86_handle_option): Don't handle OPT_mibt.
7023 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7025 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7027 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7028 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7029 (ix86_target_macros): Define __CET__ with flag_cf_protection
7030 for -fcf-protection.
7031 * config/i386/i386.c (isa2_opts): Remove -mibt.
7032 * config/i386/i386.h (TARGET_IBT): Removed.
7033 (TARGET_IBT_P): Likewise.
7034 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7035 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7036 * config/i386/i386.opt (mcet): Update help message.
7039 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7040 -mcet as an alias for -mshstk.
7042 2018-04-20 Richard Biener <rguenther@suse.de>
7045 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7046 complexity by forcing a single use of the multiply operand.
7048 2018-04-20 Martin Jambor <mjambor@suse.cz>
7051 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7052 recursion dependency to only apply to non-clones.
7054 2018-04-20 Martin Jambor <mjambor@suse.cz>
7057 * ipa-cp.c (create_specialized_node): Check that clones of
7058 self-recursive edges exist during IPA-CP.
7060 2018-04-19 Toon Moene <toon@moene.org>
7062 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7065 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7067 PR tree-optimization/85467
7068 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7069 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7070 VECTOR_CST element to type.
7072 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7075 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7076 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7077 (builtin_longjmp): Likewise.
7078 (save_stack_nonlocal): New pattern.
7079 (restore_stack_nonlocal): Likewise.
7081 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7084 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7085 Replace ASM_OUTPUT_LABEL with fprintf.
7087 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7090 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7091 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7092 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7093 define __IBT__ and __SHSTK__ for -fcf-protection.
7094 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7096 (ix86_trampoline_init): Likewise.
7097 (x86_output_mi_thunk): Likewise.
7098 (ix86_notrack_prefixed_insn_p): Likewise.
7099 (ix86_option_override_internal): Don't disallow -fcf-protection.
7100 * config/i386/i386.md (rdssp<mode>): Also enable for
7102 (incssp<mode>): Likewise.
7103 (nop_endbr): Likewise.
7104 * config/i386/i386.opt (mcet): Change help message to built-in
7108 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7109 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7110 enable CET built-in functions.
7112 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7114 * common/config/i386/i386-common.c
7115 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7116 OPTION_MASK_ISA_MOVDIRI_UNSET,
7117 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7118 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7119 * config.gcc (movdirintrin.h): New header.
7120 * config/i386/cpuid.h (bit_MOVDIRI,
7121 bit_MOVDIR64B): New bits.
7122 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7124 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7125 (VOID, PVOID, PCVOID)): New function types.
7126 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7127 __builtin_ia32_directstoreu_u64,
7128 __builtin_ia32_movdir64b): New builtins.
7129 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7130 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7132 (ix86_valid_target_attribute_inner_p): Ditto.
7133 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7134 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7135 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7136 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7137 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7138 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7139 (movdiri<mode>, movdir64b_<mode>): New.
7140 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7141 * config/i386/immintrin.h: Include movdirintrin.h.
7142 * config/i386/movdirintrin.h: New file.
7143 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7145 2018-04-19 Richard Biener <rguenther@suse.de>
7148 * cfg.c (clear_bb_flags): When loop state says we have
7149 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7151 2018-04-19 Richard Biener <rguenther@suse.de>
7153 PR tree-optimization/84737
7154 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7155 copying restrict info.
7156 (vect_setup_realignment): Use it.
7157 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7158 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7159 the first DR to all generated stores.
7160 (vectorizable_load): Likewise for loads.
7162 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7164 PR tree-optimization/85446
7165 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7166 the integral and pointer types to have the same precision.
7168 * doc/install.texi: Document --disable-cet being the default and
7171 2018-04-18 Martin Liska <mliska@suse.cz>
7173 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7176 2018-04-18 Martin Liska <mliska@suse.cz>
7179 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7182 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7183 arguments if they are comparable.
7185 2018-04-18 Martin Liska <mliska@suse.cz>
7188 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7191 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7194 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7197 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7198 ENDBR after calling __morestack.
7200 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7203 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7204 by using gcc_base_ver to generate a gcc_driver_version, and use
7205 it when generating GCC_DRIVER_NAME.
7206 * configure: Regenerate.
7208 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7211 * config.gcc: Obsolete powerpc*-*-*spe*.
7213 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7216 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7217 (stabstr_D): Change type of unum from unsigned int to
7218 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7221 2018-04-17 Jim Wilson <jimw@sifive.com>
7224 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7225 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7226 Set arg_pointer_offset after using pretend_args_size.
7228 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7230 PR rtl-optimization/85431
7231 * dse.c (record_store): Ignore zero width stores.
7234 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7235 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7236 __builtin_stack_restore rather than after it.
7237 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7238 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7239 argument instead of virtual_dynamic_stack_rtx.
7241 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7243 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7245 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7246 Add note to error message to explain internal mapping of overloaded
7247 built-in function name to non-overloaded built-in function name.
7248 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7251 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7254 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7255 where the inputs overlap with the output.
7257 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7259 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7260 (=v, v) alternative and explicit "memory" attribute.
7261 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7262 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7264 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7265 "sselog1" type instead of "sselog".
7266 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7267 "sselog". Remove explicit "memory" attribute.
7268 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7269 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7271 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7272 "sselog1" type instead of "sselog". Remove explicit "memory"
7274 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7275 use "sselog1" type instead of "sselog". Remove explicit "memory"
7277 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7278 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7280 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7281 "sselog1" type instead of "sselog". Remove explicit "memory"
7283 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7284 use "sselog1" type instead of "sselog". Remove explicit "memory"
7288 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7291 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7294 2018-04-17 Martin Jambor <mjambor@suse.cz>
7297 * ipa-cp.c (create_specialized_node): Call
7298 expand_all_artificial_thunks if necessary.
7300 2018-04-17 Martin Liska <mliska@suse.cz>
7303 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7304 in message, remote space in between '_G' and '('.
7306 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7309 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7310 avx512f_vmcmp<mode>3<round_saeonly_name>,
7311 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7312 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7313 avx512f_rndscale<mode><round_saeonly_name>,
7314 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7315 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7316 Use %<iptr>2 instead of %2 for -masm=intel.
7317 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7318 avx512f_vcvttss2usi<round_saeonly_name>,
7319 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7321 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7322 avx512f_vcvttsd2usi<round_saeonly_name>,
7323 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7324 Use %q1 instead of %1 for -masm=intel.
7325 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7326 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7327 of %3 for -masm=intel.
7328 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7330 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7332 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7334 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7335 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7337 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7339 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7340 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7341 %g1 and one with %0 and %1.
7342 (avx512er_vmrcp28<mode><round_saeonly_name>,
7343 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7345 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7346 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7347 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7348 of %0 and %{%4%} for -masm=intel.
7349 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7350 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7351 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7352 order of %0 and %{%5%}%{z%} for -masm=intel.
7354 2018-04-17 Jan Hubicka <jh@suse.cz>
7357 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7359 2018-04-17 Martin Liska <mliska@suse.cz>
7362 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7363 for target_clone error message. Make default implementation
7364 clone to be a local declaration.
7365 (separate_attrs): Add new argument and check for an empty
7367 (expand_target_clones): Handle it.
7368 (ipa_target_clone): Make redirection just for target_clones
7371 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7372 Tom de Vries <tom@codesourcery.com>
7375 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7376 tiled basic blocks without omp continue statements.
7378 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7381 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7382 vec_extract expression as having side effects to make sure it gets
7385 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7388 * config/i386/i386.c (get_builtin_code_for_version): Check
7391 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7394 * gcc/config.gcc: Support "skylake".
7395 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7397 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7398 (processor_target_table): Add "skylake".
7399 (ix86_option_override_internal): Add "skylake".
7400 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7401 PROCESSOR_CANNONLAKE.
7402 (get_builtin_code_for_version): Fix priority for
7403 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7404 PROCESSOR_SKYLAKE-AVX512.
7405 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7406 (processor_type): Add PROCESSOR_SKYLAKE.
7408 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7409 Jason Merrill <jason@redhat.com>
7412 * convert.c (convert_to_integer_1): Use direct recursion for
7413 enumeral types and types with a precision less than the number
7414 of bits in their mode.
7416 2018-04-16 Julia Koval <julia.koval@intel.com>
7419 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7420 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7422 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7425 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7426 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7427 and -mno-direct-move.
7429 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7432 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7433 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7434 Ensure negative shifts result in {0}.
7436 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7438 PR rtl-optimization/79916
7439 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7440 regs (if any) to define how to gnerate SD moves when LRA is in
7443 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7445 PR rtl-optimization/85393
7446 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7447 * except.c (expand_dw2_landing_pad_for_region): Make static.
7448 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7449 a label and unconditional jump to old_bb, rather than
7450 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7453 PR rtl-optimization/85376
7454 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7455 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7456 instead of a specific value.
7458 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7459 Bin Cheng <bin.cheng@arm.com>
7461 PR tree-optimization/82965
7462 PR tree-optimization/83991
7463 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7464 by_profile_only parameter.
7465 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7466 information if the loop was predicted to iterate too many times.
7467 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7469 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7472 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7475 2018-04-13 Martin Liska <mliska@suse.cz>
7476 Jakub Jelinek <jakub@redhat.com>
7479 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7480 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7481 * builtins.c (expand_builtin_memory_copy_args): Use
7482 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7483 handle dest_addr == pc_rtx.
7485 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7488 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7489 asked to not generate direct moves.
7490 (fix_trunc<mode>si2_stfiwx): Similar.
7491 (fix_trunc<mode>si2_internal): Similar.
7493 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7496 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7497 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7498 lookup if dest in some wider mode is known to be const0_rtx and
7499 if so, record permanent equivalence for it to be ZERO_EXTEND of
7500 the narrower mode destination.
7502 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7504 * lto-streamer-out.c (output_function): Revert 259346.
7505 * omp-expand.c (expand_oacc_for): Likewise.
7507 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7509 PR rtl-optimization/85354
7510 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7511 * sel-sched.c (sel_global_init): ... here.
7513 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7516 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7517 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7518 mode for PE-COFF targets.
7519 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7520 (i386_pe_asm_lto_end): Likewise.
7521 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7522 (TARGET_ASM_LTO_END): Likewise.
7523 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7524 (i386_pe_asm_lto_start): New function.
7525 (i386_pe_asm_lto_end): Likewise.
7527 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7528 Richard Biener <rguenther@suse.de>
7531 * lto-streamer-out.c (output_function): Fix CFG loop state before
7533 * omp-expand.c (expand_oacc_for): Handle calls to internal
7534 functions like regular functions.
7536 2018-04-12 Richard Biener <rguenther@suse.de>
7539 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7540 for the early LTO debug to properly generate references to it
7541 during DIE emission. Do not re-use that for the skeleton for
7543 (dwarf2out_early_finish): Likewise.
7545 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7548 * config/i386/sse.md
7549 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7550 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7551 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7552 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7553 and output is a reg, avoid creating invalid lowpart subreg, but
7554 instead split into a 512-bit move. Don't split if not AVX512VL,
7555 input is xmm16+ reg and output is a mem.
7556 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7557 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7558 xmm16+ reg and output is a mem.
7560 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7562 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7563 also for flag_dwarf2_cfi_asm.
7565 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7567 PR rtl-optimization/85342
7568 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7569 a bool scalar var inside of the loop instead. Don't try to update
7570 recog_data.operand after failed apply_change_group.
7572 2018-04-12 Tom de Vries <tom@codesourcery.com>
7575 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7576 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7577 array with flexible array member as array without given dimension.
7578 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7579 argument for undefined param to true.
7581 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7584 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7585 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7586 from PowerPC section.
7587 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7588 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7589 help text that is too long.
7590 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7591 help text that is too long.
7592 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7593 help text that is too long.
7595 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7597 * config/alpha/alpha.md (stack_probe_internal): Rename
7598 from "probe_stack". Update all callers.
7600 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7602 PR rtl-optimization/84566
7603 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7604 sched_macro_fuse_insns.
7606 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7609 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7610 (compute_block_dependences): ... from here.
7612 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7614 PR tree-optimization/85331
7615 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7616 from int to HOST_WIDE_INT.
7618 2018-04-11 Martin Jambor <mjambor@suse.cz>
7621 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7622 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7623 not the same as the source val.
7624 (cgraph_edge_brings_value_p): New parameter.
7625 (gather_edges_for_value): Pass destination value to
7626 cgraph_edge_brings_value_p.
7627 (perhaps_add_new_callers): Likewise.
7628 (get_info_about_necessary_edges): Likewise and exclude values brought
7629 only by self-recursive edges.
7630 (create_specialized_node): Redirect only clones of self-calling edges.
7631 (+self_recursive_pass_through_p): New function.
7632 (find_more_scalar_values_for_callers_subset): Use it.
7633 (find_aggregate_values_for_callers_subset): Likewise.
7634 (known_aggs_to_agg_replacement_list): Removed.
7635 (decide_whether_version_node): Re-calculate known constants for all
7636 remaining context clones.
7638 2018-04-11 Richard Biener <rguenther@suse.de>
7641 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7642 from early DWARF output.
7643 (dwarf2out_early_finish): Output line info unconditionally into
7644 early DWARF and add reference to it.
7646 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7649 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7650 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7651 other than V2DFmode using iptr mode attribute.
7652 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7654 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7656 PR rtl-optimization/84659
7657 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7659 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7662 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7664 (output_loc_list): Pass address of a dummy size variable even in the
7665 locview handling loop.
7666 (index_location_lists): Add comment on why skip_loc_list_entry can't
7669 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7672 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7675 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7678 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7679 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7680 and -mstring-compare-inline-limit.
7682 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7685 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7686 for stack clash protection in a register whenever we need it to be in
7689 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7691 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7692 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7694 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7697 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7699 (mlong-double-): Ditto.
7700 * config/rs6000/sysv4.opt (msdata=): Ditto.
7701 (mtls-size=): Ditto.
7703 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7705 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7706 erroneous entries for
7707 "vector int vec_ldl (int, long int *)", and
7708 "vector unsigned int vec_ldl (int, unsigned long int *)".
7709 Add comments and entries for
7710 "vector bool char vec_ldl (int, bool char *)",
7711 "vector bool short vec_ldl (int, bool short *)",
7712 "vector bool int vec_ldl (int, bool int *)",
7713 "vector bool long long vec_ldl (int, bool long long *)",
7714 "vector pixel vec_ldl (int, pixel *)",
7715 "vector long long vec_ldl (int, long long *)",
7716 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7717 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7718 type tree bool_long_long_type_node and correct definition of
7719 bool_V2DI_type_node to make reference to this new type tree.
7720 (rs6000_mangle_type): Replace erroneous reference to
7721 bool_long_type_node with bool_long_long_type_node.
7722 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7723 comments to emphasize sign distinctions for char and int types and
7724 replace RS6000_BTI_bool_long constant with
7725 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7726 use of RS6000_BTI_pixel.
7727 (bool_long_type_node): Remove this macro definition.
7728 (bool_long_long_type_node): New macro definition
7730 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7732 PR rtl-optimization/85300
7733 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7734 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7735 simplify_unary_operation fails.
7737 2018-04-10 Martin Liska <mliska@suse.cz>
7739 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7740 cgraph_edge and ipa_ref.
7742 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7746 * config/i386/sse.md
7747 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7748 computation of the VEC_MERGE selector from mask.
7749 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7750 Fix decoding of the VEC_MERGE selector into mask.
7752 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7754 PR tree-optimization/85286
7755 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7757 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7759 * final.c (final_1): Set insn_last_address as well as
7760 insn_current_address.
7762 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7765 * explow.c (emit_stack_probe): Call validize_mem on memory location
7766 before passing it to gen_probe_stack. Create address operand and
7767 legitimize it for the probe_stack_address case.
7769 2018-04-09 Jan Hubicka <jh@suse.cz>
7772 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7773 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7774 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7776 (free_lang_data): Rebuild type inheritance graph.
7778 2018-04-09 Martin Sebor <msebor@redhat.com>
7780 * invoke.texi (-finline-small-functions): Mention other optimization
7782 (-findirect-inlining, -fpartial-inlining): Same.
7783 (-finline-functions-called-once): Same.
7784 (-freorder-blocks-and-partition): Same.
7786 2018-04-09 Jan Hubicka <jh@suse.cz>
7789 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7790 jumps; choose last target that matches the criteria (i.e.
7791 no partition changes for non-crossing jumps).
7792 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7793 support for redirecting crossing jumps to non-crossing.
7795 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7797 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7798 also for naked functions.
7800 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7802 * config/arc/arc.md (add_shift): New pattern.
7803 (add_shift2): Likewise.
7804 (sub_shift): Likewise.
7805 (sub_shift_cmp0_noout): Likewise.
7806 (compare_si_ashiftsi): Likewise.
7807 (xbfu_cmp0_noout): New combine pattern.
7808 (xbfu_cmp0"): Likewise.
7809 (movsi_set_cc_insn): Place the predicable variant first.
7810 (commutative_binary_cmp0_noout): Remove clobber.
7811 (commutative_binary_cmp0): New pattern.
7812 (noncommutative_binary_cmp0): Likewise.
7813 (noncommutative_binary_cmp0_noout): Likewise.
7814 (noncommutative_binary_comparison_result_used): Removed.
7815 (rsub_cmp0): New pattern.
7816 (rsub_cmp0_noout): Likewise.
7817 (extzvsi): Changed, keep only meaningful variants.
7818 (SQH, SEZ): New iterators.
7819 (SQH_postfix): New mode attribute.
7820 (SEZ_prefix): New code attribute.
7821 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7822 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7823 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7825 (noncommutative_operator): Check the availability of barrel
7828 2018-04-09 Richard Biener <rguenther@suse.de>
7830 PR tree-optimization/85284
7831 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7832 Only use the niter constraining form of simple_iv when the exit
7835 2018-04-09 Tom de Vries <tom@codesourcery.com>
7838 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7839 (define_expand "*memory_barrier"): New define_expand.
7840 (define_insn "memory_barrier"): New insn.
7842 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7844 PR rtl-optimization/80463
7845 PR rtl-optimization/83972
7846 PR rtl-optimization/83480
7848 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7849 correct producer for the insn.
7850 (tidy_control_flow): Fixup seqnos in case of debug insns.
7852 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7854 PR rtl-optimization/83913
7856 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7857 different sched-times when merging exprs.
7859 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7861 PR rtl-optimization/83962
7863 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7864 tidy_fallthru_edge and tidy_control_flow.
7866 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7868 PR rtl-optimization/83530
7870 * sel-sched.c (force_next_insn): New global variable.
7871 (remove_insn_for_debug): When force_next_insn is true, also leave only
7872 next insn in the ready list.
7873 (sel_sched_region): When the region wasn't scheduled, make another pass
7874 over it with force_next_insn set to 1.
7876 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7878 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7880 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7881 for interrupt control.
7882 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7883 functions for interrupt control.
7884 * config/nds32/nds32-intrinsic.md: Likewise.
7885 * config/nds32/nds32_intrinsic.h: Likewise.
7886 * config/nds32/nds32.h (nds32_builtins): Likewise.
7888 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7890 * config/nds32/nds32.c (nds32_init_machine_status,
7891 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7892 strict_aligned_p field.
7893 (nds32_expand_to_rtl_hook): New function.
7894 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7895 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7897 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7898 Chung-Ju Wu <jasonwucj@gmail.com>
7900 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7901 * config/nds32/nds32-n7.md: New file.
7902 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7903 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7905 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7906 * config/nds32/nds32.md (pipeline_model): Add n7.
7907 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7908 * config/nds32/pipelines.md: Include n7 settings.
7910 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7911 Chung-Ju Wu <jasonwucj@gmail.com>
7913 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7914 * config/nds32/nds32-e8.md: New file.
7915 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7916 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7918 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7919 * config/nds32/nds32.md (pipeline_model): Add e8.
7920 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7921 * config/nds32/pipelines.md: Include e8 settings.
7923 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7924 Chung-Ju Wu <jasonwucj@gmail.com>
7926 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7927 * config/nds32/nds32-n8.md: New file.
7928 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7929 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7931 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7932 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7933 * config/nds32/nds32.md (pipeline_model): Add n8.
7934 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7935 * config/nds32/pipelines.md: Include n8 settings.
7937 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7938 Chung-Ju Wu <jasonwucj@gmail.com>
7940 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7941 * config/nds32/nds32-n9-2r1w.md: New file.
7942 * config/nds32/nds32-n9-3r2w.md: New file.
7943 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7944 nds32_register_ports): New or modify for cpu n9.
7945 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7947 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7948 * config/nds32/nds32-utils.c: New file.
7949 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7950 TARGET_MUL_SLOW): Define.
7951 * config/nds32/nds32.md (pipeline_model): New attribute.
7952 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7953 New options that support cpu n9.
7954 * config/nds32/pipelines.md: Include n9 settings.
7955 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7957 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7959 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7960 information if necessary.
7961 (output_cond_branch_compare_zero): Likewise.
7962 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7963 (nds32_target_alignment): Refine for alignment.
7964 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7965 (FUNCTION_BOUNDARY): Modify.
7966 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7968 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7970 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7972 * config/nds32/constants.md (unspec_volatile_element): Add values for
7973 TLB operation and data prefetch.
7974 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7975 functions for TLB operation and data prefetch.
7976 * config/nds32/nds32-intrinsic.md: Likewise.
7977 * config/nds32/nds32_intrinsic.h: Likewise.
7978 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7979 (nds32_print_operand): Likewise.
7980 * config/nds32/nds32.h (nds32_builtins): Likewise.
7982 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7983 Andrew Pinski <pinsika@gcc.gnu.org>
7986 * match.pd: Use constant_boolean_node of correct type instead of
7987 boolean_true_node or boolean_false_node for simplifying
7988 pointer comparisons to zero.
7990 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7992 PR tree-optimization/80021
7993 * tree.c (verify_type_variant): Make error call in verify_variant_match
7994 translatable and remove final full stop.
7996 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7998 * config/nds32/constants.md (unspec_volatile_element): Add
7999 UNSPEC_VOLATILE_EH_RETURN.
8000 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8001 nds32_output_stack_pop): Support dwarf exception handling process.
8002 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8003 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8004 exception handling process.
8005 (nds32_compute_stack_frame): Likewise.
8006 (nds32_return_addr_rtx): Likewise.
8007 (nds32_initial_elimination_offset): Likewise.
8008 (nds32_expand_prologue): Likewise.
8009 (nds32_expand_epilogue): Likewise.
8010 (nds32_dynamic_chain_address): New function.
8011 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8013 (DYNAMIC_CHAIN_ADDRESS): Define.
8014 (EH_RETURN_DATA_REGNO): Define.
8015 (EH_RETURN_STACKADJ_RTX): Define.
8016 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8017 patterns for dwarf exception handling.
8019 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8021 * config/nds32/nds32.h: Clean up obsolete macros.
8023 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8025 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8026 Add enum values for particular instructions.
8027 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8028 particular intrinsic functions.
8029 * config/nds32/nds32-intrinsic.md: Likewise.
8030 * config/nds32/nds32_intrinsic.h: Likewise.
8031 * config/nds32/nds32.h (nds32_builtins): Likewise.
8032 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8033 (btst, ave): New patterns for particular instructions.
8035 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8037 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8038 Add enum values for atomic load/store and memory sync.
8039 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8041 * config/nds32/nds32-intrinsic.md: Likewise.
8042 * config/nds32/nds32_intrinsic.h: Likewise.
8043 * config/nds32/nds32.h (nds32_builtins): Likewise.
8045 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8047 PR tree-optimization/85257
8048 * fold-const.c (native_encode_vector): If not all elts could fit
8049 and off is -1, return 0 rather than offset.
8050 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8051 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8052 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8053 adjust buffer in native_interpret_expr call.
8055 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8057 * config/nds32/constants.md (unspec_volatile_element): Add cache
8058 control enum values.
8059 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8060 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8061 * config/nds32/nds32.c (nds32_cctl_names): New.
8062 (nds32_print_operand): Handle cache control register names.
8063 * config/nds32/nds32.h (nds32_builtins): New enum values.
8064 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8066 * config/nds32/nds32.md (type): Add mmu.
8067 * config/nds32/pipelines.md (simple_insn): Add mmu.
8069 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8071 * config/nds32/nds32.md (type): Remove call.
8072 * config/nds32/pipelines.md (simple_insn): Likewise.
8074 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8076 * config/nds32/constants.md (unspec_volatile_element): Add
8077 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8078 UNSPEC_VOLATILE_FMFCFG.
8079 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8080 description for fmfcfg and fmfcsr.
8081 (bdesc_1arg): Add fmtcsr.
8082 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8083 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8084 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8085 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8086 unspec_fmfcfg): New patterns.
8087 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8088 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8089 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8090 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8091 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8092 __nds32__fmfcfg): Define.
8094 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8096 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8097 intrinsic register names.
8098 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8099 intrinsic register enum values and macros.
8101 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8103 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8104 for load/store addressing form.
8105 (nds32_print_operand_address): Likewise.
8107 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8110 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8111 based on LABEL_REF. Remove useless assertion.
8112 (pic_address_needs_scratch): Fix formatting.
8113 (sparc_legitimize_pic_address): Minor tweaks.
8114 (sparc_delegitimize_address): Adjust assertion accordingly.
8115 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8116 into symbolic_operand.
8117 (movsi_high_pic_label_ref): Likewise.
8118 (movsi_lo_sum_pic_label_ref): Likewise.
8119 (movdi_pic_label_ref): Likewise.
8120 (movdi_high_pic_label_ref): Likewise.
8121 (movdi_lo_sum_pic_label_ref): Likewise.
8123 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8125 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8126 custom LIB_SPEC setup.
8128 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8129 Kito Cheng <kito.cheng@gmail.com>
8131 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8132 * config/riscv/freebsd.h: New.
8134 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8136 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8137 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8140 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8141 Kito Cheng <kito.cheng@gmail.com>
8143 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8144 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8145 * config/nds32/nds32-protos.h (nds32_output_call,
8146 nds32_output_return): Declare.
8147 * config/nds32/nds32.md: Refine all the call and return patterns.
8149 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8152 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8153 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8155 PR rtl-optimization/84872
8156 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8157 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8160 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8162 * expr.c (copy_blkmode_to_reg): Revert 254862.
8163 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8165 2018-04-06 Richard Biener <rguenther@suse.de>
8168 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8169 after seeing a component reference with an adjacent field. Treat
8170 refs to arrays at struct end of external decls similar to
8171 refs to unconstrained commons.
8173 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8176 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8177 look through SAVE_EXPRs with non-side-effects argument. Adjust
8179 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8182 2018-04-06 Richard Biener <rguenther@suse.de>
8185 * alias.c (find_base_term): New wrapper around find_base_term
8186 unwinding CSELIB_VAL_PTR changes.
8187 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8190 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8192 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8194 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8195 constant definitions.
8196 ("nop"): lr 0,0 -> nopr r0
8197 ("nop_lr0", "nop_lr1"): New insn definitions.
8199 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8201 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8202 NDS32_V3PUSH_AVAILABLE_P macro.
8204 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8205 Chung-Ju Wu <jasonwucj@gmail.com>
8207 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8208 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8209 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8210 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8211 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8212 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8213 * config/nds32/constraints.md: New constraints and checking for hard
8214 float configuration.
8215 * config/nds32/iterators.md: New mode iterator and attribute for hard
8216 float configuration.
8217 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8219 * config/nds32/nds32-fpu.md: New file.
8220 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8221 deal with hard float code generation.
8222 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8224 (abi_type, float_reg_number): New enum type.
8225 * config/nds32/nds32-predicates.c: New predicates for hard float.
8226 * config/nds32/nds32-protos.h: Declare functions for hard float.
8227 * config/nds32/nds32.c: Implementation for hard float configuration.
8228 * config/nds32/nds32.h: Definitions for hard float configuration.
8229 * config/nds32/nds32.md: Include hard float machine description and
8230 modify patterns for hard float configuration.
8231 * config/nds32/nds32.opt: New options for hard float configuration.
8232 * config/nds32/predicates.md: New predicates for hard float
8235 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8237 * common/config/nds32/nds32-common.c
8238 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8240 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8243 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8244 CONSTRUCTOR_ELT (ctor, ...)->value.
8246 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8249 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8251 2018-04-05 Tom de Vries <tom@codesourcery.com>
8254 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8257 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8258 Kito Cheng <kito.cheng@gmail.com>
8260 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8261 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8262 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8263 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8265 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8266 Kito Cheng <kito.cheng@gmail.com>
8268 * config/nds32/constraints.md (Ufe): New memory constraint.
8269 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8270 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8271 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8273 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8274 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8276 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8278 * config/nds32/nds32.md: Use optimize_size in the condition for
8279 alu-shift instructions.
8281 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8283 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8285 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8287 * config/nds32/nds32.md (negsi2): Refine pattern.
8289 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8290 Chung-Ju Wu <jasonwucj@gmail.com>
8292 * config/nds32/iterators.md (shift_rotate): New code iterator.
8293 (shift): New code attribute.
8294 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8295 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8296 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8297 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8298 bit-wise operations.
8299 (andsi3, *andsi3): Ditto.
8300 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8301 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8302 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8303 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8304 nds32_ior_operand, nds32_xor_operand): New predicates.
8306 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8308 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8309 (addsi3, subsi3): ... this.
8311 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8313 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8315 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8317 * config/nds32/nds32.md: Adjust indention.
8319 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8321 * config/nds32/nds32.md (feature): New attribute.
8323 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8325 * config/nds32/nds32.md (subtype): New attribute.
8327 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8330 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8331 expansion to perform a bitwise AND of the argument followed by a
8332 boolean negation of the result.
8334 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8336 PR rtl-optimization/84878
8337 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8338 the basic block. Assert the use reference is not artificial and that
8339 it has an associated insn.
8341 2018-04-04 Michael Matz <matz@suse.de>
8343 * builtins.c (compute_objsize): Pass correct operand
8344 to array_at_struct_end_p.
8346 2018-04-04 Richard Biener <rguenther@suse.de>
8349 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8350 from contexts for DINFO_LEVEL_TERSE and below.
8352 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8354 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8355 resiter_operand condition.
8356 * config/nds32/nds32.md (*move<mode>): Ditto.
8358 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8359 Monk Chiang <sh.chiang04@gmail.com>
8361 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8363 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8365 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8367 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8368 Kito Cheng <kito.cheng@gmail.com>
8370 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8371 nds32_cond_code_str, output_cond_branch,
8372 output_cond_branch_compare_zero, nds32_expand_cbranch,
8373 nds32_expand_cstore, nds32_expand_movcc,
8374 nds32_output_cbranchsi4_equality_zero,
8375 nds32_output_cbranchsi4_equality_reg,
8376 nds32_output_cbranchsi4_equality_reg_or_const_int,
8377 nds32_output_cbranchsi4_greater_less_zero: New functions.
8378 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8379 nds32_expand_cstore, nds32_expand_movcc,
8380 nds32_output_cbranchsi4_equality_zero,
8381 nds32_output_cbranchsi4_equality_reg,
8382 nds32_output_cbranchsi4_equality_reg_or_const_int,
8383 nds32_output_cbranchsi4_greater_less_zero): Declare.
8384 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8385 nds32_rimm11s_operand): New predicates.
8386 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8387 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8390 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8392 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8393 * config/nds32/nds32.md: Ditto.
8394 * config/nds32/pipelines.md: Ditto.
8396 2018-04-04 Richard Biener <rguenther@suse.de>
8398 PR tree-optimization/85168
8399 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8400 propagating abnormals.
8402 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8404 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8406 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8407 Kito Cheng <kito.cheng@gmail.com>
8409 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8410 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8411 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8412 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8413 * config/nds32/nds32.md (sibcall_internal): New.
8414 (sibcall_register): Remove.
8415 (sibcall_immediate): Remove.
8416 (sibcall_value_internal): New.
8417 (sibcall_value_register): Remove.
8418 (sibcall_value_immediate): Remove.
8419 * config/nds32/predicates.md (nds32_general_register_operand): New.
8420 (nds32_call_address_operand): New.
8422 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8424 PR rtl-optimization/85167
8425 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8426 bb_defs if *split_p, instead preinitialize it to NULL.
8428 PR tree-optimization/85156
8429 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8430 evaluating the argument multiple times.
8432 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8434 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8436 (_mm_cvtpd_ps): Likewise.
8437 (_mm_cvttpd_epi32): Likewise.
8438 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8439 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8440 vector, pixel, and bool following altivec.h include.
8442 2018-04-03 Martin Sebor <msebor@redhat.com>
8444 * doc/extend.texi (Common Function Attributes): Clarify.
8445 (const attribute): Likewise.
8446 (pure attribute): Likewise.
8448 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8451 * config/i386/i386.c (ix86_expand_vector_set): Use
8452 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8454 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8456 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8457 instructions when changing rounding bits to preserve precision bits
8458 in the x87 control word.
8460 2018-04-03 Martin Liska <mliska@suse.cz>
8462 PR tree-optimization/82491
8463 * rtl.h (strip_offset_and_add): Replace += suboffset with
8464 poly_uint64 () + suboffset.
8466 2018-03-29 Martin Liska <mliska@suse.cz>
8467 Martin Jambor <mjambor@suse.cz>
8470 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8471 param_type is not an integral or pointer type.
8473 2018-04-03 Richard Biener <rguenther@suse.de>
8475 * sese.h (recompute_all_dominators): Remove.
8477 2018-04-02 Martin Sebor <msebor@redhat.com>
8479 * doc/invoke.texi (-Wrestrict): Fix typos.
8481 2018-04-02 Jim Wilson <jimw@sifive.com>
8483 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8484 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8485 (<optab>di3, <optab>si3_extend): Likewise.
8486 (<optab>si3_mask, <optab>si3_mask_1): New.
8487 (<optab>di3_mask, <optab>di3_mask_1): New.
8488 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8489 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8490 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8492 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8494 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8497 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8499 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8500 (nds32_canonicalize_comparison): New function.
8502 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8503 Kito Cheng <kito.cheng@gmail.com>
8504 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8506 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8507 * config/nds32/constants.md (unspec_volatile_element): Add
8508 UNSPEC_VOLATILE_RELAX_GROUP.
8509 * config/nds32/nds32-relax-opt.c: New file.
8510 * config/nds32/nds32-predicates.c
8511 (nds32_symbol_load_store_p): New function.
8512 * config/nds32/nds32-protos.h
8513 (nds32_symbol_load_store_p): Declare function.
8514 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8515 * config/nds32/nds32.c
8516 (nds32_register_pass): New function to register pass.
8517 (nds32_register_passes): New function to register passes.
8518 * config/nds32/nds32.md (relax_group): New pattern.
8519 * config/nds32/nds32.opt (mrelax-hint): New option.
8520 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8522 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8524 * config/nds32/t-nds32: Modify files dependency.
8526 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8528 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8529 (PROFILE_HOOK): Define its implementation.
8531 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8533 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8534 type and 32-bit size.
8536 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8539 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8540 (V_128_256): New mode iterator.
8541 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8542 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8543 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8545 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8546 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8548 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8551 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8552 NaN inputs correctly.
8554 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8557 * config/rs6000/vsx.md (??r): New mode attribute.
8558 (*vsx_mov<mode>_64bit): Use it.
8559 (*vsx_mov<mode>_32bit): Likewise.
8561 2018-03-30 Martin Sebor <msebor@redhat.com>
8563 PR tree-optimization/84818
8564 * builtins.c (check_access): Use warning_n.
8566 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8569 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8571 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8574 2018-03-30 Julia Koval <julia.koval@intel.com>
8577 * x86-tune.def (movx, partial_reg_dependency): Enable for
8580 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8583 * lra-constraints.c (process_alt_operands): Move setting
8584 this_alternative_matches below.
8586 2018-03-29 Martin Liska <mliska@suse.cz>
8589 * doc/invoke.texi: Document how LTO works with debug info.
8590 Describe auto-load support of binutils. Mention 'x86-64'
8591 as valid option value of -march option.
8593 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8595 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8598 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8599 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8600 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8603 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8606 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8607 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8608 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8609 for RS6000_BTM_POWERPC64.
8610 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8611 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8612 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8617 2018-03-28 Carl Love <cel@us.ibm.com>
8620 2017-09-27 Carl Love <cel@us.ibm.com>
8622 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8623 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8624 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8627 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8629 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8630 instead of __vector bool.
8631 (_mm_max_pu8): Likewise.
8632 (_mm_min_pi16): Likewise.
8634 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8637 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8638 (DIVWEUO): Likewise.
8640 (DIVDEUO): Likewise.
8641 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8642 DIVWEUO and DIVDEUO.
8643 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8644 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8645 (div_extend): Likewise.
8646 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8648 (__builtin_divweuo): Likewise.
8649 (__builtin_divdeo): Likewise.
8650 (__builtin_divdeuo): Likewise.
8652 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8655 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8656 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8658 PR tree-optimization/82004
8659 * gimple-match-head.c (optimize_pow_to_exp): New function.
8660 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8661 Don't fold to exp if optimize_pow_to_exp is false.
8663 2018-03-28 Martin Liska <mliska@suse.cz>
8666 * calls.c (initialize_argument_information): Fix trailing space.
8667 * common.opt: Fix typo and provide better explanation for
8668 -fsanitize-coverage option.
8669 * config/i386/i386.opt: Fix typo.
8671 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8672 Martin Liska <mliska@suse.cz>
8675 * gimplify.c (asan_poison_variable): Don't do the check for
8676 gimplify_omp_ctxp here.
8677 (gimplify_decl_expr): Do it here.
8678 (gimplify_target_expr): Likewise.
8680 2018-03-28 Martin Liska <mliska@suse.cz>
8683 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8684 chkp_type_bounds_count if MPX is not enabled.
8686 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8688 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8690 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8693 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8694 function to create the function decl for complex long double
8695 multiply and divide for -mabi=ieeelongdouble.
8696 (init_float128_ieee): Call it.
8698 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8701 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8702 -fcf-protection=branch -mibt.
8703 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8705 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8708 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8709 arm_word_relocations.
8711 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8714 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8715 extern array declarations.
8717 2018-03-27 Richard Biener <rguenther@suse.de>
8720 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8721 explicit single_use checks.
8723 2018-03-27 Richard Biener <rguenther@suse.de>
8725 PR tree-optimization/85082
8726 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8729 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8731 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8732 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8733 Turn on fasynchronous-unwind-tables and funwind-tables.
8735 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8738 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8739 (*bmi_blsr_<mode>_ccz): Ditto.
8741 2018-03-26 Tom de Vries <tom@codesourcery.com>
8743 PR tree-optimization/85063
8744 * omp-general.c (offloading_function_p): New function. Factor out
8746 * omp-offload.c (pass_omp_target_link::gate): ... here.
8747 * omp-general.h (offloading_function_p): Declare.
8748 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8749 with attribute omp declare target for offloading functions.
8751 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8753 PR tree-optimization/84005
8754 * tree-data-ref.h (get_base_for_alignment): Declare.
8755 * tree-data-ref.c (get_base_for_alignment_1): New function.
8756 (get_base_for_alignment): Likewise.
8757 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8758 get_base_for_alignment to find a suitable base object, instead
8759 of always using drb->base_address.
8761 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8764 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8765 known size by default.
8767 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8770 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8771 and non BLKmode operands.
8773 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8776 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8777 Clean up attributes.
8779 2018-03-23 Richard Biener <rguenther@suse.de>
8782 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8783 we are going to emit early debug for LTO.
8785 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8788 * function.c (match_asm_constraints_1): Don't optimize if input
8789 doesn't satisfy general_operand predicate for output's mode.
8792 * alias.c (write_dependence_p): Don't require for x_canonicalized
8793 non-VOIDmode if x has VOIDmode.
8796 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8797 just don't try to optimize it rather than assert it never happens.
8799 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8801 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8802 macro expansions for definition of ST_INTERNAL_<mode> and
8803 LD_INTERNAL_<mode> builtins.
8804 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8806 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8808 (altivec_expand_st_builtin): Likewise.
8809 (altivec_expand_builtin): Remove calls to deleted functions.
8810 (rs6000_address_for_altivec): Delete this function.
8811 * config/rs6000/vector.md: Remove expands for
8812 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8814 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8817 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8818 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8819 re-computing once computed.
8820 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8821 (arm_init_machine_status): Initialize
8822 machine->static_chain_stack_bytes.
8824 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8827 * doc/extend.texi: Add four new prototypes for vec_ld.
8828 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8829 definitions for more logical presentation.
8830 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8831 entries for V1TI variants of __builtin_altivec_ld builtin.
8832 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8833 handling of V1TI variant of LVX icode pattern.
8834 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8835 (rs6000_gimple_fold_builtin): Likewise.
8836 (altivec_init_builtins): Add code to define
8837 __builtin_altivec_lvx_v1ti function.
8839 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8842 * function.c (match_asm_constraints_1): Don't do the optimization
8843 if input isn't a REG, SUBREG, MEM or constant.
8845 2018-03-22 Tom de Vries <tom@codesourcery.com>
8847 PR tree-optimization/84956
8848 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8849 bb_has_abnormal_pred.
8851 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8854 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8855 DECL_INITIAL (decl) to decl at the end.
8856 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8859 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8861 * doc/extend.texi (__builtin_tgmath): Document when complex
8862 integer types are treated as _Complex _Float64.
8864 2018-03-21 Tom de Vries <tom@codesourcery.com>
8866 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8868 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8870 PR tree-optimization/84960
8871 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8872 if it is ENTRY block, move them into single succ of ENTRY in that case.
8874 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8876 PR tree-optimization/84811
8877 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8878 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8879 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8880 (known_subrange_p): Update accordingly. Cast each value involved
8881 in the size comparison, rather than casting the result of the
8884 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8886 PR tree-optimization/84982
8887 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8888 by flipping the least significant bit rather than all bits from
8889 bitpos to bitpos + bitsize - 1.
8891 2018-03-21 Nathan Sidwell <nathan@acm.org>
8893 * doc/extend.texi (Deprecated Features): Remove mention of
8894 long-deleted deprecations.
8896 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8899 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8900 * configure: Regenerate.
8902 2018-03-21 Tom de Vries <tom@codesourcery.com>
8904 PR tree-optimization/83126
8905 * tree-parloops.c (num_phis): New function.
8906 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8908 2018-03-21 Nathan Sidwell <nathan@acm.org>
8910 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8911 mention anon-struct/union members and trailing attributes.
8913 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8915 PR tree-optimization/84969
8916 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8917 builtin memset partitions if they set different rhs values.
8919 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8921 PR rtl-optimization/84989
8922 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8923 VEC_DUPLICATE with scalar result mode.
8925 2018-03-21 Martin Liska <mliska@suse.cz>
8928 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8929 not intended return statement.
8931 2018-03-21 Martin Liska <mliska@suse.cz>
8934 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8935 (chkp_find_bound_slots_1): Limit number of iterations.
8937 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8940 * Minor grammar fixes for x86 options.
8942 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8945 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8946 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8948 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8951 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8952 (altivec_lvx_<mode>_1op): Likewise.
8953 (altivec_stvx_<mode>_2op): Likewise.
8954 (altivec_stvx_<mode>_1op): Likewise.
8955 (altivec_lvx_<VM2:mode>): New define_expand.
8956 (altivec_stvx_<VM2:mode>): Likewise.
8957 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8958 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8959 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8960 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8961 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8962 (rs6000_gen_lvx): Likewise.
8963 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8964 (altivec_expand_stv_builtin): Likewise.
8965 (altivec_expand_builtin): Likewise.
8966 * config/rs6000/vector.md: Likewise.
8968 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8971 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8974 2018-03-20 Richard Biener <rguenther@suse.de>
8977 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8978 sign-conversions as zero, fall back to standard scalar_stmt
8981 2018-03-20 Martin Liska <mliska@suse.cz>
8984 * predict.c (rebuild_frequencies): Handle case when we have
8985 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8987 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8990 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8991 flag_section_anchors.
8992 * varasm.c (use_blocks_for_decl_p): Remove hack for
8993 dw2_force_const_mem.
8996 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8998 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8999 be created, use lowpart_subreg of operands[0] rather than operands[0]
9001 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9002 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9003 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9004 and n constraint instead of aarch64_shift_imm_di and Usd.
9005 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9006 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9008 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9011 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9012 to favor GPR over NEON registers.
9013 (<shift>di3_neon): Likewise.
9015 2018-03-20 Tom de Vries <tom@codesourcery.com>
9018 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9019 (nvptx_process_pars): Emit bar.sync asap and alap.
9021 2018-03-20 Tom de Vries <tom@codesourcery.com>
9024 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9025 seen_label if seen_label is already set.
9027 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9030 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9031 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9032 Use 1U instead of 1. Formatting fixes.
9035 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9036 instead of TREE_TYPE (s1) for the return value.
9038 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9040 PR tree-optimization/84946
9041 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9042 bitsize + bitsize in poly_uint64 rather than poly_int64.
9045 * dwarf2asm.c: Include fold-const.c.
9046 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9047 of decl rather than decl itself.
9049 PR rtl-optimization/84643
9050 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9052 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9055 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9056 calling assemble_variable.
9058 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9061 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9062 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9064 2018-03-19 Jim Wilson <jimw@sifive.com>
9067 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9068 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9069 (riscv_first_stack_step): Likewise.
9070 (riscv_option_override): Use STACK_BOUNDARY instead of
9072 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9074 (BIGGEST_ALIGNMENT): Set to 128.
9075 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9076 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9079 2018-03-19 Richard Biener <rguenther@suse.de>
9081 PR tree-optimization/84933
9082 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9083 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9085 2018-03-19 Richard Biener <rguenther@suse.de>
9087 PR tree-optimization/84859
9088 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9089 (cond_if_else_store_replacement): Perform sinking operation on
9090 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9091 Generalize what a BB with a single eligible store is.
9093 2018-03-19 Richard Biener <rguenther@suse.de>
9095 PR tree-optimization/84929
9096 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9097 chrec_is_positive against non-chrec arg.
9099 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9102 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9104 2018-03-18 Martin Liska <mliska@suse.cz>
9106 PR rtl-optimization/84635
9107 * regrename.c (build_def_use): Use matches_mode only when
9110 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9112 PR tree-optimization/84913
9113 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9114 vectorize chains of COND_EXPRs.
9116 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9118 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9120 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9122 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9124 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9126 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9128 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9129 Kito Cheng <kito.cheng@gmail.com>
9131 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9132 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9133 (nds32_adjust_reg_alloc_order): New function.
9134 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9136 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9138 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9139 nds32_print_operand, nds32_print_operand_address): Use
9140 HOST_WIDE_INT_PRINT_DEC instead.
9142 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9144 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9146 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9149 * config/i386/i386.c (initial_ix86_tune_features,
9150 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9152 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9153 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9154 rather than 1u << ix86_tune. Formatting fix.
9155 (ix86_option_override_internal): Change ix86_arch_mask from
9156 unsigned int to unsigned HOST_WIDE_INT, initialize to
9157 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9158 (ix86_function_specific_restore): Likewise.
9160 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9163 * postreload.c (reload_combine_recognize_pattern): Perform
9164 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9165 truncate_int_for_mode the result for the destination's mode.
9168 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9169 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9171 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9174 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9177 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9178 regno_allocno_class_array and sorted_pseudos.
9179 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9180 insns where regno is used.
9182 2018-03-16 Martin Liska <mliska@suse.cz>
9185 * multiple_target.c (create_dispatcher_calls): Redirect
9186 reference in the symbol table.
9188 2018-03-16 Martin Liska <mliska@suse.cz>
9191 * multiple_target.c (create_dispatcher_calls): Redirect also
9194 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9198 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9199 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9200 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9202 2018-03-16 Julia Koval <julia.koval@intel.com>
9204 * doc/invoke.texi (Skylake Server): Add CLWB.
9205 Cannonlake): Remove CLWB.
9207 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9209 PR tree-optimization/84841
9210 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9212 (FLOAT_ONE_CONST_TYPE): Define.
9213 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9214 (sort_by_operand_rank): Put entries with higher constant_type last
9215 rather than first to match comments.
9217 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9219 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9222 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9225 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9226 check and use address of target always.
9228 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9231 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9232 (indirect_thunk_bnd_needed): Likewise.
9233 (indirect_thunks_used): Likewise.
9234 (indirect_thunks_bnd_used): Likewise.
9235 (indirect_return_needed): New.
9236 (indirect_return_bnd_needed): Likewise.
9237 (output_indirect_thunk_function): Add a bool argument for
9239 (output_indirect_thunk_function): Don't generate alias for
9240 function return thunk.
9241 (ix86_code_end): Call output_indirect_thunk_function to generate
9242 function return thunks.
9243 (ix86_output_function_return): Set indirect_return_bnd_needed
9244 and indirect_return_needed instead of indirect_thunk_bnd_needed
9245 and indirect_thunk_needed.
9247 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9249 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9250 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9251 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9253 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9254 Paul Hua <paul.hua.gm@gmail.com>
9257 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9259 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9261 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9262 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9265 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9268 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9269 instead of GET_MODE_SIZE when comparing Units.
9271 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9274 * varasm.c (hash_section): Return an unchangeble hash value
9275 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9276 Return !aarch64_can_use_per_function_literal_pools_p ().
9278 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9281 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9282 rather than address of cmode as last argument to prepare_cmp_insn.
9284 2018-03-15 Julia Koval <julia.koval@intel.com>
9286 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9287 F_AVX512VNNI, F_AVX512BITALG): New.
9289 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9292 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9293 insn for floating-point loads and stores.
9295 2018-03-14 Carl Love <cel@us.ibm.com>
9297 * config/rs6000/rs6000-c.c: Add macro definitions for
9298 ALTIVEC_BUILTIN_VEC_PERMXOR.
9299 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9300 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9301 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9302 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9304 * config/doc/extend.texi: Add prototypes for vec_permxor.
9306 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9309 * diagnostic-show-locus.c (class layout_point): Convert m_line
9310 from int to linenum_type.
9311 (line_span::comparator): Use linenum "compare" function when
9312 comparing line numbers.
9313 (test_line_span): New function.
9314 (layout_range::contains_point): Convert param "row" from int to
9316 (layout_range::intersects_line_p): Likewise.
9317 (layout::will_show_line_p): Likewise.
9318 (layout::print_source_line): Likewise.
9319 (layout::should_print_annotation_line_p): Likewise.
9320 (layout::print_annotation_line): Likewise.
9321 (layout::print_leading_fixits): Likewise.
9322 (layout::annotation_line_showed_range_p): Likewise.
9323 (struct line_corrections): Likewise for field m_row.
9324 (line_corrections::line_corrections): Likewise for param "row".
9325 (layout::print_trailing_fixits): Likewise.
9326 (layout::get_state_at_point): Likewise.
9327 (layout::get_x_bound_for_row): Likewise.
9328 (layout::print_line): Likewise.
9329 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9330 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9331 * input.c (selftest::test_linenum_comparisons): New function.
9332 (selftest::input_c_tests): Call it.
9333 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9334 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9335 * selftest.h (ASSERT_GT): New macro.
9336 (ASSERT_GT_AT): New macro.
9337 (ASSERT_LT): New macro.
9338 (ASSERT_LT_AT): New macro.
9340 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9342 PR rtl-optimization/84780
9343 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9345 2018-03-14 Martin Liska <mliska@suse.cz>
9347 * tree.c (record_node_allocation_statistics): Use
9348 get_stats_node_kind.
9349 (get_stats_node_kind): New function extracted from
9350 record_node_allocation_statistics.
9351 (free_node): Use get_stats_node_kind.
9353 2018-03-14 Richard Biener <rguenther@suse.de>
9355 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9356 that the value-set of ANTIC_IN doesn't grow.
9359 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9361 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9362 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9364 2018-03-14 Julia Koval <julia.koval@intel.com>
9366 * config.gcc (icelake-client, icelake-server): New.
9368 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9369 (initial_ix86_arch_features): Ditto.
9370 (PTA_SKYLAKE): Add SGX.
9371 (PTA_ICELAKE): Remove.
9372 (PTA_ICELAKE_CLIENT): New.
9373 (PTA_ICELAKE_SERVER): New.
9374 (ix86_option_override_internal): Split up icelake on icelake client and
9376 (get_builtin_code_for_version): Ditto.
9377 (fold_builtin_cpu): Ditto.
9378 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9379 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9380 * config/i386/i386.h (processor_type): Ditto.
9381 * doc/invoke.texi: Ditto.
9383 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9386 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9387 INTEGER_CST offset, add it together with bitpos / 8 and
9388 sign extend based on POINTER_SIZE.
9392 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9395 * config/i386/constraints.md (Yc): New register constraint.
9396 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9397 Use Yc constraint for alternative 2 of operand 0. Remove
9398 preferred_for_speed attribute.
9400 2018-03-14 Richard Biener <rguenther@suse.de>
9402 PR tree-optimization/84830
9403 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9404 with the old one to avoid oscillations.
9406 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9409 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9411 (assign_by_spills): Return a flag of reload assignment failure.
9412 Do not process the reload assignment failures. Do not spill other
9413 reload pseudos if they has the same reg class. Update n if
9415 (lra_assign): Add a return arg. Set up from the result of
9416 assign_by_spills call.
9417 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9418 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9419 usage_insns if it is not NULL.
9420 (spill_hard_reg_in_range): New function.
9421 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9422 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9423 function prototypes.
9424 (lra_assign): Change prototype.
9425 * lra.c (lra): Add code to deal with fails by splitting hard reg
9428 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9430 * config/riscv/riscv.opt (mrelax): New option.
9431 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9432 "norelax" when riscv_mrelax is disabled.
9433 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9435 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9438 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9439 reassociation for int modes.
9441 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9443 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9444 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9446 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9447 * config/aarch64/aarch64-sve.md
9448 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9449 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9450 (*extend<mode><Vwide>2): Rename to...
9451 (aarch64_sve_extend<mode><Vwide>2): ...this.
9452 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9453 renaming the old pattern to...
9454 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9456 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9457 define_expand, renaming the old pattern to...
9458 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9459 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9460 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9461 account when deciding which SVE instruction the optab should use.
9462 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9464 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9466 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9467 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9468 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9469 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9470 (tlsdesc_small_<mode>): Turn a define_expand and use
9471 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9472 (tlsdesc_small_advsimd_<mode>): ...this.
9473 (tlsdesc_small_sve_<mode>): New pattern.
9475 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9477 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9478 (UNSPEC_UMUL_HIGHPART): New constants.
9479 (MUL_HIGHPART): New int iteraor.
9480 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9481 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9483 (*<su>mul<mode>3_highpart): New define_insn.
9485 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9488 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9491 2018-03-13 Martin Liska <mliska@suse.cz>
9494 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9496 (sem_item_optimizer::~sem_item_optimizer): Release it.
9497 (sem_item_optimizer::merge_classes): Register variable aliases.
9498 (sem_item_optimizer::fixup_pt_set): New function.
9499 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9500 * ipa-icf.h: Declare new variables and functions.
9502 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9505 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9506 integer_pow2p@2 and test integer_pow2p in condition.
9507 (A < 0 ? C : 0): Similarly for @1.
9510 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9511 characters starting at p contain '\0' character, don't look beyond
9515 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9516 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9519 * reg-stack.c (change_stack): Change update_end var from int to
9520 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9521 also call set_block_for_insn on the newly added insns and rescan.
9524 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9525 on the last operand.
9528 * tree.c (stabilize_reference_1): Return save_expr (e) for
9529 STATEMENT_LIST even if it doesn't have side-effects.
9531 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9533 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9535 2018-03-12 Renlin Li <renlin.li@arm.com>
9537 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9538 aarch64_output_scalar_simd_mov_immediate.
9540 2018-03-12 Martin Sebor <msebor@redhat.com>
9542 PR tree-optimization/83456
9543 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9544 for perfectly overlapping calls to memcpy.
9545 (gimple_fold_builtin_memory_chk): Same.
9546 (gimple_fold_builtin_strcpy): Handle no-warning.
9547 (gimple_fold_builtin_stxcpy_chk): Same.
9548 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9550 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9552 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9553 parameter. Use it for SFmode.
9554 (rs6000_function_arg_advance_1): Adjust.
9555 (rs6000_function_arg): Adjust.
9556 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9558 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9560 PR rtl-optimization/84169
9561 PR rtl-optimization/84780
9562 * combine.c (can_combine_p): Check for a 2-insn combination whether
9563 the destination register is used between the two insns, too.
9565 2018-03-12 Richard Biener <rguenther@suse.de>
9567 PR tree-optimization/84803
9568 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9569 for refs DR analysis didn't process.
9571 2018-03-12 Richard Biener <rguenther@suse.de>
9573 PR tree-optimization/84777
9574 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9575 force-vectorize loops ignore whether we are optimizing for size.
9577 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9579 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9580 (TARGET_MD_ASM_ADJUST): Define.
9582 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9583 Kito Cheng <kito.cheng@gmail.com>
9584 Chung-Ju Wu <jasonwucj@gmail.com>
9586 * config/nds32/nds32.c (nds32_compute_stack_frame,
9587 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9588 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9589 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9590 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9591 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9592 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9593 * config/nds32/nds32.md (prologue, epilogue): Use macro
9594 NDS32_V3PUSH_AVAILABLE_P to do checking.
9596 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9599 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9600 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9601 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9602 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9603 addition of most attributes on !orig_type_die or the attribute not
9604 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9606 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9607 Chung-Ju Wu <jasonwucj@gmail.com>
9609 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9611 * config/nds32/nds32.opt (mvh): New option.
9613 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9614 Chung-Ju Wu <jasonwucj@gmail.com>
9616 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9618 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9619 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9622 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9623 Chung-Ju Wu <jasonwucj@gmail.com>
9625 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9627 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9628 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9630 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9631 Kito Cheng <kito.cheng@gmail.com>
9632 Chung-Ju Wu <jasonwucj@gmail.com>
9634 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9635 UNSPEC_FFMISM and UNSPEC_FLMISM.
9636 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9637 for ffb, ffmism and flmism.
9638 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9639 (unspec_ffmism): Ditto.
9640 (unspec_flmism): Ditto.
9641 (nds32_expand_builtin_impl): Check if string extension is available.
9642 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9643 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9645 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9648 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9651 * lra-assigns.c (assign_by_spills): Return a flag of reload
9652 assignment failure. Do not process the reload assignment
9653 failures. Do not spill other reload pseudos if they has the same
9655 (lra_assign): Add a return arg. Set up from the result of
9656 assign_by_spills call.
9657 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9658 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9659 usage_insns if it is not NULL.
9660 (spill_hard_reg_in_range): New function.
9661 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9662 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9663 function prototypes.
9664 (lra_assign): Change prototype.
9665 * lra.c (lra): Add code to deal with fails by splitting hard reg
9668 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9671 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9673 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9676 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9677 (dw_val_node): Add val_symbolic_view.
9678 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9679 (symview_upper_bound): New.
9680 (new_line_info_table): Initialize symviews_since_reset.
9681 (dwarf2out_source_line): Count symviews_since_reset and set
9682 symview_upper_bound.
9683 (dw_val_equal_p): Handle symview.
9684 (add_AT_symview): New.
9685 (print_dw_val): Handle symview.
9686 (attr_checksum, attr_checksum_ordered): Likewise.
9687 (same_dw_val_p, size_of_die): Likewise.
9688 (value_format, output_die): Likewise.
9689 (add_high_low_attributes): Use add_AT_symview for entry_view.
9690 (dwarf2out_finish): Reset symview_upper_bound, clear
9693 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9696 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9697 Add strict argument and use it.
9698 (rs6000_split_multireg_move): Update for new strict argument.
9699 (mem_operand_gpr): Disallow all non-offsettable addresses.
9700 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9702 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9705 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9706 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9707 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9710 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9711 decl, use remap_type if we want to use the type.
9713 2018-03-09 Martin Sebor <msebor@redhat.com>
9715 PR tree-optimization/84526
9716 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9718 (builtin_access::generic_overlap): Be prepared to handle non-array
9721 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9723 PR rtl-optimization/84682
9724 * lra-constraints.c (process_address_1): Check is_address flag
9725 for address constraints.
9726 (process_alt_operands): Likewise.
9727 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9728 preprocess_constraints.
9729 * recog.h (preprocess_constraints): Add oploc parameter.
9731 * recog.c (preprocess_constraints): Test address_operand for
9732 CT_ADDRESS constraints.
9734 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9737 * lra-assigns.c (assign_by_spills): Return a flag of reload
9738 assignment failure. Do not process the reload assignment
9739 failures. Do not spill other reload pseudos if they has the same
9741 (lra_assign): Add a return arg. Set up from the result of
9742 assign_by_spills call.
9743 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9744 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9745 usage_insns if it is not NULL.
9746 (spill_hard_reg_in_range): New function.
9747 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9748 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9749 function prototypes.
9750 (lra_assign): Change prototype.
9751 * lra.c (lra): Add code to deal with fails by splitting hard reg
9754 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9757 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9758 Accept complain bool parameter. Only emit errors if it is true.
9759 (arm_parse_cpu_option_name): Likewise.
9760 (arm_target_thumb_only): Adjust callers of the above.
9761 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9762 prototype to take a default true bool parameter.
9763 (arm_parse_arch_option_name): Likewise.
9765 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9766 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9770 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9771 * configure: Regenerate.
9772 * configure.ac ("linker --version-script option"): New.
9773 ("linker soname option"): New.
9775 2018-03-09 Richard Biener <rguenther@suse.de>
9777 PR tree-optimization/84775
9778 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9779 immediate uses of predicate stmts and mark them modified.
9782 PR tree-optimization/84178
9783 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9785 (version_loop_for_if_conversion): Delay update_ssa call.
9786 (tree_if_conversion): Delay update_ssa until after predicate
9789 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9792 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9793 when the function accesses prior frames.
9795 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9798 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9799 gen_llsym, otherwise call maybe_gen_llsym.
9802 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9803 has ',' character inside of it.
9805 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9808 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9809 as clobbering CC_REGNUM.
9811 2018-03-08 Richard Biener <rguenther@suse.de>
9814 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9815 (follow_copies_to_constant): Do not follow SSA names registered
9818 2018-03-08 Richard Biener <rguenther@suse.de>
9820 PR tree-optimization/84178
9821 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9823 (version_loop_for_if_conversion): Delay update_ssa call.
9824 (tree_if_conversion): Delay update_ssa until after predicate
9827 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9829 PR tree-optimization/84178
9830 * tree-if-conv.c (release_bb_predicate): Remove the
9831 the assertion that the stmts have NULL use_ops.
9832 Discard the statements, asserting that they haven't
9833 yet been added to a BB.
9835 2018-03-08 Richard Biener <rguenther@suse.de>
9837 PR tree-optimization/84746
9838 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9839 (phi_translate): Pass in destination ANTIC_OUT set.
9840 (phi_translate_1): Likewise. For a simplified result lookup
9841 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9842 (phi_translate_set): Adjust.
9843 (do_pre_regular_insertion): Likewise.
9844 (do_pre_partial_partial_insertion): Likewise.
9846 2018-03-08 Martin Liska <mliska@suse.cz>
9848 PR gcov-profile/84735
9849 * doc/gcov.texi: Document usage of profile files.
9850 * gcov-io.h: Document changes in the format.
9852 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9856 * dwarf2out.c (struct dw_line_info_table): Update comments for
9858 (FORCE_RESET_NEXT_VIEW): New.
9859 (FORCE_RESETTING_VIEW_P): New.
9860 (RESETTING_VIEW_P): Check for -1 too.
9861 (ZERO_VIEW_P): Likewise.
9862 (new_line_info_table): Force-reset next view.
9863 (dwarf2out_begin_function): Likewise.
9864 (dwarf2out_source_line): Simplify zero_view_p initialization.
9865 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9866 view directly. Omit view when omitting .loc at line 0.
9868 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9870 PR tree-optimization/84740
9871 * tree-switch-conversion.c (process_switch): Call build_constructors
9872 only if info.phi_count is non-zero.
9874 PR tree-optimization/84739
9875 * tree-tailcall.c (find_tail_calls): Check call arguments against
9876 DECL_ARGUMENTS (current_function_decl) rather than
9877 DECL_ARGUMENTS (func) when checking for tail recursion.
9879 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9881 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9882 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9883 Volker Reichelt's entry and add entries for people that perform
9884 GCC fuzzy testing and report numerous bugs.
9886 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9889 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9890 readonly data in sdata, if that is disabled.
9891 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9892 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9893 -mreadonly-in-sdata option.
9895 2018-03-07 Martin Sebor <msebor@redhat.com>
9897 PR tree-optimization/84468
9898 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9899 basic block when looking for nul assignment.
9901 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9904 * except.h (output_function_exception_table): Adjust prototype.
9905 * except.c (output_function_exception_table): Remove FNNAME parameter
9906 and add SECTION parameter. Ouput one part of the table at a time.
9907 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9908 the first part of the exception table and emit unwind directives.
9909 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9910 (i386_pe_seh_cold_init): Likewise.
9911 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9912 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9913 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9914 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9915 * config/i386/winnt.c: Include except.h.
9916 (struct seh_frame_state): Add reg_offset, after_prologue and
9917 in_cold_section fields.
9918 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9919 (i386_pe_seh_cold_init): New function.
9920 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9921 to seh->in_cold_section.
9922 (seh_emit_push): Record the offset of the push.
9923 (seh_emit_save): Record the offet of the save.
9924 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9925 Test seh->after_prologue to disregard the epilogue.
9926 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9927 (i386_pe_end_cold_function): New function.
9929 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9932 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9933 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9936 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9938 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9939 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9942 * multiple_target.c: Include tree-inline.h and intl.h.
9943 (expand_target_clones): Diagnose and fail if node->definition and
9944 !tree_versionable_function_p (node->decl).
9946 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9948 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9950 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9951 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9952 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9954 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9957 * combine.c (try_combine): Use reg_or_subregno instead of handling
9958 just paradoxical SUBREGs and REGs.
9960 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9962 * config/arc/arc.c (arc_finalize_pic): Remove function.
9963 (arc_must_save_register): We use single base PIC register, remove
9964 checks to save/restore the PIC register.
9965 (arc_expand_prologue): Likewise.
9966 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9968 (arc_verify_short): Likewise.
9969 (arc_attr_type): Likewise.
9970 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9971 (walk_stores): Likewise.
9972 (arc_address_cost): Make it static.
9973 (arc_verify_short): Likewise.
9974 (branch_dest): Likewise.
9975 (arc_attr_type): Likewise.
9976 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9977 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9978 (arc_final_prescan_insn): Remove inserting the nops due to
9979 hardware hazards. It is done in reorg step.
9980 (insn_length_variant_t): Remove.
9981 (insn_length_parameters_t): Likewise.
9982 (arc_insn_length_parameters): Likewise.
9983 (arc_get_insn_variants): Likewise.
9984 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9986 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9989 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9992 PR tree-optimization/84687
9993 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9995 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9997 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9999 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10000 Rename to ppc_speculation_barrier.
10001 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10002 __builtin_ppc_speculation_barrier.
10004 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10007 * combine.c (combine_simplify_rtx): Don't try to simplify if
10008 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10011 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10013 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10014 to 32 bytes when compiling for POWER9.
10016 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10019 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10020 regparm >= 3 with no arg reg available also for calls with
10021 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10024 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10026 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10028 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10031 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10033 2018-03-05 Richard Biener <rguenther@suse.de>
10035 PR tree-optimization/84486
10036 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10037 When inserting a __builtin_assume_aligned call set the LHS
10038 SSA name alignment info accordingly.
10040 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10042 PR tree-optimization/84114
10043 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10044 Avoid reassociation of FLOAT_MODE addition.
10046 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10048 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10049 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10050 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10051 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10052 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10053 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10054 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10056 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10057 __builtin_ia32_wbinvd): New builtins.
10058 (SPECIAL_ARGS2): New.
10059 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10060 (SPECIAL_ARGS2): New.
10061 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10062 (ix86_valid_target_attribute_inner_p): Ditto.
10063 (ix86_init_mmx_sse_builtins): Add special_args2.
10064 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10065 TARGET_WBNOINVD_P): New.
10066 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10067 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10068 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10069 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10070 * config/i386/pconfigintrin.h: New file.
10071 * config/i386/wbnoinvdintrin.h: Ditto.
10072 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10074 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10076 2018-03-05 Richard Biener <rguenther@suse.de>
10078 PR tree-optimization/84670
10079 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10081 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10082 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10083 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10084 doesn't grow if all successors (recursively) were visited at least
10087 2018-03-05 Richard Biener <rguenther@suse.de>
10089 PR tree-optimization/84650
10090 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10091 if executed in the loop pipeline.
10093 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10095 * doc/configfiles.texi (Configuration Files): Move info about
10096 conditionalizing $target-protos.h to...
10097 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10098 differs from $target-protos.h.
10100 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10101 Chung-Ju Wu <jasonwucj@gmail.com>
10103 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10104 * config/nds32/nds32-multiple.md (setmemsi): Define.
10105 * config/nds32/nds32-memory-manipulation.c
10106 (nds32_gen_dup_4_byte_to_word_value): New.
10107 (emit_setmem_word_loop): New.
10108 (emit_setmem_byte_loop): New.
10109 (nds32_expand_setmem_loop): New.
10110 (nds32_expand_setmem_loop_v3m): New.
10111 (nds32_expand_setmem_unroll): New.
10112 (nds32_expand_setmem): New.
10114 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10115 Chung-Ju Wu <jasonwucj@gmail.com>
10117 * config/nds32/nds32-memory-manipulation.c
10118 (nds32_emit_load_store): New.
10119 (nds32_emit_post_inc_load_store): New.
10120 (nds32_emit_mem_move): New.
10121 (nds32_emit_mem_move_block): New.
10122 (nds32_expand_movmemsi_loop_unknown_size): New.
10123 (nds32_expand_movmemsi_loop_known_size): New.
10124 (nds32_expand_movmemsi_loop): New.
10125 (nds32_expand_movmemsi_unroll): New.
10126 (nds32_expand_movmemqi): Rename ...
10127 (nds32_expand_movmemsi): ... to this.
10128 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10129 (movmemsi): ... to this.
10130 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10131 (nds32_expand_movmemsi): ... to this.
10133 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10134 Monk Chiang <sh.chiang04@gmail.com>
10135 Chung-Ju Wu <jasonwucj@gmail.com>
10137 * config/nds32/nds32-protos.h
10138 (nds32_expand_load_multiple): New arguments.
10139 (nds32_expand_store_multiple): Ditto.
10140 (nds32_valid_multiple_load_store): Rename ...
10141 (nds32_valid_multiple_load_store_p): ... to this.
10142 * config/nds32/nds32-memory-manipulation.c
10143 (nds32_expand_load_multiple): Refine implementation.
10144 (nds32_expand_store_multiple): Ditto.
10145 * config/nds32/nds32-multiple.md
10146 (load_multiple): Update nds32_expand_load_multiple interface.
10147 (store_multiple): Update nds32_expand_store_multiple interface.
10148 * config/nds32/nds32-predicates.c
10149 (nds32_valid_multiple_load_store): Rename ...
10150 (nds32_valid_multiple_load_store_p): ... to this and refine
10152 * config/nds32/predicates.md
10153 (nds32_load_multiple_and_update_address_operation): New predicate.
10154 (nds32_store_multiple_and_update_address_operation): New predicate.
10156 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10157 Chung-Ju Wu <jasonwucj@gmail.com>
10159 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10160 (combo): New attribute.
10161 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10163 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10165 * config/nds32/nds32.opt: Change -mcmodel= default value.
10167 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10168 Monk Chiang <sh.chiang04@gmail.com>
10169 Chung-Ju Wu <jasonwucj@gmail.com>
10171 * config/nds32/constants.md (unspec_element): New enum.
10172 * config/nds32/constraints.md (Umw): New constraint.
10173 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10174 * config/nds32/nds32-intrinsic.md: Likewise.
10175 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10176 (nds32_valid_smw_lwm_base_p): New.
10177 (nds32_output_smw_single_word): New.
10178 (nds32_output_lmw_single_word): New.
10179 (nds32_expand_unaligned_load): New.
10180 (nds32_expand_unaligned_store): New.
10181 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10182 (nds32_output_smw_single_word): Declare.
10183 (nds32_output_lmw_single_word): Declare.
10184 (nds32_expand_unaligned_load): Declare.
10185 (nds32_expand_unaligned_store): Declare.
10186 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10187 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10188 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10189 NDS32_BUILTIN_UASTORE_DW.
10190 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10193 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10194 Kito Cheng <kito.cheng@gmail.com>
10195 Chung-Ju Wu <jasonwucj@gmail.com>
10197 * config/nds32/nds32-intrinsic.c
10198 (nds32_expand_builtin_null_ftype_reg): Delete.
10199 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10200 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10201 (nds32_read_argument): New.
10202 (nds32_legitimize_target): Ditto.
10203 (nds32_legitimize_argument): Ditto.
10204 (nds32_check_constant_argument): Ditto.
10205 (nds32_expand_unop_builtin): Ditto.
10206 (nds32_expand_unopimm_builtin): Ditto.
10207 (nds32_expand_binop_builtin): Ditto.
10208 (nds32_builtin_decl_impl): Ditto.
10209 (builtin_description): Ditto.
10210 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10211 (nds32_init_builtins_impl): Ditto.
10212 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10213 (nds32_builtin_decl): New.
10214 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10215 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10217 2018-03-02 Jeff Law <law@redhat.com>
10219 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10220 (redundant_insn, fill_simple_delay_slots): Likewise.
10221 (fill_slots_from_thread): Likewise.
10222 * resource.c (mark_referenced_resources): Likewise.
10223 (mark_set_resources, find_dead_or_set_registers): Likewise.
10225 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10227 * substring-locations.h (format_warning_va): Formatting fix for
10228 ATTRIBUTE_GCC_DIAG.
10229 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10231 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10232 * substring-locations.c: Include intl.h.
10233 (format_warning_va): Turned into small wrapper around
10234 format_warning_n_va, renamed to ...
10235 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10236 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10238 (format_warning_at_substring_n): New function.
10239 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10240 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10241 format_warning_at_substring with just a shorter name instead of
10242 const function pointer.
10243 (fmtwarn_n): New function.
10244 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10245 appropriate, get rid of all the fmtstr temporaries, move conditionals
10246 with G_() wrapped string literals directly into fmtwarn arguments,
10247 cast dir.len to (int), formatting fixes.
10249 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10251 * doc/invoke.texi: Remove "Cilk Plus" references.
10253 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10254 Richard Biener <rguenther@suse.de>
10257 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10258 for error or warning attributes if CALL_FROM_THUNK_P is set.
10261 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10264 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10265 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10268 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10269 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10271 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10272 instead of -1U in last predictors element's probability member.
10274 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10277 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10278 arguments if they are comparable.
10280 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10282 PR tree-optimization/84634
10283 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10284 masks and masked_loop_p with a single loop_masks, making sure it's
10285 null for bb vectorization.
10287 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10289 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10290 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10291 loop->force_vectorize to check whether there is no alias.
10293 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10296 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10298 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10300 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10301 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10302 instead of a loop around prev_real_insn.
10303 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10306 PR inline-asm/84625
10307 * config/i386/i386.c (ix86_print_operand): Use conditional
10308 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10311 2018-03-02 Richard Biener <rguenther@suse.de>
10313 PR tree-optimization/84427
10314 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10315 (bitmap_set_subtract_values): Rewrite to handle multiple
10318 (prune_clobbered_mems): Likewise.
10319 (phi_translate): Take edge instead of pred/phiblock.
10320 (phi_translate_1): Likewise.
10321 (phi_translate_set): Likewise. Insert all translated
10322 exprs for a value into the set, keeping possibly multiple
10323 expressions per value.
10324 (compute_antic_aux): Adjust for phi_translate changes.
10325 When intersecting union the expressions and prune those
10326 not in the final value set, keeping possibly multiple
10327 expressions per value. Do not use value-insertion
10328 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10329 all expressions. Add verification that the value-sets
10330 only shrink during iteration.
10331 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10332 (do_pre_regular_insertion): Likewise.
10333 (do_pre_partial_partial_insertion): Likewise.
10335 2018-03-02 Richard Biener <rguenther@suse.de>
10338 * config/darwin.c (saved_debug_info_level): New static global.
10339 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10340 (darwin_asm_lto_end): Restore debug info generation settings.
10342 2018-03-01 Martin Liska <mliska@suse.cz>
10345 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10346 volatile arguments.
10348 2018-03-01 Richard Biener <rguenther@suse.de>
10351 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10352 type annotation with LTO when debug was disabled at compile-time.
10354 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10356 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10358 (mips_final_postscan_insn): Likewise.
10360 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10362 PR rtl-optimization/84528
10363 * alias.c (init_alias_target): Add commentary.
10364 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10365 a unique base value if the frame pointer is not eliminated
10366 to the stack pointer.
10368 2018-03-01 Tom de Vries <tom@codesourcery.com>
10370 PR rtl-optimization/83327
10371 * lra-int.h (hard_regs_spilled_into): Declare.
10372 * lra.c (hard_regs_spilled_into): Define.
10373 (init_reg_info): Init hard_regs_spilled_into.
10374 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10375 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10376 (process_bb_lives): Handle hard_regs_spilled_into.
10377 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10378 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10380 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10382 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10383 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10384 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10385 * config/rs6000/aix72.h: New file.
10387 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10389 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10390 instead of warning_at with conditional singular and plural messages
10394 * stor-layout.c (update_alignment_for_field): For
10395 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10396 && !DECL_PACKED (field), do the alignment update, just use
10397 only desired_align instead of MAX (type_align, desired_align)
10399 (place_field): Don't do known_align < desired_align handling
10400 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10401 is non-NULL, instead do it after rli->prev_field handling and
10402 only if not within a bitfield word. For DECL_PACKED (field)
10403 use type_align of BITS_PER_UNIT.
10405 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10407 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10408 superfluous parentheses and trailing spaces.
10410 2018-02-28 Richard Biener <rguenther@suse.de>
10412 PR tree-optimization/84584
10413 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10414 SCoPs with fake exit edge.
10416 2018-02-28 Martin Liska <mliska@suse.cz>
10419 * timevar.c (timer::print): Fix format to properly print 100%
10422 2018-02-28 Richard Biener <rguenther@suse.de>
10424 PR middle-end/84607
10425 * genmatch.c (capture_info::walk_match): Do not mark
10426 captured expressions without operands as expr_p given
10427 they act more like predicates and should be subject to
10428 "lost tail" side-effect preserving.
10430 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10432 PR rtl-optimization/81611
10433 * auto-inc-dec.c (attempt_change): Move dead note from
10434 mem_insn if it's the next use of regno
10435 (find_address): Take address use of reg holding
10436 non-incremented value. Add parm to limit search to the named
10438 (merge_in_block): Attempt to use a mem insn that is the next
10439 use of the original regno.
10441 2018-02-27 Martin Sebor <msebor@redhat.com>
10444 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10445 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10447 2018-02-27 Martin Sebor <msebor@redhat.com>
10449 PR translation/84207
10450 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10451 n argument to unsigned HOST_WIDE_INT.
10452 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10453 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10454 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10455 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10457 2018-02-27 Richard Biener <rguenther@suse.de>
10459 PR tree-optimization/84512
10460 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10461 Do not use the estimate returned from record_stmt_cost for
10462 the scalar iteration cost but sum properly using add_stmt_cost.
10464 2018-02-27 Richard Biener <rguenther@suse.de>
10466 PR tree-optimization/84466
10467 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10468 Adjust last change to less strictly validate use operands.
10470 2018-02-27 Martin Liska <mliska@suse.cz>
10472 PR gcov-profile/84548
10473 * gcov.c (process_file): Allow partial overlap and consider it
10474 also as group functions.
10475 (output_lines): Properly calculate range of lines for a group.
10477 2018-02-27 Martin Liska <mliska@suse.cz>
10479 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10480 'ggc' suffixes. Change first column width.
10481 (timer::print): Fix formatting of the column.
10483 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10485 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10486 preserve inline entry blocks for the sake of debug inline
10487 entry point markers alone.
10488 (remove_unused_locals): Suggest in comments a better place to
10489 force the preservation of inline entry blocks that are
10490 otherwise unused, but do not preserve them.
10492 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10494 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10496 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10499 * config/i386/constraints.md (Bs): Replace
10500 ix86_indirect_branch_register with
10501 TARGET_INDIRECT_BRANCH_REGISTER.
10503 * config/i386/i386.md (indirect_jump): Likewise.
10504 (tablejump): Likewise.
10505 (*sibcall_memory): Likewise.
10506 (*sibcall_value_memory): Likewise.
10507 Peepholes of indirect call and jump via memory: Likewise.
10508 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10509 (*sibcall_value_GOT_32): Likewise.
10510 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10511 (GOT_memory_operand): Likewise.
10512 (call_insn_operand): Likewise.
10513 (sibcall_insn_operand): Likewise.
10514 (GOT32_symbol_operand): Likewise.
10515 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10517 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10519 PR rtl-optimization/83496
10520 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10521 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10522 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10523 redundant insn, if any.
10524 (relax_delay_slots): Likewise.
10525 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10527 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10529 PR tree-optimization/83965
10530 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10531 that grouped statements are part of a reduction chain. Return
10532 true if the statement is not marked as a reduction itself but
10533 is part of a group.
10534 (vect_recog_dot_prod_pattern): Don't check whether the statement
10535 is part of a group here.
10536 (vect_recog_sad_pattern): Likewise.
10537 (vect_recog_widen_sum_pattern): Likewise.
10539 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10542 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10544 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10547 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10549 (ix86_output_indirect_function_return): New prototype.
10550 (ix86_split_simple_return_pop_internal): Likewise.
10551 * config/i386/i386.c (indirect_return_via_cx): New.
10552 (indirect_return_via_cx_bnd): Likewise.
10553 (indirect_thunk_name): Handle return va CX_REG.
10554 (output_indirect_thunk_function): Create alias for
10555 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10556 (ix86_output_indirect_jmp): Remove the bool argument.
10557 (ix86_output_indirect_function_return): New function.
10558 (ix86_split_simple_return_pop_internal): Likewise.
10559 * config/i386/i386.md (*indirect_jump): Don't pass false
10560 to ix86_output_indirect_jmp.
10561 (*tablejump_1): Likewise.
10562 (simple_return_pop_internal): Change it to define_insn_and_split.
10563 Call ix86_split_simple_return_pop_internal to split it for
10564 -mfunction-return=.
10565 (simple_return_indirect_internal): Call
10566 ix86_output_indirect_function_return instead of
10567 ix86_output_indirect_jmp.
10569 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10572 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10573 memset and value initialization afterwards.
10575 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10577 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10579 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10582 * common/config/aarch64/aarch64-common.c
10583 (aarch_option_optimization_table[]): Switch
10584 off fomit-frame-pointer
10586 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10587 Chung-Ju Wu <jasonwucj@gmail.com>
10589 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10591 (store_multiple): Ditto.
10593 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10595 * config.gcc: Add --with-cpu support for nds32 target.
10596 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10597 * config/nds32/nds32.opt: Add -mcpu= option.
10599 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10601 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10602 isel=yes): Warn for these deprecated options.
10604 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10606 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10607 ISA_2_5_MASKS_EMBEDDED.
10609 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10611 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10612 p->max as pointers rather than using iterative_hash_expr.
10614 2018-02-23 Carl Love <cel@us.ibm.com>
10616 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10617 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10619 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10620 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10621 P8V_BUILTIN_VEC_VUNSIGNED2.
10623 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10626 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10627 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10629 * lra-constraints.c (curr_insn_transform): Set up
10630 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10632 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10633 * lra-eliminations.c (spill_pseudos): Ditto.
10634 (process_insn_for_elimination): Ditto.
10635 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10636 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10637 LRA_NON_CLOBBERED_ALT.
10639 2018-02-22 Martin Sebor <msebor@redhat.com>
10641 PR tree-optimization/84480
10642 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10643 to maybe_diag_stxncpy_trunc. Call it.
10644 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10645 from gimple_fold_builtin_strcpy. Print inlining stack.
10646 (handle_builtin_stxncpy): Print inlining stack.
10647 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10649 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10652 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10653 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10654 and -fcheck-pointer-bounds are used together.
10655 (indirect_thunk_prefix): New enum.
10656 (indirect_thunk_need_prefix): New function.
10657 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10658 "_nt" instead of "_bnd" for NOTRACK prefix.
10659 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10660 (output_indirect_thunk_function): Likewise.
10662 (ix86_code_end): Update output_indirect_thunk_function calls.
10663 (ix86_output_indirect_branch_via_reg): Replace
10664 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10665 (ix86_output_indirect_branch_via_push): Likewise.
10666 (ix86_output_function_return): Likewise.
10667 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10668 incompatible with -fcf-protection=branch and
10669 -fcheck-pointer-bounds.
10671 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10674 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10675 Change gcc_assert call to output_operand_lossage.
10677 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10679 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10681 2018-02-22 DJ Delorie <dj@redhat.com>
10682 Sebastian Perta <sebastian.perta@renesas.com>
10683 Oleg Endo <olegendo@gcc.gnu.org>
10685 * config/rx/rx.c (rx_rtx_costs): New function.
10686 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10688 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10690 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10692 2018-02-22 Martin Liska <mliska@suse.cz>
10695 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10696 Add "native" as a possible value.
10698 2018-02-22 Martin Liska <mliska@suse.cz>
10701 * config/i386/i386.c (ix86_option_override_internal):
10702 Add "native" as a possible value for -march and -mtune.
10704 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10707 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10708 to all type variants.
10710 PR tree-optimization/84503
10711 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10712 width as info->bitpos + info->bitsize - start.
10713 (merged_store_group::merge_overlapping): Simplify width computation.
10714 (check_no_overlap): New function.
10715 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10716 start + width and last_order of the group, fail if check_no_overlap
10718 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10719 to group if check_no_overlap fails.
10721 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10723 * config/rs6000/altivec.md: Delete contraint arguments to
10724 define_expand, define_split, and define_peephole2, and in
10725 define_insn_and_split if always unused.
10726 * config/rs6000/darwin.md: Ditto.
10727 * config/rs6000/dfp.md: Ditto.
10728 * config/rs6000/rs6000.md: Ditto.
10729 * config/rs6000/sync.md: Ditto.
10730 * config/rs6000/vector.md: Ditto.
10731 * config/rs6000/vsx.md: Ditto.
10733 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10735 * config/rs6000/altivec.md: Write output control strings as braced
10736 blocks instead of double-quoted strings.
10737 * config/rs6000/darwin.md: Ditto.
10738 * config/rs6000/rs6000.md: Ditto.
10739 * config/rs6000/vector.md: Ditto.
10740 * config/rs6000/vsx.md: Ditto.
10742 2018-02-21 Jason Merrill <jason@redhat.com>
10744 PR c++/84314 - ICE with templates and fastcall attribute.
10745 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10747 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10749 * ipa-cp.c (determine_versionability): Fix comment typos.
10751 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10754 * ipa-cp.c (determine_versionability): Do not version functions caling
10757 2018-02-21 Martin Liska <mliska@suse.cz>
10760 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10761 Add "native" as a possible value.
10762 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10763 the macro when native cpu detection is available.
10765 2018-02-21 Martin Liska <mliska@suse.cz>
10768 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10769 Add "native" as a possible value.
10770 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10771 when native cpu detection is available.
10773 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10774 Martin Sebor <msebor@redhat.com>
10776 PR tree-optimization/84478
10777 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10779 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10780 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10781 support which is conservatively correct, for 2 only stay conservative
10782 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10783 argument to the 2 argument get_range_strlen, adjust 6 arg
10784 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10786 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10787 (gimple_fold_builtin_strlen): Pass true as last argument to
10790 2018-02-20 Martin Sebor <msebor@redhat.com>
10792 PR middle-end/84095
10793 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10794 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10795 (builtin_memref::builtin_memref): Factor out parts into
10796 set_base_and_offset and call it.
10798 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10800 PR middle-end/84406
10801 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10802 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10803 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10804 search at the associated MODE_INT.
10806 2018-02-20 Jeff Law <law@redhat.com>
10808 PR middle-end/82123
10809 PR tree-optimization/81592
10810 PR middle-end/79257
10811 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10812 for range data rather than using global data.
10813 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10814 range data rather than using global data.
10815 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10816 pass it to children as needed.
10817 (struct directive::fmtresult): Similarly.
10818 (struct directive::set_width): Similarly.
10819 (struct directive::set_precision): Similarly.
10820 (format_integer, format_directive, parse_directive): Similarly.
10821 (format_none): Accept unnamed vr_values parameter.
10822 (format_percent, format_floating, format_character): Similarly.
10823 (format_string, format_plain): Similarly.
10824 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10825 the EVRP range analyzer for range data rather than using global data.
10826 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10827 gimple-ssa-evrp-analyze.h
10828 (class sprintf_dom_walker): Add after_dom_children member function.
10829 Add evrp_range_analyzer member.
10830 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10831 range analyzer as needed.
10832 (sprintf_dom_walker::after_dom_children): New member function.
10833 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10835 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10836 (evrp_range_analyzer::pop_to_marker): Likewise.
10838 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10840 PR tree-optimization/84419
10841 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10842 with the required type if its current type is compatible but
10845 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10847 PR middle-end/82004
10848 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10849 after vectorization.
10851 2018-02-20 Martin Liska <mliska@suse.cz>
10854 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10855 possible values if we don't have a hint.
10857 2018-02-20 Martin Liska <mliska@suse.cz>
10861 * final.c (shorten_branches): Build align_tab array with one
10863 * opts.c (finish_options): Add alignment option limit check.
10864 (MAX_CODE_ALIGN): Likewise.
10865 (MAX_CODE_ALIGN_VALUE): Likewise.
10866 * doc/invoke.texi: Document maximum allowed option value for
10867 all -falign-* options.
10869 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10872 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10873 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10874 * var-tracking.c (emit_note_insn_var_location): Remove all references
10875 to NOTE_INSN_CALL_ARG_LOCATION.
10876 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10877 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10878 Use copy_rtx_if_shared.
10879 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10880 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10881 (dwarf2out_var_location): Remove handling of
10882 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10884 * final.c (final_scan_insn): Remove all references to
10885 NOTE_INSN_CALL_ARG_LOCATION.
10886 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10887 before dumping final insns.
10888 * except.c (emit_note_eh_region_end): Remove all references to
10889 NOTE_INSN_CALL_ARG_LOCATION.
10890 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10891 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10892 * config/arc/arc.c (hwloop_optimize): Likewise.
10893 * config/arm/arm.c (create_fix_barrier): Likewise.
10894 * config/s390/s390.c (s390_chunkify_start): Likewise.
10895 * config/sh/sh.c (find_barrier): Likewise.
10896 * config/i386/i386.c (rest_of_insert_endbranch,
10897 ix86_seh_fixup_eh_fallthru): Likewise.
10898 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10899 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10900 * config/frv/frv.c (frv_function_prologue): Likewise.
10901 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10903 (note_outside_basic_block_p): Remove all references to
10904 NOTE_INSN_CALL_ARG_LOCATION.
10905 * gengtype.c (adjust_field_rtx_def): Likewise.
10906 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10908 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10909 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10912 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10915 PR tree-optimization/84452
10916 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10917 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10920 2018-02-19 Martin Liska <mliska@suse.cz>
10923 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10925 2018-02-19 Martin Liska <mliska@suse.cz>
10926 Richard Sandiford <richard.sandiford@linaro.org>
10928 PR tree-optimization/82491
10929 * gimple-fold.c (get_base_constructor): Make earlier bail out
10932 2018-02-19 Carl Love <cel@us.ibm.com>
10934 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10935 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10937 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10938 P8V_BUILTIN_VEC_NEG.
10940 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10942 * config/rl78/rl78.md (movdf): New define expand.
10944 2018-02-19 Martin Liska <mliska@suse.cz>
10947 * doc/invoke.texi: Fix typo.
10948 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10950 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10952 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10953 handle rs6000_single_float and rs6000_double_float specially for
10956 2018-02-16 Jeff Law <law@redhat.com>
10958 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10959 (pop_regs): Use it.
10961 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10964 * ipa-inline.c (inline_small_functions): Fix a typo.
10966 2018-02-16 Nathan Sidwell <nathan@acm.org>
10968 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10970 2018-02-16 Carl Love <cel@us.ibm.com>
10972 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10973 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10974 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10975 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10976 expansion to P8V_BUILTIN_VEC_FLOAT2.
10978 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10980 PR rtl-optimization/70023
10981 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10982 src_regno into account.
10984 2018-02-16 Carl Love <cel@us.ibm.com>
10986 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10987 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10988 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10989 * config/rs6000/rs6000.c: Remove case statements for
10990 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10991 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10992 and P9V_BUILTIN_VEC_VINSERT4B.
10993 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10994 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10995 * config/rs6000/vsx.md:
10996 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10999 2018-02-16 Carl Love <cel@us.ibm.com>
11001 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11003 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11005 * config/rs6000/rs6000-c.c: Add the definitions for
11006 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11007 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11008 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11009 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11010 definition for insert4b and define insn *insert3b_internal.
11011 * doc/extend.texi: Add documentation for vec_extract4b.
11013 2018-02-16 Nathan Sidwell <nathan@acm.org>
11015 * doc/extend.texi (Backwards Compatibility): Mention friend
11016 injection. Note for-scope is deprecated.
11017 * doc/invoke.texi (-ffriend-injection): Deprecate.
11019 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11021 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11022 that moved to I2, also allow destinations that are a paradoxical
11023 subreg (instead of a normal reg).
11025 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11028 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11031 2018-02-16 Richard Biener <rguenther@suse.de>
11033 PR tree-optimization/84037
11034 PR tree-optimization/84016
11036 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11037 Adjust vec_construct for the fact we need additional higher latency
11038 128bit inserts for AVX256 and AVX512 vector builds.
11039 (ix86_add_stmt_cost): Scale vector construction cost for
11042 2018-02-16 Richard Biener <rguenther@suse.de>
11044 PR tree-optimization/84417
11045 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11046 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11047 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11049 2018-02-16 Martin Liska <mliska@suse.cz>
11052 * internal-fn.def (ASAN_CHECK): Set proper flags.
11053 (ASAN_MARK): Likewise.
11055 2018-02-16 Julia Koval <julia.koval@intel.com>
11057 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11058 from PTA_CANNONLAKE.
11060 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11063 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11064 Use ++iter rather than iter++ for std::list iterators.
11065 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11066 defer deleting them until all nodes in the forest are processed. Do
11067 free even leaf nodes. Change to_process into auto_vec.
11070 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11071 * vec.h (vec_default_construct): Use memset instead of placement new
11072 if BROKEN_VALUE_INITIALIZATION is defined.
11073 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11074 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11077 PR rtl-optimization/83723
11078 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11079 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11080 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11082 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11084 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11086 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11088 PR rtl-optimization/81443
11089 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11090 from inner REGs to paradoxical SUBREGs.
11092 2018-02-16 Richard Biener <rguenther@suse.de>
11094 PR tree-optimization/84399
11095 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11096 For operands we can analyze at their definition make sure we can
11097 analyze them at each use as well.
11099 2018-02-16 Richard Biener <rguenther@suse.de>
11101 PR tree-optimization/84190
11102 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11103 volatile accesses if the decl isn't volatile.
11105 2018-02-15 Jason Merrill <jason@redhat.com>
11107 PR c++/84314 - ICE with templates and fastcall attribute.
11108 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11109 TYPE_CANONICAL on an existing type.
11111 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11113 PR tree-optimization/84383
11114 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11115 dstoff nor call operand_equal_p if dstbase is NULL.
11117 PR tree-optimization/84334
11118 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11119 also a CONSTANT_CLASS_P, punt.
11121 2018-02-14 Jim Wilson <jimw@sifive.com>
11123 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11124 first SMALL_OPERAND check. New local min_second_step. Move assert
11125 to where locals are set. Add TARGET_RVC support.
11126 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11128 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11130 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11132 2018-02-14 Martin Sebor <msebor@redhat.com>
11134 PR tree-optimization/83698
11135 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11136 arrays constrain the offset range to their bounds.
11137 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11138 (builtin_access::overlap): Avoid setting the size of overlap if it's
11140 (maybe_diag_overlap): Also consider arrays when deciding what values
11141 of offsets to include in diagnostics.
11143 2018-02-14 Martin Sebor <msebor@redhat.com>
11146 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11147 that correspond to the kind of a declaration.
11149 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11152 * config/pa/pa.md: Load address of PIC label using the linkage table
11153 if the label is nonlocal.
11155 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11157 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11158 warning message if user requests -maltivec=be.
11159 * doc/invoke.texi: Document deprecation of -maltivec=be.
11161 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11164 * config/rs6000/rs6000-c.c: Update definitions for
11165 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11166 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11168 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11171 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11172 add _get_ssp intrinsics. Remove argument from
11173 __builtin_ia32_rdssp[d|q].
11174 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11175 * config/i386/i386-builtin.def: Remove argument from
11176 __builtin_ia32_rdssp[d|q].
11177 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11178 ix86_expand_special_args_builtin for _rdssp[d|q].
11179 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11180 Clear register before usage.
11181 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11182 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11184 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11186 PR tree-optimization/84357
11187 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11188 operand 1 of an ARRAY_REF too.
11190 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11193 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11194 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11196 (set_of_reg): New struct.
11197 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11198 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11199 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11201 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11202 Split into bitclr, bitset, bitinvert patterns if appropriate.
11203 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11204 use rx_fuse_in_memory_bitop.
11205 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11206 to named insn, correct maximum insn length.
11208 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11211 * machmode.def: Define a complex mode for PARTIAL_INT.
11212 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11214 * doc/rtl.texi: Document CSPImode.
11215 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11217 (msp430_hard_regno_nregs_with_padding): Likewise.
11219 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11222 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11224 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11226 PR rtl-optimization/84169
11227 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11228 we generated a parallel as new i3 and we split that to new i2 and i3
11229 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11230 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11231 those to i2, not i1. Partially rewrite this scan code.
11233 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11236 * stor-layout.c (place_field): For variable length fields, adjust
11237 offset_align afterwards not just based on the field's alignment,
11238 but also on the size.
11240 PR middle-end/84309
11241 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11242 of exps and logs in the use_exp2 case.
11244 2018-02-13 Jeff Law <law@redhat.com>
11246 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11247 entry for "vector".
11249 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11252 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11256 * common.opt (gas-loc-support, gas-locview-support): New.
11257 (ginline-points, ginternal-reset-location-views): New.
11258 * doc/invoke.texi: Document them. Use @itemx where intended.
11259 (gvariable-location-views): Adjust.
11260 * target.def (reset_location_view): New.
11261 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11262 (TARGET_RESET_LOCATION_VIEW): New.
11263 * doc/tm.texi: Rebuilt.
11264 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11265 (dwarf2out_default_as_locview_support): New.
11266 (output_asm_line_debug_info): Use option variables.
11267 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11268 (output_loc_list): Likewise.
11269 (add_high_low_attributes): Check option variables.
11270 Don't output entry view attribute in strict mode.
11271 (gen_inlined_subroutine_die): Check option variables.
11272 (dwarf2out_inline_entry): Likewise.
11273 (init_sections_and_labels): Likewise.
11274 (dwarf2out_early_finish): Likewise.
11275 (maybe_reset_location_view): New, from...
11276 (dwarf2out_var_location): ... here. Call it.
11277 * debug.h (dwarf2out_default_as_loc_support): Declare.
11278 (dwarf2out_default_as_locview_support): Declare.
11279 * hooks.c (hook_int_rtx_insn_0): New.
11280 * hooks.h (hook_int_rtx_insn_0): Declare.
11281 * toplev.c (process_options): Take -gas-loc-support and
11282 -gas-locview-support from dwarf2out. Enable
11283 -gvariable-location-views by default only with locview
11284 assembler support. Enable -ginternal-reset-location-views by
11285 default only if the target defines the corresponding hook.
11286 Enable -ginline-points by default if location views are
11287 enabled; force it disabled if statement frontiers are
11289 * tree-inline.c (expand_call_inline): Check option variables.
11290 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11292 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11294 PR tree-optimization/84321
11295 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11296 handling. Also check whether the anti-range contains any values
11297 that satisfy the mask; switch to a VR_RANGE if not.
11299 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11302 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11304 2018-02-13 Martin Jambor <mjambor@suse.cz>
11307 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11308 of call statements, also set location of a load to a temporary.
11310 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11312 * config/rl78/rl78.c (add_vector_labels): New function.
11313 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11314 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11315 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11316 which checks that no arguments are passed.
11317 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11318 * doc/extend.texi: Documentation for the new attribute.
11320 2018-02-13 Andreas Schwab <schwab@suse.de>
11322 * config/riscv/linux.h (CPP_SPEC): Define.
11324 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11327 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11328 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11329 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11330 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11331 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11332 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11333 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11334 temporarily for AES and PCLMUL builtins.
11336 PR tree-optimization/84339
11337 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11338 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11341 PR middle-end/84309
11342 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11343 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11344 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11346 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11348 * omp-simd-clone.h: New file.
11349 * omp-simd-clone.c: Include omp-simd-clone.h.
11350 (expand_simd_clones): No longer static.
11351 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11352 cgraph.h and omp-simd-clone.h.
11353 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11354 (vect_recog_widen_shift_pattern): Formatting fix.
11355 (vect_pattern_recog_1): Don't check optab for calls.
11358 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11359 operands[2] into a REG before using gen_lowpart on it.
11361 2018-02-12 Jeff Law <law@redhat.com>
11364 * config/sh/sh.c (find_barrier): Consider a sibling call
11367 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11368 successfully back substituting a reg.
11370 2018-02-12 Richard Biener <rguenther@suse.de>
11372 PR tree-optimization/84037
11373 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11374 parameter, move visited init to caller.
11375 (vect_slp_analyze_operations): Separate cost from validity
11376 check, initialize visited once for all instances.
11377 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11379 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11381 (vect_model_promotion_demotion_cost): Likewise.
11382 (vectorizable_bswap): Guard cost modeling with !slp_node
11383 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11385 (vectorizable_call): Likewise.
11386 (vectorizable_conversion): Likewise.
11387 (vectorizable_assignment): Likewise.
11388 (vectorizable_shift): Likewise.
11389 (vectorizable_operation): Likewise.
11390 (vectorizable_store): Likewise.
11391 (vectorizable_load): Likewise.
11392 (vectorizable_condition): Likewise.
11393 (vectorizable_comparison): Likewise.
11395 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11398 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11399 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11400 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11402 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11404 PR middle-end/83665
11405 * params.def (inline-min-speedup): Increase from 8 to 15.
11406 (max-inline-insns-auto): Decrease from 40 to 30.
11407 * ipa-split.c (consider_split): Add some buffer for function to
11408 be considered inlining candidate.
11409 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11412 2018-02-12 Richard Biener <rguenther@suse.de>
11414 PR tree-optimization/84037
11415 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11416 matched stmts if we cannot swap the non-matched ones.
11418 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11420 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11421 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11422 _mm_maskz_scalef_round_ss): New intrinsics.
11423 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11424 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11425 __builtin_ia32_scalefss_round): Remove.
11426 (__builtin_ia32_scalefsd_mask_round,
11427 __builtin_ia32_scalefss_mask_round): New intrinsics.
11428 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11429 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11430 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11431 "<round_constraint>")): Changed to ...
11432 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11433 "<round_scalar_constraint>")): ... this.
11434 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11435 %0, %1, %2<round_op3>}"): Changed to ...
11436 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11437 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11438 %2<round_scalar_mask_op3>}"): ... this.
11439 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11441 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11443 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11444 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11445 (_mm_maskz_sqrt_round_ss): New intrinsics.
11446 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11447 (__builtin_ia32_sqrtsd_mask_round)
11448 (__builtin_ia32_sqrtss_mask_round): New builtins.
11449 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11450 (__builtin_ia32_sqrtss_round): Remove.
11451 (__builtin_ia32_sqrtsd_mask_round)
11452 (__builtin_ia32_sqrtss_mask_round): New builtins.
11453 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11454 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11455 ((match_operand:VF_128 1 "vector_operand"
11456 "xBm,<round_constraint>")): Changed to ...
11457 ((match_operand:VF_128 1 "vector_operand"
11458 "xBm,<round_scalar_constraint>")): ... this.
11459 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11460 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11461 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11462 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11463 %<iptr>1<round_scalar_mask_op3>}): ... this.
11464 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11465 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11467 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11470 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11471 Cast vec_cmpeq result to correct type.
11472 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11473 Cast vec_cmpgt result to correct type.
11475 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11477 * final.c (final_scan_insn_1): Renamed from...
11478 (final_scan_insn): ... this. New wrapper, to recover
11479 seen from the outermost call in recursive ones.
11480 * config/sparc/sparc.c (output_return): Drop seen from call.
11481 (output_sibcall): Likewise.
11482 * config/visium/visium.c (output_branch): Likewise.
11484 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11486 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11489 2018-02-10 Alan Modra <amodra@gmail.com>
11492 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11493 Specify LR as an input.
11495 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11498 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11499 remove_member_access_dummy_vars): New functions.
11500 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11501 lower_omp_1, execute_lower_omp): Use them.
11503 PR rtl-optimization/84308
11504 * shrink-wrap.c (spread_components): Release todo vector.
11506 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11508 PR rtl-optimization/57193
11509 * ira-color.c (struct allocno_color_data): Add member
11510 conflict_allocno_hard_prefs.
11511 (update_conflict_allocno_hard_prefs): New.
11512 (bucket_allocno_compare_func): Add a preference based on
11513 conflict_allocno_hard_prefs.
11514 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11515 (color_allocnos): Remove a dead code. Initiate
11516 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11518 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11521 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11522 constraint from =wa to wa. Avoid a subreg on the output operand,
11523 instead use a pseudo and subreg it in a move.
11524 (p9_xxbrd_<mode>): Changed to ...
11525 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11526 (p9_xxbrd_v2df): New expander.
11527 (p9_xxbrw_<mode>): Changed to ...
11528 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11529 (p9_xxbrw_v4sf): New expander.
11531 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11533 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11535 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11538 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11539 multiply in 32-bit mode.
11540 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11541 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11544 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11546 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11547 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11548 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11549 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11551 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11554 * dwarf2out.c (is_trivial_indirect_ref): New function.
11555 (dwarf2out_late_global_decl): Do not generate a location
11556 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11557 and that are not defined in the current unit.
11559 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11561 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11562 instead of a libcall for UNORDERED.
11564 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11567 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11568 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11570 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11573 * config/s390/s390.c (s390_set_current_function): Invoke
11574 s390_indirect_branch_settings also if fndecl didn't change.
11576 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11578 * config/rs6000/rs6000.md (blockage): Set length to zero.
11580 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11582 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11584 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11587 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11588 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11592 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11593 PARALLEL incoming that failed vt_get_decl_and_offset check.
11595 PR middle-end/84237
11596 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11597 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11599 (get_variable_section): For decls in named .bss* sections pass true as
11600 second argument to bss_initializer_p.
11602 2018-02-09 Marek Polacek <polacek@redhat.com>
11603 Jakub Jelinek <jakub@redhat.com>
11606 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11607 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11608 Sync some changes from cxx_fold_indirect_ref.
11610 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11612 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11614 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11615 (BLOCK_INLINE_ENTRY_LABEL): New.
11616 (dwarf2out_var_location): Disregard inline entry markers.
11617 (inline_entry_data): New struct.
11618 (inline_entry_data_hasher): New hashtable type.
11619 (inline_entry_data_hasher::hash): New.
11620 (inline_entry_data_hasher::equal): New.
11621 (inline_entry_data_table): New variable.
11622 (add_high_low_attributes): Add DW_AT_entry_pc and
11623 DW_AT_GNU_entry_view attributes if a pending entry is found
11624 in inline_entry_data_table. Add old entry_pc attribute only
11625 if debug nonbinding markers are disabled.
11626 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11627 markers are enabled.
11628 (block_within_block_p, dwarf2out_inline_entry): New.
11629 (dwarf2out_finish): Check that no entries remained in
11630 inline_entry_data_table.
11631 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11632 (final_scan_insn, notice_source_line): Likewise.
11633 (rest_of_clean_state): Skip inline entry markers.
11634 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11636 * gimple.c (gimple_build_debug_inline_entry): New.
11637 * gimple.h (enum gimple_debug_subcode): Add
11638 GIMPLE_DEBUG_INLINE_ENTRY.
11639 (gimple_build_debug_inline_entry): Declare.
11640 (gimple_debug_inline_entry_p): New.
11641 (gimple_debug_nonbind_marker_p): Adjust.
11642 * insn-notes.def (INLINE_ENTRY): New.
11643 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11644 inline entry marker notes.
11645 (print_insn): Likewise.
11646 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11647 (INSN_DEBUG_MARKER_KIND): Likewise.
11648 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11649 * tree-inline.c (expand_call_inline): Build and insert
11650 debug_inline_entry stmt.
11651 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11652 inline entry blocks early, if nonbind markers are enabled.
11653 (dump_scope_block): Dump fragment info.
11654 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11655 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11656 (gimple_build_debug_inline_entry): New.
11657 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11658 Enable/disable inline entry points too.
11659 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11660 (DEBUG_INSN): Describe inline entry markers.
11662 * common.opt (gvariable-location-views): New.
11663 (gvariable-location-views=incompat5): New.
11664 * config.in: Rebuilt.
11665 * configure: Rebuilt.
11666 * configure.ac: Test assembler for view support.
11667 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11668 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11669 * dwarf2out.c (var_loc_view): New typedef.
11670 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11671 (dwarf2out_locviews_in_attribute): New.
11672 (dwarf2out_locviews_in_loclist): New.
11673 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11674 (enum dw_line_info_opcode): Add LI_adv_address.
11675 (struct dw_line_info_table): Add view.
11676 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11677 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11678 (zero_view_p): New variable.
11679 (ZERO_VIEW_P): New macro.
11680 (output_asm_line_debug_info): New.
11681 (struct var_loc_node): Add view.
11682 (add_AT_view_list, AT_loc_list): New.
11683 (add_var_loc_to_decl): Add view param. Test it against last.
11684 (new_loc_list): Add view params. Record them.
11685 (AT_loc_list_ptr): Handle loc and view lists.
11686 (view_list_to_loc_list_val_node): New.
11687 (print_dw_val): Handle dw_val_class_view_list.
11688 (size_of_die): Likewise.
11689 (value_format): Likewise.
11690 (loc_list_has_views): New.
11691 (gen_llsym): Set vl_symbol too.
11692 (maybe_gen_llsym, skip_loc_list_entry): New.
11693 (dwarf2out_maybe_output_loclist_view_pair): New.
11694 (output_loc_list): Output view list or entries too.
11695 (output_view_list_offset): New.
11696 (output_die): Handle dw_val_class_view_list.
11697 (output_dwarf_version): New.
11698 (output_compilation_unit_header): Use it.
11699 (output_skeleton_debug_sections): Likewise.
11700 (output_rnglists, output_line_info): Likewise.
11701 (output_pubnames, output_aranges): Update version comments.
11702 (output_one_line_info_table): Output view numbers in asm comments.
11703 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11704 Call maybe_gen_llsym.
11705 (loc_list_from_tree_1): Adjust.
11706 (add_AT_location_description): Create view list attribute if
11707 needed, check it's absent otherwise.
11708 (convert_cfa_to_fb_loc_list): Adjust.
11709 (maybe_emit_file): Call output_asm_line_debug_info for test.
11710 (dwarf2out_var_location): Reset views as needed. Precompute
11711 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11712 attribute. Set view.
11713 (new_line_info_table): Reset next view.
11714 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11715 (dwarf2out_source_line): Likewise. Output view resets and labels to
11716 the assembler, or select appropriate line info opcodes.
11717 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11718 (optimize_string_length): Catch it. Adjust.
11719 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11720 dw_val_class_view_list, and remove it if no longer needed.
11721 (hash_loc_list): Hash view numbers.
11722 (loc_list_hasher::equal): Compare them.
11723 (optimize_location_lists): Check whether a view list symbol is
11724 needed, and whether the locview attribute is present, and
11725 whether they match. Remove the locview attribute if no longer
11727 (index_location_lists): Call skip_loc_list_entry for test.
11728 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11729 Use output_dwarf_version.
11730 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11731 (struct dw_val_node): Add val_view_list.
11732 * final.c (SEEN_NEXT_VIEW): New.
11733 (set_next_view_needed): New.
11734 (clear_next_view_needed): New.
11735 (maybe_output_next_view): New.
11736 (final_start_function): Rename to...
11737 (final_start_function_1): ... this. Take pointer to FIRST,
11738 add SEEN parameter. Emit param bindings in the initial view.
11739 (final_start_function): Reintroduce SEEN-less interface.
11740 (final): Rename to...
11741 (final_1): ... this. Take SEEN parameter. Output final pending
11742 next view at the end.
11743 (final): Reintroduce seen-less interface.
11744 (final_scan_insn): Output pending next view before switching
11745 sections or ending a block. Mark the next view as needed when
11746 outputting variable locations. Notify debug backend of section
11747 changes, and of location view changes.
11748 (rest_of_handle_final): Adjust.
11749 * toplev.c (process_options): Autodetect value for debug variable
11750 location views option. Warn on incompat5 without -gdwarf-5.
11751 * doc/invoke.texi (gvariable-location-views): New.
11752 (gvariable-location-views=incompat5): New.
11753 (gno-variable-location-views): New.
11755 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11757 PR tree-optimization/84136
11758 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11759 that the result of find_edge is non-NULL.
11761 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11764 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11765 storing integer register in SImode. Fix cost of 256 and 512
11766 byte aligned SSE register store.
11768 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11770 * config/i386/i386.c (ix86_multiplication_cost): Fix
11771 multiplication cost for TARGET_AVX512DQ.
11773 2018-02-08 Marek Polacek <polacek@redhat.com>
11775 PR tree-optimization/84238
11776 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11779 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11781 PR tree-optimization/84265
11782 * tree-vect-stmts.c (vectorizable_store): Don't treat
11783 VMAT_CONTIGUOUS accesses as grouped.
11784 (vectorizable_load): Likewise.
11786 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11788 PR tree-optimization/81635
11789 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11790 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11791 (test_round_for_mask): New functions.
11792 (wide_int_cc_tests): Call test_round_for_mask.
11793 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11794 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11795 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11796 range returned by get_range_info.
11798 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11801 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11802 * symtab.c: Include builtins.h
11803 (symtab_node::output_to_lto_symbol_table_p): Move here
11804 from lto-streamer-out.c:output_symbol_p.
11805 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11806 (output_symbol_p): Move all logic to symtab.c
11807 (produce_symtab): Update.
11809 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11811 * config/s390/s390-opts.h (enum indirect_branch): Define.
11812 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11813 (s390_indirect_branch_via_thunk)
11814 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11815 (enum s390_indirect_branch_type): Define.
11816 * config/s390/s390.c (struct s390_frame_layout, struct
11817 machine_function): Remove.
11818 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11819 (indirect_branch_table_label_no, indirect_branch_table_name):
11821 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11822 (enum s390_indirect_branch_option): Define.
11823 (s390_return_addr_from_memory): New function.
11824 (s390_handle_string_attribute): New function.
11825 (s390_attribute_table): Add new attribute handler.
11826 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11827 (s390_indirect_branch_via_thunk): New function.
11828 (s390_indirect_branch_via_inline_thunk): New function.
11829 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11830 sibling call optimization for non z10 compiles.
11831 (s390_emit_call): Force indirect branch target to be a single
11832 register. Add r1 clobber for non-z10 compiles.
11833 (s390_emit_epilogue): Emit return jump via return_use expander.
11834 (s390_reorg): Handle JUMP_INSNs as execute targets.
11835 (s390_option_override_internal): Perform validity checks for the
11836 new command line options.
11837 (s390_indirect_branch_attrvalue): New function.
11838 (s390_indirect_branch_settings): New function.
11839 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11840 (s390_output_indirect_thunk_function): New function.
11841 (s390_code_end): Implement target hook.
11842 (s390_case_values_threshold): Implement target hook.
11843 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11845 * config/s390/s390.h (struct s390_frame_layout)
11846 (struct machine_function): Move here from s390.c.
11847 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11848 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11849 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11850 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11851 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11852 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11853 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11854 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11855 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11856 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11857 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11858 (mnemonic attribute): Add values which aren't recognized
11860 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11861 pattern for branch conversion. Fix mnemonic attribute.
11862 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11863 indirect branch via thunk if requested.
11864 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11865 ("*indirect_jump"): Disable for branch conversion using out of
11867 ("indirect_jump_via_thunk<mode>_z10")
11868 ("indirect_jump_via_thunk<mode>")
11869 ("indirect_jump_via_inlinethunk<mode>_z10")
11870 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11871 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11872 ("casesi_jump_via_inlinethunk<mode>_z10")
11873 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11874 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11875 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11876 ("*indirect2_jump"): Disable for branch conversion.
11877 ("casesi_jump"): Turn into expander and expand patterns for branch
11879 ("return_use"): New expander.
11880 ("*return"): Emit return via thunk and rename it to ...
11881 ("*return<mode>"): ... this one.
11882 * config/s390/s390.opt: Add new options and and enum for the
11885 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11887 * lra-constraints.c (match_reload): Unconditionally use
11888 gen_lowpart_SUBREG, rather than selecting between that
11889 and equivalent gen_rtx_SUBREG code.
11891 2018-02-08 Richard Biener <rguenther@suse.de>
11893 PR tree-optimization/84233
11894 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11895 changed flag instead of boguously re-using phi_inserted.
11897 2018-02-08 Martin Jambor <mjambor@suse.cz>
11899 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11900 static local variables.
11902 2018-02-08 Richard Biener <rguenther@suse.de>
11904 PR tree-optimization/84278
11905 * tree-vect-stmts.c (vectorizable_store): When looking for
11906 smaller vector types to perform grouped strided loads/stores
11907 make sure the mode is supported by the target.
11908 (vectorizable_load): Likewise.
11910 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11912 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11913 Increase LDP/STP opportunities by adding adjacent callee-saves.
11915 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11917 PR rtl-optimization/84068
11918 PR rtl-optimization/83459
11919 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11921 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11923 PR tree-optimization/84224
11924 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11925 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11926 non-zero arguments.
11928 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11931 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11932 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11933 count, remove one USE.
11935 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11937 * doc/install.texi (Configuration): Document the
11938 --with-long-double-format={ibm,ieee} PowerPC configuration
11942 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11943 Convert from define_expand to be define_insn_and_split. Rework
11944 float/double/_Float128 conversions to QI/HI/SImode to work with
11945 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11946 conversions to QI/HImode types did a store and then a load to
11947 truncate the value. For conversions to VSX registers, don't split
11948 the insn, instead emit the code directly. Use the code iterator
11949 any_fix to combine signed and unsigned conversions.
11950 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11951 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11952 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11953 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11954 (fix_<mode>di2_hw): Likewise.
11955 (fixuns_<mode>di2_hw): Likewise.
11956 (fix_<mode>si2_hw): Likewise.
11957 (fixuns_<mode>si2_hw): Likewise.
11958 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11959 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11960 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11961 fix<uns>_trunc<SFDF:mode>si2_p8.
11962 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11963 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11964 (fix<uns>_<mode>_mem): Likewise.
11965 (fctiw<u>z_<mode>_mem): Likewise.
11966 (fix<uns>_<mode>_mem): Likewise.
11967 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11968 the register allocator from doing a direct move to the GPRs to do
11969 a store, and instead use the ISA 3.0 store byte/half-word from
11970 vector register instruction. For IEEE 128-bit floating point,
11971 also optimize stores of 32-bit ints.
11972 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11974 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11976 * genextract.c (push_pathstr_operand): New function to support
11978 (walk_rtx): Call push_pathstr_operand.
11979 (print_path): Support [a-zA-Z].
11981 2018-02-07 Richard Biener <rguenther@suse.de>
11983 PR tree-optimization/84037
11984 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11985 (cse_and_gimplify_to_preheader): Declare.
11986 (vect_get_place_in_interleaving_chain): Likewise.
11987 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11989 (_loop_vec_info::~_loop_vec_info): Delete it.
11990 (cse_and_gimplify_to_preheader): New function.
11991 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11992 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11993 (vectorizable_load): Likewise. For grouped stores always base
11994 the IV on the first element.
11995 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11996 condition before gimplifying.
11998 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12000 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12001 *DIV_EXPR and *MOD_EXPR.
12003 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12006 * config/i386/i386.c (ix86_option_override_internal): Mask out
12007 the CF_SET bit when checking -fcf-protection.
12009 2018-02-07 Tom de Vries <tom@codesourcery.com>
12012 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12015 2018-02-07 Richard Biener <rguenther@suse.de>
12017 PR tree-optimization/84204
12018 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12021 PR tree-optimization/84205
12022 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12023 special-case isl_ast_op_zdiv_r.
12025 PR tree-optimization/84223
12026 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12027 Only add conditions from within the region.
12028 (gather_bbs::after_dom_children): Adjust.
12030 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12033 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12034 * config/avr/avr.md: Only post-reload split REG-REG moves if
12035 either register is GENERAL_REG_P.
12037 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12039 PR tree-optimization/84235
12040 * tree-ssa-scopedtables.c
12041 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12042 if the subtraction is performed in floating point type where NaNs are
12043 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12044 build 1. Formatting fix.
12046 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12049 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12050 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12051 and skip it regardless of bb boundaries. Use CALL_P macro,
12052 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12053 unnecessarily, formatting fix.
12055 2018-02-06 Michael Collison <michael.collison@arm.com>
12057 * config/arm/thumb2.md:
12058 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12059 (*thumb_mov_notscc): Ditto.
12061 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12064 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12065 unsigned_fix, not "s".
12067 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12069 * configure.ac (gcc_fn_eh_frame_ro): New function.
12070 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12071 correct .eh_frame permissions.
12072 * configure: Regenerate.
12074 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12076 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12077 irrelevant options.
12079 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12081 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12082 Display warning message for -mno-speculate-indirect-jumps.
12084 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12086 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12088 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12090 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12092 PR tree-optimization/84225
12093 * tree-eh.c (find_trapping_overflow): Only call
12094 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12096 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12099 * config/i386/i386.c: Reimplement the check of possible options
12100 -mibt/-mshstk conbination. Change error messages.
12101 * doc/invoke.texi: Fix a typo: remove extra '='.
12103 2018-02-06 Marek Polacek <polacek@redhat.com>
12105 PR tree-optimization/84228
12106 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12108 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12111 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12112 emitted arch directives.
12113 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12114 __ARM_FEATURE_COPROC before changing architectures.
12116 2018-02-06 Richard Biener <rguenther@suse.de>
12118 * config/i386/i386.c (print_reg): Fix typo.
12119 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12121 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12123 * configure: Regenerate.
12125 2018-02-05 Martin Sebor <msebor@redhat.com>
12127 PR tree-optimization/83369
12128 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12131 2018-02-05 Martin Liska <mliska@suse.cz>
12133 * doc/invoke.texi: Cherry-pick upstream r323995.
12135 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12137 * ira.c (ira_init_register_move_cost): Adjust comment.
12139 2018-02-05 Martin Liska <mliska@suse.cz>
12141 PR gcov-profile/84137
12142 * doc/gcov.texi: Fix typo in documentation.
12144 2018-02-05 Martin Liska <mliska@suse.cz>
12146 PR gcov-profile/83879
12147 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12148 using dlopen functionality.
12150 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12152 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12153 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12154 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12155 _mm_maskz_range_round_ss): New intrinsics.
12156 (__builtin_ia32_rangesd128_round)
12157 (__builtin_ia32_rangess128_round): Remove.
12158 (__builtin_ia32_rangesd128_mask_round,
12159 __builtin_ia32_rangess128_mask_round): New builtins.
12160 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12161 __builtin_ia32_rangess128_round): Remove.
12162 (__builtin_ia32_rangesd128_mask_round,
12163 __builtin_ia32_rangess128_mask_round): New builtins.
12164 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12165 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12166 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12167 "<round_saeonly_constraint>")): Changed to ...
12168 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12169 "<round_saeonly_scalar_constraint>")): ... this.
12170 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12171 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12172 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12173 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12174 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12176 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12178 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12180 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12181 Remove all values except native, 8540 and 8548.
12183 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12185 * config/i386/i386.c (ix86_output_function_return): Pass
12186 INVALID_REGNUM, instead of -1, as invalid register number to
12187 indirect_thunk_name and output_indirect_thunk.
12189 2018-02-02 Julia Koval <julia.koval@intel.com>
12191 * config.gcc: Add -march=icelake.
12192 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12193 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12194 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12195 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12196 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12197 (processor_target_table): Add icelake.
12198 (ix86_option_override_internal): Handle new PTAs.
12199 (get_builtin_code_for_version): Handle icelake.
12200 (M_INTEL_COREI7_ICELAKE): New.
12201 (fold_builtin_cpu): Handle icelake.
12202 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12203 * doc/invoke.texi: Add -march=icelake.
12205 2018-02-02 Julia Koval <julia.koval@intel.com>
12207 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12208 to wide_int_bitmask.
12209 * wide-int-bitmask.h: New.
12211 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12214 * config/i386/i386.md: Replace Pmode with word_mode in
12215 builtin_setjmp_setup and builtin_longjmp to support x32.
12217 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12221 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12223 (rs6000_supported_cpu_names): New static variable.
12224 (linux_cpu_translation_table): Likewise.
12225 (elf_platform) <cpu>: Define new static variable and use it.
12226 Translate kernel AT_PLATFORM name to canonical name if needed.
12227 Error if platform name is unknown.
12229 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12232 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12234 2018-02-01 Jeff Law <law@redhat.com>
12237 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12238 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12239 the scratch if RELEASE_VIA_POP is false.
12240 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12241 If we have to save a temporary register, decrement SIZE appropriately.
12242 Pass new arguments to release_scratch_register_on_entry.
12243 (ix86_adjust_stack_and_probe): Likewise.
12244 (ix86_emit_probe_stack_range): Pass new arguments to
12245 release_scratch_register_on_entry.
12247 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12249 PR rtl-optimization/84157
12250 * combine.c (change_zero_ext): Use REG_P predicate in
12251 front of HARD_REGISTER_P predicate.
12253 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12255 * config/avr/avr.c (avr_option_override): Move disabling of
12256 -fdelete-null-pointer-checks to...
12257 * common/config/avr/avr-common.c (avr_option_optimization_table):
12260 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12262 PR tree-optimization/81635
12263 * tree-data-ref.c (split_constant_offset_1): For types that
12264 wrap on overflow, try to use range info to prove that wrapping
12267 2018-02-01 Renlin Li <renlin.li@arm.com>
12270 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12271 TAILCALL_ADDR_REGS.
12272 (aarch64_register_move_cost): Likewise.
12273 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12274 TAILCALL_ADDR_REGS.
12275 (REG_CLASS_NAMES): Likewise.
12276 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12277 TAILCALL_ADDR_REGS. Remove IP registers.
12278 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12280 2018-02-01 Richard Biener <rguenther@suse.de>
12282 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12283 for specifying RPO order and allow NULL for that.
12284 * domwalk.c (dom_walker::dom_walker): Likewise.
12285 (dom_walker::walk): Handle NULL RPO order.
12286 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12288 (rewrite_update_dom_walker): Likewise.
12289 (mark_def_dom_walker): Likewise.
12291 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12293 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12294 (aarch64_maybe_expand_sve_subreg_move): Declare.
12295 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12296 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12298 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12299 that are semantically a reverse operation.
12300 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12301 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12302 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12304 (aarch64_can_change_mode_class): For big-endian, forbid changes
12305 between two SVE modes if they have different element sizes.
12307 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12309 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12310 the TImode handling for big-endian targets.
12312 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12314 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12315 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12317 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12318 Remove BSWAP handing for big-endian targets and use the form of
12319 LD1RQ appropariate for the mode.
12321 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12323 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12324 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12325 duplicated element.
12327 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12330 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12331 check for operands that need to go through aarch64_sve_reload_be.
12333 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12335 PR tree-optimization/81661
12336 PR tree-optimization/84117
12337 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12338 * tree-eh.c: Include gimplify.h.
12339 (find_trapping_overflow, replace_trapping_overflow,
12340 rewrite_to_non_trapping_overflow): New functions.
12341 * tree-vect-loop.c: Include tree-eh.h.
12342 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12343 * tree-data-ref.c: Include tree-eh.h.
12344 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12346 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12348 PR rtl-optimization/84123
12349 * combine.c (change_zero_ext): Check if hard register satisfies
12350 can_change_dest_mode before calling gen_lowpart_SUBREG.
12352 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12355 * ira.c (ira_init_register_move_cost): Remove assert.
12357 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12359 PR rtl-optimization/84071
12360 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12361 * doc/tm.texi: Regenerate.
12363 2018-01-31 Richard Biener <rguenther@suse.de>
12365 PR tree-optimization/84132
12366 * tree-data-ref.c (analyze_miv_subscript): Properly
12367 check whether evolution_function_is_affine_multivariate_p
12368 before calling gcd_of_steps_may_divide_p.
12370 2018-01-31 Julia Koval <julia.koval@intel.com>
12373 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12374 * config/i386/i386.md (rdpid_rex64) New.
12375 (rdpid): Make 32bit only.
12377 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12380 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12381 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12383 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12386 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12388 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12390 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12392 PR rtl-optimization/84071
12393 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12394 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12396 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12398 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12399 (arc_attribute_table): Add 'aux' attribute.
12400 (arc_in_small_data_p): Consider aux like variables.
12401 (arc_is_aux_reg_p): New function.
12402 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12403 (arc_get_aux_arg): New function.
12404 (prepare_move_operands): Handle aux-register access.
12405 (arc_handle_aux_attribute): New function.
12406 * doc/extend.texi (ARC Variable attributes): Add subsection.
12408 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12410 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12411 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12412 (arc_attribute_table): Add 'uncached' attribute.
12413 (arc_print_operand): Print '.di' flag for uncached memory
12415 (arc_in_small_data_p): Do not consider for small data the uncached
12417 (arc_is_uncached_mem_p): New function.
12418 * config/arc/predicates.md (compact_store_memory_operand): Check
12419 for uncached memory accesses.
12420 (nonvol_nonimm_operand): Likewise.
12421 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12423 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12426 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12427 falign-loops=): Add Optimization flag.
12429 2018-01-30 Jeff Law <law@redhat.com>
12432 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12433 INT_REGISTERS_SAVED. Check it prior to calling
12434 get_scratch_register_on_entry.
12435 (ix86_adjust_stack_and_probe): Similarly.
12436 (ix86_emit_probe_stack_range): Similarly.
12437 (ix86_expand_prologue): Corresponding changes.
12439 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12442 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12443 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12445 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12448 * lra-constraints.c (curr_insn_transform): Process AND in the
12451 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12453 PR rtl-optimization/83986
12454 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12455 dependence against last_pending_memory_flush in addition to
12456 pending_jump_insns.
12458 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12460 PR tree-optimization/81611
12461 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12464 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12467 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12470 2018-01-30 Richard Biener <rguenther@suse.de>
12471 Jakub Jelinek <jakub@redhat.com>
12473 PR tree-optimization/84111
12474 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12475 inner loops added during recursion, as they don't have up-to-date
12478 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12481 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12482 (can_inline_edge_by_limits_p): ... here.
12483 (can_early_inline_edge_p, check_callers,
12484 update_caller_keys, update_callee_keys, recursive_inlining,
12485 add_new_edges_to_heap, speculation_useful_p,
12486 inline_small_functions,
12487 inline_small_functions, flatten_function,
12488 inline_to_all_callers_1): Update.
12490 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12492 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12495 2018-01-30 Richard Biener <rguenther@suse.de>
12497 PR tree-optimization/83008
12498 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12499 invariant and constant vector uses in stmts when they need
12500 more than one stmt.
12502 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12505 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12506 * configure: Regenerate.
12508 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12510 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12512 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12513 Use gen_rtx_REG rather than gen_lowpart.
12515 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12517 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12518 rather than 0 when creating partial subregs.
12520 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12522 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12525 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12528 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12529 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12530 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12531 flags. This restores the settings used before the 2017-07-24.
12532 Turning off pre increment/decrement/modify allows IVOPTS to
12533 optimize DF/SF loops where the index is an int.
12535 2018-01-29 Richard Biener <rguenther@suse.de>
12536 Kelvin Nilsen <kelvin@gcc.gnu.org>
12539 * tree-vect-stmts.c (vectorizable_call): Don't call
12540 targetm.vectorize_builtin_md_vectorized_function if callee is
12543 2018-01-22 Carl Love <cel@us.ibm.com>
12545 * doc/extend.tex: Fix typo in second arg in
12546 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12548 2018-01-29 Richard Biener <rguenther@suse.de>
12550 PR tree-optimization/84086
12551 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12552 (flush_ssaname_freelist): When SSA names were released reset
12553 the SCEV hash table.
12555 2018-01-29 Richard Biener <rguenther@suse.de>
12557 PR tree-optimization/84057
12558 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12559 removed paths when removing edges.
12561 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12563 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12564 -mfunction-return=@var{choice}.
12566 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12568 PR diagnostic/84034
12569 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12570 Handle CR like TAB.
12571 (layout::print_source_line): Likewise.
12572 (test_get_line_width_without_trailing_whitespace): Add test cases.
12574 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12576 PR middle-end/84040
12577 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12580 2018-01-26 Jim Wilson <jimw@sifive.com>
12582 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12584 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12587 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12589 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12590 and CMP + SUB-immediate -> SUBS.
12592 2018-01-26 Martin Sebor <msebor@redhat.com>
12594 PR tree-optimization/83896
12595 * tree-ssa-strlen.c (get_string_len): Rename...
12596 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12597 Avoid assuming length is constant.
12598 (handle_char_store): Use HOST_WIDE_INT for string length.
12600 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12603 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12604 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12606 2018-01-26 Richard Biener <rguenther@suse.de>
12608 PR rtl-optimization/84003
12609 * dse.c (record_store): Only record redundant stores when
12610 the earlier store aliases at least all accesses the later one does.
12612 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12614 PR rtl-optimization/83985
12615 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12616 REG_CFA_RESTORE insns.
12617 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12618 REG_CFA_RESTORE insns here.
12621 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12622 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12624 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12626 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12627 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12628 (arc_init): Likewise.
12629 (arc_override_options): Likewise.
12630 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12632 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12634 * config/arc/arc.h (TARGET_DBNZ): Define.
12635 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12636 properly set the tune attribute.
12637 (dbnz): Use TARGET_DBNZ guard.
12638 * config/arc/arc.opt (mtune): Add core3 option.
12640 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12642 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12643 recognize new pic like addresses.
12644 (arc_delegitimize_address): Clean up.
12646 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12648 * config/arc/arc-arches.def: Option mrf16 valid for all
12650 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12651 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12652 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12653 * config/arc/arc-tables.opt: Regenerate.
12654 * config/arc/arc.c (arc_conditional_register_usage): Handle
12655 reduced register file case.
12656 (arc_file_start): Set must have build attributes.
12657 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12658 mrf16 option value.
12659 * config/arc/arc.opt (mrf16): Add new option.
12660 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12661 * config/arc/genmultilib.awk: Handle new mrf16 option.
12662 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12663 * config/arc/t-multilib: Regenerate.
12664 * doc/invoke.texi (ARC Options): Document mrf16 option.
12666 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12668 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12669 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12670 (arc_attribute_table): Add 'secure_call' attribute.
12671 (arc_print_operand): Print secure call operand.
12672 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12674 (arc_is_secure_call_p): New function. * config/arc/arc.md
12675 (call_i): Add support for sjli instruction.
12676 (call_value_i): Likewise.
12677 * config/arc/constraints.md (Csc): New constraint.
12679 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12680 John Eric Martin <John.Martin@emmicro-us.com>
12682 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12683 * config/arc/arc.c (_arc_jli_section): New struct.
12684 (arc_jli_section): New type.
12685 (rc_jli_sections): New static variable.
12686 (arc_handle_jli_attribute): New function.
12687 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12688 (arc_file_end): New function.
12689 (TARGET_ASM_FILE_END): Define.
12690 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12691 (arc_add_jli_section): New function.
12692 (jli_call_scan): Likewise.
12693 (arc_reorg): Call jli_call_scan.
12694 (arc_output_addsi): Remove 'S' from printing asm operand.
12695 (arc_is_jli_call_p): New function.
12696 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12698 (movhi_insn): Likewise.
12699 (movsi_insn): Likewise.
12700 (movsi_set_cc_insn): Likewise.
12701 (loadqi_update): Likewise.
12702 (load_zeroextendqisi_update): Likewise.
12703 (load_signextendqisi_update): Likewise.
12704 (loadhi_update): Likewise.
12705 (load_zeroextendhisi_update): Likewise.
12706 (load_signextendhisi_update): Likewise.
12707 (loadsi_update): Likewise.
12708 (loadsf_update): Likewise.
12709 (movsicc_insn): Likewise.
12710 (bset_insn): Likewise.
12711 (bxor_insn): Likewise.
12712 (bclr_insn): Likewise.
12713 (bmsk_insn): Likewise.
12714 (bicsi3_insn): Likewise.
12715 (cmpsi_cc_c_insn): Likewise.
12716 (movsi_ne): Likewise.
12717 (movsi_cond_exec): Likewise.
12718 (clrsbsi2): Likewise.
12719 (norm_f): Likewise.
12729 (call_i): Remove 'S' asm letter, add jli instruction.
12730 (call_value_i): Likewise.
12731 * config/arc/arc.op (mjli-always): New option.
12732 * config/arc/constraints.md (Cji): New constraint.
12733 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12735 (subsf3_fpx): Likewise.
12736 (mulsf3_fpx): Likewise.
12737 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12739 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12740 function attrbutes.
12741 * doc/invoke.texi (ARC): Document mjli-always option.
12743 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12745 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12746 avoid addition with 0 and use incw and decw where possible.
12748 2018-01-26 Richard Biener <rguenther@suse.de>
12750 PR tree-optimization/81082
12751 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12752 association if it requires casting to unsigned.
12753 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12754 from fold_plusminus_mult_expr to catch important cases late when
12755 range info is available.
12757 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12759 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12760 * configure.ac (hidden_linkonce): New test.
12761 * configure: Regenerate.
12762 * config.in: Regenerate.
12764 2018-01-26 Julia Koval <julia.koval@intel.com>
12766 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12767 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12768 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12769 _mm_mask_bitshuffle_epi64_mask): Fix type.
12770 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12771 USI_FTYPE_V4DI_V4DI_USI): Remove.
12772 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12773 __builtin_ia32_vpshufbitqmb256_mask,
12774 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12775 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12776 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12778 2018-01-26 Alan Modra <amodra@gmail.com>
12781 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12782 UNSPEC_VBPERMQ. Sort other unspecs.
12784 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12786 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12788 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12790 PR middle-end/83055
12791 * predict.c (drop_profile): Do not push/pop cfun; update also
12793 (handle_missing_profiles): Fix logic looking for zero profiles.
12795 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12797 PR middle-end/83977
12798 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12799 on functions with #pragma omp declare simd or functions with simd
12801 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12802 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12803 Remove trailing \n from warning_at calls.
12805 2018-01-25 Tom de Vries <tom@codesourcery.com>
12808 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12809 for neutered workers.
12811 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12814 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12815 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12817 2018-01-24 Jeff Law <law@redhat.com>
12820 * i386.c (get_probe_interval): Move to earlier point.
12821 (ix86_compute_frame_layout): If -fstack-clash-protection and
12822 the frame is larger than the probe interval, then use pushes
12823 to save registers rather than reg->mem moves.
12824 (ix86_expand_prologue): Remove conditional for int_registers_saved
12827 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12830 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12831 min/max for never referenced object.
12833 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12835 PR middle-end/83977
12836 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12838 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12839 attributes from DECL_ATTRIBUTES (decl) without affecting
12840 DECL_ATTRIBUTES (current_function_decl).
12841 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12842 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12844 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12846 PR tree-optimization/83979
12847 * fold-const.c (fold_comparison): Use constant_boolean_node
12848 instead of boolean_{true,false}_node.
12850 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12852 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12855 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12857 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12858 Simplify the clause that sets the length attribute.
12859 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12860 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12861 clause that sets the length attribute.
12862 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12864 2018-01-24 Tom de Vries <tom@codesourcery.com>
12867 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12868 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12869 Add strict parameter.
12870 (prevent_branch_around_nothing): Insert dummy insn between branch to
12871 label and label with no ptx insn inbetween.
12872 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12874 2018-01-24 Tom de Vries <tom@codesourcery.com>
12877 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12878 for neutered threads in warp.
12879 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12881 2018-01-24 Richard Biener <rguenther@suse.de>
12883 PR tree-optimization/83176
12884 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12887 2018-01-24 Richard Biener <rguenther@suse.de>
12889 PR tree-optimization/82819
12890 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12891 code generating pluses that are no-ops in the target precision.
12893 2018-01-24 Richard Biener <rguenther@suse.de>
12895 PR middle-end/84000
12896 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12898 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12900 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12901 to merge probabilities.
12902 * predict.c (probably_never_executed): Also mark as cold functions
12903 with global 0 profile and guessed local profile.
12904 * profile-count.c (profile_probability::combine_with_count): New
12906 * profile-count.h (profile_probability::operator*,
12907 profile_probability::operator*=, profile_probability::operator/,
12908 profile_probability::operator/=): Reduce precision to adjusted
12909 and set value to guessed on contradictory divisions.
12910 (profile_probability::combine_with_freq): Remove.
12911 (profile_probability::combine_wiht_count): Declare.
12912 (profile_count::force_nonzero):: Set to adjusted.
12913 (profile_count::probability_in):: Set quality to adjusted.
12914 * tree-ssa-tail-merge.c (replace_block_by): Use
12915 combine_with_count.
12917 2018-01-23 Andrew Waterman <andrew@sifive.com>
12918 Jim Wilson <jimw@sifive.com>
12920 * config/riscv/riscv.c (riscv_stack_boundary): New.
12921 (riscv_option_override): Set riscv_stack_boundary. Handle
12922 riscv_preferred_stack_boundary_arg.
12923 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12924 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12925 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12926 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12927 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12928 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12930 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12933 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12934 of struct ix86_frame.
12935 (ix86_expand_epilogue): Likewise. Add a local variable for
12936 the reg_save_offset field in struct ix86_frame.
12938 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12940 PR tree-optimization/82604
12941 * tree-loop-distribution.c (enum partition_kind): New enum item
12942 PKIND_PARTIAL_MEMSET.
12943 (partition_builtin_p): Support above new enum item.
12944 (generate_code_for_partition): Ditto.
12945 (compute_access_range): Differentiate cases that equality can be
12946 proven at all loops, the innermost loops or no loops.
12947 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12948 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12949 (finalize_partitions, distribute_loop): Don't fuse partition of
12950 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12951 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12952 parloop is enabled.
12954 2018-01-23 Martin Liska <mliska@suse.cz>
12956 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12957 order to ignore the predictor.
12958 (PRED_POLYMORPHIC_CALL): Likewise.
12959 (PRED_RECURSIVE_CALL): Likewise.
12961 2018-01-23 Martin Liska <mliska@suse.cz>
12963 * tree-profile.c (tree_profiling): Print function header to
12964 aware reader which function we are working on.
12965 * value-prof.c (gimple_find_values_to_profile): Do not print
12966 not interesting value histograms.
12968 2018-01-23 Martin Liska <mliska@suse.cz>
12970 * profile-count.h (enum profile_quality): Add
12971 profile_uninitialized as the first value. Do not number values
12972 as they are zero based.
12973 (profile_count::verify): Update sanity check.
12974 (profile_probability::verify): Likewise.
12976 2018-01-23 Nathan Sidwell <nathan@acm.org>
12978 * doc/invoke.texi (ffor-scope): Deprecate.
12980 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12982 PR tree-optimization/83510
12983 * domwalk.c (set_all_edges_as_executable): New function.
12984 (dom_walker::dom_walker): Convert bool param
12985 "skip_unreachable_blocks" to enum reachability. Move setup of
12986 edge flags to set_all_edges_as_executable and only do it when
12987 reachability is REACHABLE_BLOCKS.
12988 * domwalk.h (enum dom_walker::reachability): New enum.
12989 (dom_walker::dom_walker): Convert bool param
12990 "skip_unreachable_blocks" to enum reachability.
12991 (set_all_edges_as_executable): New decl.
12992 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12993 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12995 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12996 but converting true to REACHABLE_BLOCKS.
12997 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12999 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13000 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13001 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13003 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13004 if check_all_array_refs will be called.
13006 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13008 * tree.c (selftest::test_location_wrappers): Add more test
13011 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13013 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13014 (selftest::test_bit_in_range): Likewise.
13016 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13019 * doc/sourcebuild.texi (vect_float): Say that the selector
13020 only describes the situation when -funsafe-math-optimizations is on.
13021 (vect_float_strict): Document.
13023 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13025 PR tree-optimization/83965
13026 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13027 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13028 instead of checking only for a reduction.
13029 (vect_recog_widen_sum_pattern): Likewise.
13031 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13033 * predict.c (probably_never_executed): Only use precise profile info.
13034 (compute_function_frequency): Skip after inlining hack since we now
13035 have quality checking.
13037 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13039 * profile-count.h (profile_probability::very_unlikely,
13040 profile_probability::unlikely, profile_probability::even): Set
13041 precision to guessed.
13043 2018-01-23 Richard Biener <rguenther@suse.de>
13045 PR tree-optimization/83963
13046 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13047 Properly terminate dominator walk when crossing the exit edge not
13048 when visiting its source block.
13050 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13053 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13054 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13056 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13058 PR tree-optimization/83957
13059 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13060 semicolon after for body surrounded by braces.
13062 PR tree-optimization/83081
13063 * profile-count.h (profile_probability::split): New method.
13064 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13065 Use profile_probability::split.
13066 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13067 when splitting a single conditional jump into 2.
13069 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13071 PR tree-optimization/69452
13072 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13077 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13078 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13079 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13081 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13083 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13085 * config/rl78/rl78.md (movdi): New define_expand.
13086 * config/rl78/rl78.c (rl78_split_movdi): New function.
13088 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13091 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13093 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13094 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13095 128-bit to produce an UNSPEC move to get the double word with the
13096 signbit and then a shift directly to do signbit.
13097 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13098 implementation with a new version that just does either a direct
13099 move or a regular move. Move memory interface to separate insns.
13100 Move insns so they are next to the expander.
13101 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13102 with signbit move. Split big and little endian case.
13103 (signbit<mode>2_dm_mem_le): Likewise.
13104 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13105 (signbit<mode>2_dm2): Likewise.
13107 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13109 * config/rl78/rl78.md (anddi3): New define_expand.
13111 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13113 * config/rl78/rl78.md (umindi3): New define_expand.
13115 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13117 * config/rl78/rl78.md (smindi3): New define_expand.
13119 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13121 * config/rl78/rl78.md (smaxdi3): New define_expand.
13123 2018-01-22 Carl Love <cel@us.ibm.com>
13125 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13126 LVX_V1TI): Add macro expansion.
13127 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13128 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13129 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13130 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13131 Change check to determine if the instruction is a byte reversing
13132 entry. Fix typo in comment.
13133 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13134 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13135 Add def_builtin calls for new builtins.
13136 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13137 Add define_insn expansion.
13139 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13141 * config/rl78/rl78.md (umaxdi3): New define_expand.
13143 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13145 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13146 for non-QImode registers.
13148 2018-01-22 Richard Biener <rguenther@suse.de>
13150 PR tree-optimization/83963
13151 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13152 including the loop exit block.
13153 (scop_detection::merge_sese): Likewise.
13154 (scop_detection::add_scop): Do it here instead.
13156 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13158 * doc/sourcebuild.texi (arm_softfloat): Document.
13160 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13163 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13164 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13165 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13167 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13168 David Edelsohn <dje.gcc@gmail.com>
13171 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13172 Change "crset eq" to "crset 2".
13173 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13174 (*call_indirect_aix<mode>_nospec): Likewise.
13175 (*call_value_indirect_aix<mode>_nospec): Likewise.
13176 (*call_indirect_elfv2<mode>_nospec): Likewise.
13177 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13178 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13179 change assembly output from . to $.
13180 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13181 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13182 (*tablejump<mode>_internal1_nospec): Likewise.
13184 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13187 * config/sh/sh_optimize_sett_clrt.cc:
13188 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13190 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13192 PR tree-optimization/83940
13193 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13194 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13195 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13196 use it to pass back the definition type.
13197 (vect_check_store_rhs): Likewise rhs_dt_out.
13198 (vect_build_gather_load_calls): Add a mask_dt argument and use
13199 it instead of a call to vect_is_simple_use.
13200 (vectorizable_store): Update calls to vect_check_load_store_mask
13201 and vect_check_store_rhs. Use the dt returned by the latter instead
13202 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13203 instead of calls to vect_is_simple_use. Pass the scalar rather
13204 than the vector operand to vect_is_simple_use when handling
13205 second and subsequent copies of an rhs value.
13206 (vectorizable_load): Update calls to vect_check_load_store_mask
13207 and vect_build_gather_load_calls. Use the cached mask_dt and
13208 gs_info.offset_dt instead of calls to vect_is_simple_use.
13210 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13212 PR middle-end/83945
13213 * tree-emutls.c: Include gimplify.h.
13214 (lower_emutls_2): New function.
13215 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13216 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13217 it before further processing.
13220 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13221 UINTVAL (trueop1) instead of INTVAL (op1).
13223 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13227 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13228 INCOMING_FRAME_SP_OFFSET if not defined.
13229 (scan_trace): Add ENTRY argument. If true and
13230 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13231 emit a note to adjust the CFA offset.
13232 (create_cfi_notes): Adjust scan_trace callers.
13233 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13234 INCOMING_FRAME_SP_OFFSET in the CIE.
13235 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13236 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13238 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13239 * doc/tm.texi: Regenerated.
13241 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13243 PR rtl-optimization/83147
13244 * lra-constraints.c (remove_inheritance_pseudos): Use
13245 lra_substitute_pseudo_within_insn.
13247 2018-01-19 Tom de Vries <tom@codesourcery.com>
13248 Cesar Philippidis <cesar@codesourcery.com>
13251 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13253 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13256 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13257 spaces for function labels.
13259 2018-01-19 Martin Liska <mliska@suse.cz>
13261 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13262 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13263 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13264 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13265 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13266 (PRED_CONST_RETURN): Change from 69 to 65.
13267 (PRED_NULL_RETURN): Change from 91 to 71.
13268 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13269 (PRED_LOOP_GUARD): Change from 66 to 73.
13271 2018-01-19 Martin Liska <mliska@suse.cz>
13273 * predict.c (predict_insn_def): Add new assert.
13274 (struct branch_predictor): Change type to signed integer.
13275 (test_prediction_value_range): Amend test to cover
13276 PROB_UNINITIALIZED.
13277 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13278 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13279 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13280 (PRED_LOOP_IV_COMPARE): Likewise.
13281 * predict.h (PROB_UNINITIALIZED): Define new constant.
13283 2018-01-19 Martin Liska <mliska@suse.cz>
13285 * predict.c (dump_prediction): Add new format for
13286 analyze_brprob.py script which is enabled with -details
13288 * profile-count.h (precise_p): New function.
13290 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13292 PR tree-optimization/83922
13293 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13294 there are no statements that need masking.
13295 (vect_active_double_reduction_p): New function.
13296 (vect_analyze_loop_operations): Use it when handling phis that
13297 are not in the loop header.
13299 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13301 PR tree-optimization/83914
13302 * tree-vect-loop.c (vectorizable_induction): Don't convert
13303 init_expr or apply the peeling adjustment for inductions
13304 that are nested within the vectorized loop.
13306 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13308 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13311 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13315 * function.h (gimplify_parameters): Add gimple_seq * argument.
13316 * function.c: Include gimple.h and options.h.
13317 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13318 for the added local temporaries if needed.
13319 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13320 if there are any parameter cleanups, wrap whole body into a
13321 try/finally with the cleanups.
13323 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13326 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13327 Use GET_MODE_CLASS for scalar floating point.
13329 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13333 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13334 Fix call of call_cgraph_insertion_hooks.
13336 2018-01-18 Martin Sebor <msebor@redhat.com>
13338 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13340 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13343 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13346 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13349 * common.opt: (-ffile-prefix-map): New option.
13350 * opts.c (common_handle_option): Defer it.
13351 * opts-global.c (handle_common_deferred_options): Handle it.
13352 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13353 * file-prefix-map.h: New file.
13354 (remap_debug_filename, add_debug_prefix_map): ...here.
13355 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13356 * final.c (debug_prefix_map, add_debug_prefix_map
13357 remap_debug_filename): Move to...
13358 * file-prefix-map.c: New file.
13359 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13360 generalize, get rid of alloca(), use strrchr() instead of strchr().
13361 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13362 Implement in terms of add_prefix_map().
13363 (remap_macro_filename, remap_debug_filename): Implement in term of
13365 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13366 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13367 * dbxout.c: Include file-prefix-map.h.
13368 * varasm.c: Likewise.
13369 * vmsdbgout.c: Likewise.
13370 * xcoffout.c: Likewise.
13371 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13372 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13373 * doc/invoke.texi (-ffile-prefix-map): Document.
13374 (-fdebug-prefix-map): Update description.
13376 2018-01-18 Martin Liska <mliska@suse.cz>
13378 * config/i386/i386.c (indirect_thunk_name): Document that also
13380 (output_indirect_thunk): Document why both instructions
13381 (pause and lfence) are generated.
13383 2018-01-18 Richard Biener <rguenther@suse.de>
13385 PR tree-optimization/83887
13386 * graphite-scop-detection.c
13387 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13388 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13389 (scop_detection::merge_sese): Re-implement with a flood-fill
13390 algorithm that properly finds a SESE region if it exists.
13392 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13395 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13396 pointer_diff optimizations use view_convert instead of convert.
13398 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13400 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13401 Generate different code for -mno-speculate-indirect-jumps.
13402 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13403 (*call_indirect_aix<mode>): Disable for
13404 -mno-speculate-indirect-jumps.
13405 (*call_indirect_aix<mode>_nospec): New define_insn.
13406 (*call_value_indirect_aix<mode>): Disable for
13407 -mno-speculate-indirect-jumps.
13408 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13409 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13410 -mno-speculate-indirect-jumps.
13411 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13413 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13415 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13416 long double type, set the flags for noting the default long double
13417 type, even if we don't pass or return a long double type.
13419 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13422 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13425 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13427 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13428 support for merge[hl].
13429 (fold_mergehl_helper): New helper function.
13430 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13431 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13432 (altivec_vmrglw_direct): Add xxmrglw insn.
13434 2018-01-17 Andrew Waterman <andrew@sifive.com>
13436 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13437 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13439 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13442 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13443 call the lto_location_cache before reading the
13444 DECL_SOURCE_LOCATION of the types.
13446 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13447 Richard Sandiford <richard.sandiford@linaro.org>
13449 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13450 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13451 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13452 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13453 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13455 * config/aarch64/constraints.md (aarch64_movti_operand):
13457 * config/aarch64/predicates.md (Uti): Add new constraint.
13459 2018-01-17 Carl Love <cel@us.ibm.com>
13461 * config/rs6000/vsx.md (define_expand xl_len_r,
13462 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13463 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13465 (define_expand, define_insn): Move the shift left from the
13466 define_insn to the define_expand for lxvl and stxvl instructions.
13467 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13468 and XL_LEN_R definitions to PURE.
13470 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13472 * config/i386/i386.c (indirect_thunk_name): Declare regno
13473 as unsigned int. Compare regno with INVALID_REGNUM.
13474 (output_indirect_thunk): Ditto.
13475 (output_indirect_thunk_function): Ditto.
13476 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13477 in the call to output_indirect_thunk_function.
13479 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13481 PR middle-end/83884
13482 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13483 rather than the size of inner_type to determine the stack slot size
13484 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13486 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13489 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13492 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13494 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13495 endian Linux systems to optionally enable multilibs for selecting
13496 the long double type if the user configured an explicit type.
13497 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13498 have no long double multilibs if not defined.
13499 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13500 warn if the user used -mabi={ieee,ibm}longdouble and we built
13501 multilibs for long double.
13502 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13503 appropriate multilib option.
13504 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13506 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13507 for building long double multilibs.
13508 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13510 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13512 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13515 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13517 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13520 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13523 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13526 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13528 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13529 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13531 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13533 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13534 different rtl trees depending on TARGET_64BIT.
13535 (rs6000_gen_lvx): Likewise.
13537 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13539 * config/visium/visium.md (nop): Tweak comment.
13540 (hazard_nop): Likewise.
13542 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13544 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13545 -mspeculate-indirect-jumps.
13546 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13547 for -mno-speculate-indirect-jumps.
13548 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13549 (*call_value_indirect_elfv2<mode>): Disable for
13550 -mno-speculate-indirect-jumps.
13551 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13552 (indirect_jump): Emit different RTL for
13553 -mno-speculate-indirect-jumps.
13554 (*indirect_jump<mode>): Disable for
13555 -mno-speculate-indirect-jumps.
13556 (*indirect_jump<mode>_nospec): New define_insn.
13557 (tablejump): Emit different RTL for
13558 -mno-speculate-indirect-jumps.
13559 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13560 (tablejumpsi_nospec): New define_expand.
13561 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13562 (tablejumpdi_nospec): New define_expand.
13563 (*tablejump<mode>_internal1): Disable for
13564 -mno-speculate-indirect-jumps.
13565 (*tablejump<mode>_internal1_nospec): New define_insn.
13566 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13569 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13571 * caller-save.c (insert_save): Drop unnecessary parameter. All
13574 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13575 Richard Biener <rguenth@suse.de>
13578 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13579 return early, inline manually is_gimple_sizepos. Make sure if we
13580 call gimplify_expr we don't end up with a gimple constant.
13581 * tree.c (variably_modified_type_p): Don't return true for
13582 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13583 * gimplify.h (is_gimple_sizepos): Remove.
13585 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13587 PR tree-optimization/83857
13588 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13589 vectorizable_live_operation for pure SLP statements.
13590 (vectorizable_live_operation): Handle PHIs.
13592 2018-01-16 Richard Biener <rguenther@suse.de>
13594 PR tree-optimization/83867
13595 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13596 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13598 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13601 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13602 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13603 If off is not INTEGER_CST, issue a may not be aligned warning
13604 rather than isn't aligned. Use isn%'t rather than isn't.
13605 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13607 <case MULT_EXPR>: Improve the case when bottom and one of the
13608 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13609 operand, in that case check if the other operand is multiple of
13610 bottom divided by the INTEGER_CST operand.
13612 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13615 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13616 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13617 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13618 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13619 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13620 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13621 (pa_function_arg_size): New function.
13623 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13625 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13626 in a separate statement.
13628 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13630 PR tree-optimization/83847
13631 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13632 group gathers and scatters.
13634 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13636 PR rtl-optimization/86620
13637 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13639 PR rtl-optimization/83213
13640 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13641 to last if both are JUMP_INSNs.
13643 PR tree-optimization/83843
13644 * gimple-ssa-store-merging.c
13645 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13646 store_immediate_info for bswap/nop orig_stores.
13648 2018-01-15 Andrew Waterman <andrew@sifive.com>
13650 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13652 <UDIV>: Increase cost if !TARGET_DIV.
13654 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13656 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13657 (define_attr "cr_logical_3op"): New.
13658 (cceq_ior_compare): Adjust.
13659 (cceq_ior_compare_complement): Adjust.
13660 (*cceq_rev_compare): Adjust.
13661 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13662 (is_cracked_insn): Adjust.
13663 (insn_must_be_first_in_group): Adjust.
13664 * config/rs6000/40x.md: Adjust.
13665 * config/rs6000/440.md: Adjust.
13666 * config/rs6000/476.md: Adjust.
13667 * config/rs6000/601.md: Adjust.
13668 * config/rs6000/603.md: Adjust.
13669 * config/rs6000/6xx.md: Adjust.
13670 * config/rs6000/7450.md: Adjust.
13671 * config/rs6000/7xx.md: Adjust.
13672 * config/rs6000/8540.md: Adjust.
13673 * config/rs6000/cell.md: Adjust.
13674 * config/rs6000/e300c2c3.md: Adjust.
13675 * config/rs6000/e500mc.md: Adjust.
13676 * config/rs6000/e500mc64.md: Adjust.
13677 * config/rs6000/e5500.md: Adjust.
13678 * config/rs6000/e6500.md: Adjust.
13679 * config/rs6000/mpc.md: Adjust.
13680 * config/rs6000/power4.md: Adjust.
13681 * config/rs6000/power5.md: Adjust.
13682 * config/rs6000/power6.md: Adjust.
13683 * config/rs6000/power7.md: Adjust.
13684 * config/rs6000/power8.md: Adjust.
13685 * config/rs6000/power9.md: Adjust.
13686 * config/rs6000/rs64.md: Adjust.
13687 * config/rs6000/titan.md: Adjust.
13689 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13691 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13692 ix86_indirect_branch_register logic.
13694 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13696 * config/i386/constraints.md (Bs): Update
13697 ix86_indirect_branch_register check. Don't check
13698 ix86_indirect_branch_register with GOT_memory_operand.
13700 * config/i386/predicates.md (GOT_memory_operand): Don't check
13701 ix86_indirect_branch_register here.
13702 (GOT32_symbol_operand): Likewise.
13704 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13706 * config/i386/predicates.md (constant_call_address_operand):
13707 Rewrite ix86_indirect_branch_register logic.
13708 (sibcall_insn_operand): Likewise.
13710 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13712 * config/i386/constraints.md (Bs): Replace
13713 ix86_indirect_branch_thunk_register with
13714 ix86_indirect_branch_register.
13716 * config/i386/i386.md (indirect_jump): Likewise.
13717 (tablejump): Likewise.
13718 (*sibcall_memory): Likewise.
13719 (*sibcall_value_memory): Likewise.
13720 Peepholes of indirect call and jump via memory: Likewise.
13721 * config/i386/i386.opt: Likewise.
13722 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13723 (GOT_memory_operand): Likewise.
13724 (call_insn_operand): Likewise.
13725 (sibcall_insn_operand): Likewise.
13726 (GOT32_symbol_operand): Likewise.
13728 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13730 PR middle-end/83837
13731 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13732 type rather than type addr's type points to.
13733 (expand_omp_atomic_mutex): Likewise.
13734 (expand_omp_atomic): Likewise.
13736 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13739 * config/i386/i386.c (output_indirect_thunk_function): Use
13740 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13741 for __x86_return_thunk.
13743 2018-01-15 Richard Biener <rguenther@suse.de>
13745 PR middle-end/83850
13746 * expmed.c (extract_bit_field_1): Fix typo.
13748 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13751 * config/arm/iterators.md (VF): New mode iterator.
13752 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13753 Remove integer-related logic from pattern.
13754 (neon_vabd<mode>_3): Likewise.
13756 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13758 PR middle-end/82694
13759 * common.opt (fstrict-overflow): No longer an alias.
13760 (fwrapv-pointer): New option.
13761 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13762 also for pointer types based on flag_wrapv_pointer.
13763 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13764 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13765 opts->x_flag_wrapv got set.
13766 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13767 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13768 POINTER_TYPE_OVERFLOW_UNDEFINED.
13769 * match.pd: Likewise in address comparison pattern.
13770 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13772 2018-01-15 Richard Biener <rguenther@suse.de>
13775 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13776 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13777 Reset type names to their identifier if their TYPE_DECL doesn't
13778 have linkage (and thus is used for ODR and devirt).
13779 (save_debug_info_for_decl): Remove.
13780 (save_debug_info_for_type): Likewise.
13781 (add_tree_to_fld_list): Adjust.
13782 * tree-pretty-print.c (dump_generic_node): Make dumping of
13783 type names more robust.
13785 2018-01-15 Richard Biener <rguenther@suse.de>
13787 * BASE-VER: Bump to 8.0.1.
13789 2018-01-14 Martin Sebor <msebor@redhat.com>
13792 * builtins.c (check_access): Avoid warning when the no-warning bit
13795 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13797 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13798 * ira-color (allocno_hard_regs_compare): Likewise.
13800 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13803 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13804 Use .pushsection/.popsection.
13806 2018-01-14 Martin Sebor <msebor@redhat.com>
13809 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13811 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13813 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13814 entry from extra_headers.
13815 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13816 extra_headers, make the list bitwise identical to the i?86-*-* one.
13818 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13820 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13821 -mcmodel=large with -mindirect-branch=thunk,
13822 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13823 -mfunction-return=thunk-extern.
13824 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13825 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13826 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13828 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13830 * config/i386/i386.c (print_reg): Print the name of the full
13831 integer register without '%'.
13832 (ix86_print_operand): Handle 'V'.
13833 * doc/extend.texi: Document 'V' modifier.
13835 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13837 * config/i386/constraints.md (Bs): Disallow memory operand for
13838 -mindirect-branch-register.
13840 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13841 (GOT_memory_operand): Likewise.
13842 (call_insn_operand): Likewise.
13843 (sibcall_insn_operand): Likewise.
13844 (GOT32_symbol_operand): Likewise.
13845 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13846 for -mindirect-branch-register.
13847 (tablejump): Likewise.
13848 (*sibcall_memory): Likewise.
13849 (*sibcall_value_memory): Likewise.
13850 Disallow peepholes of indirect call and jump via memory for
13851 -mindirect-branch-register.
13852 (*call_pop): Replace m with Bw.
13853 (*call_value_pop): Likewise.
13854 (*sibcall_pop_memory): Replace m with Bs.
13855 * config/i386/i386.opt (mindirect-branch-register): New option.
13856 * doc/invoke.texi: Document -mindirect-branch-register option.
13858 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13860 * config/i386/i386-protos.h (ix86_output_function_return): New.
13861 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13862 set function_return_type.
13863 (indirect_thunk_name): Add ret_p to indicate thunk for function
13865 (output_indirect_thunk_function): Pass false to
13866 indirect_thunk_name.
13867 (ix86_output_indirect_branch_via_reg): Likewise.
13868 (ix86_output_indirect_branch_via_push): Likewise.
13869 (output_indirect_thunk_function): Create alias for function
13870 return thunk if regno < 0.
13871 (ix86_output_function_return): New function.
13872 (ix86_handle_fndecl_attribute): Handle function_return.
13873 (ix86_attribute_table): Add function_return.
13874 * config/i386/i386.h (machine_function): Add
13875 function_return_type.
13876 * config/i386/i386.md (simple_return_internal): Use
13877 ix86_output_function_return.
13878 (simple_return_internal_long): Likewise.
13879 * config/i386/i386.opt (mfunction-return=): New option.
13880 (indirect_branch): Mention -mfunction-return=.
13881 * doc/extend.texi: Document function_return function attribute.
13882 * doc/invoke.texi: Document -mfunction-return= option.
13884 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13886 * config/i386/i386-opts.h (indirect_branch): New.
13887 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13888 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13889 with local indirect jump when converting indirect call and jump.
13890 (ix86_set_indirect_branch_type): New.
13891 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13892 (indirectlabelno): New.
13893 (indirect_thunk_needed): Likewise.
13894 (indirect_thunk_bnd_needed): Likewise.
13895 (indirect_thunks_used): Likewise.
13896 (indirect_thunks_bnd_used): Likewise.
13897 (INDIRECT_LABEL): Likewise.
13898 (indirect_thunk_name): Likewise.
13899 (output_indirect_thunk): Likewise.
13900 (output_indirect_thunk_function): Likewise.
13901 (ix86_output_indirect_branch_via_reg): Likewise.
13902 (ix86_output_indirect_branch_via_push): Likewise.
13903 (ix86_output_indirect_branch): Likewise.
13904 (ix86_output_indirect_jmp): Likewise.
13905 (ix86_code_end): Call output_indirect_thunk_function if needed.
13906 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13908 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13909 (ix86_attribute_table): Add indirect_branch.
13910 * config/i386/i386.h (machine_function): Add indirect_branch_type
13911 and has_local_indirect_jump.
13912 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13914 (tablejump): Likewise.
13915 (*indirect_jump): Use ix86_output_indirect_jmp.
13916 (*tablejump_1): Likewise.
13917 (simple_return_indirect_internal): Likewise.
13918 * config/i386/i386.opt (mindirect-branch=): New option.
13919 (indirect_branch): New.
13922 (thunk-inline): Likewise.
13923 (thunk-extern): Likewise.
13924 * doc/extend.texi: Document indirect_branch function attribute.
13925 * doc/invoke.texi: Document -mindirect-branch= option.
13927 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13930 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13932 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13934 * ipa-inline.c (want_inline_small_function_p): Return false if
13935 inlining has already failed with CIF_FINAL_ERROR.
13936 (update_caller_keys): Call want_inline_small_function_p before
13938 (update_callee_keys): Likewise.
13940 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13942 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13944 (rs6000_quadword_masked_address_p): Likewise.
13945 (quad_aligned_load_p): Likewise.
13946 (quad_aligned_store_p): Likewise.
13947 (const_load_sequence_p): Add comment to describe the outer-most loop.
13948 (mimic_memory_attributes_and_flags): New function.
13949 (rs6000_gen_stvx): Likewise.
13950 (replace_swapped_aligned_store): Likewise.
13951 (rs6000_gen_lvx): Likewise.
13952 (replace_swapped_aligned_load): Likewise.
13953 (replace_swapped_load_constant): Capitalize argument name in
13954 comment describing this function.
13955 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13956 and stores that access quad-word aligned addresses and replace
13957 with stvx or lvx instructions when appropriate.
13958 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13959 New function prototype.
13960 (rs6000_quadword_masked_address_p): Likewise.
13961 (rs6000_gen_lvx): Likewise.
13962 (rs6000_gen_stvx): Likewise.
13963 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13964 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13965 when memory address is aligned.
13966 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13967 this split to select lvx instruction when memory address is aligned.
13968 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13969 instruction when memory address is aligned.
13970 (*vsx_le_perm_load_v16qi): Likewise.
13971 (four unnamed splitters): Modify to select the stvx instruction
13972 when memory is aligned.
13974 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13976 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13977 which appears in the queue multiple times.
13979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13980 Alan Hayward <alan.hayward@arm.com>
13981 David Sherwood <david.sherwood@arm.com>
13983 * tree-vectorizer.h (vec_lower_bound): New structure.
13984 (_loop_vec_info): Add check_nonzero and lower_bounds.
13985 (LOOP_VINFO_CHECK_NONZERO): New macro.
13986 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13987 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13988 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13989 fields. Make seg_len the distance travelled, not including the
13991 (dr_direction_indicator): Declare.
13992 (dr_zero_step_indicator): Likewise.
13993 (dr_known_forward_stride_p): Likewise.
13994 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13996 (runtime_alias_check_p): Allow runtime alias checks with
13998 (operator ==): Compare access_size and align.
13999 (prune_runtime_alias_test_list): Rework for new distinction between
14000 the access_size and seg_len.
14001 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14003 (get_segment_min_max): New function.
14004 (create_intersect_range_checks): Use it.
14005 (dr_step_indicator): New function.
14006 (dr_direction_indicator): Likewise.
14007 (dr_zero_step_indicator): Likewise.
14008 (dr_known_forward_stride_p): Likewise.
14009 * tree-loop-distribution.c (data_ref_segment_size): Return
14010 DR_STEP * (niters - 1).
14011 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14013 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14014 (vect_preserves_scalar_order_p): New function, split out from...
14015 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14016 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14017 (vect_vfa_access_size): New function.
14018 (vect_vfa_align): Likewise.
14019 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14020 (dump_lower_bound): New function.
14021 (vect_check_lower_bound): Likewise.
14022 (vect_small_gap_p): Likewise.
14023 (vectorizable_with_step_bound_p): Likewise.
14024 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14025 depencies if the vectorization factor is 1. Convert the checks
14026 for nonzero steps into checks on the bounds of DR_STEP. Try using
14027 a bunds check for variable steps if the minimum required step is
14028 relatively small. Update calls to the dr_with_seg_len
14029 constructor and to vect_compile_time_alias.
14030 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14032 (vect_loop_versioning): Call it.
14033 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14035 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14038 Alan Hayward <alan.hayward@arm.com>
14039 David Sherwood <david.sherwood@arm.com>
14041 * doc/sourcebuild.texi (vect_scatter_store): Document.
14042 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14044 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14046 * genopinit.c (main): Add supports_vec_scatter_store and
14047 supports_vec_scatter_store_cached to target_optabs.
14048 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14049 IFN_MASK_SCATTER_STORE.
14050 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14052 * internal-fn.h (internal_store_fn_p): Declare.
14053 (internal_fn_stored_value_index): Likewise.
14054 * internal-fn.c (scatter_store_direct): New macro.
14055 (expand_scatter_store_optab_fn): New function.
14056 (direct_scatter_store_optab_supported_p): New macro.
14057 (internal_store_fn_p): New function.
14058 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14059 IFN_MASK_SCATTER_STORE.
14060 (internal_fn_mask_index): Likewise.
14061 (internal_fn_stored_value_index): New function.
14062 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14063 for scatter stores.
14064 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14065 * optabs-query.c (supports_vec_scatter_store_p): New function.
14066 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14067 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14068 true for scatter stores.
14069 (vect_gather_scatter_fn_p): Handle scatter stores too.
14070 (vect_check_gather_scatter): Consider using scatter stores if
14071 supports_vec_scatter_store_p.
14072 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14073 scatter stores too.
14074 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14075 internal_fn_stored_value_index.
14076 (check_load_store_masking): Handle scatter stores too.
14077 (vect_get_store_rhs): Make public.
14078 (vectorizable_call): Use internal_store_fn_p.
14079 (vectorizable_store): Handle scatter store internal functions.
14080 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14081 when deciding whether the end of the group has been reached.
14082 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14083 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14084 (mask_scatter_store<mode>): New insns.
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_gather_scatter_fn_p): Declare.
14091 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14092 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14094 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14095 Use vect_truncate_gather_scatter_offset if we can't treat the
14096 operation as a normal gather load or scatter store.
14097 (get_group_load_store_type): Take the gather_scatter_info
14098 as argument. Try using a gather load or scatter store for
14099 single-element groups.
14100 (get_load_store_type): Update calls to get_group_load_store_type
14101 and vect_use_strided_gather_scatters_p.
14103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14104 Alan Hayward <alan.hayward@arm.com>
14105 David Sherwood <david.sherwood@arm.com>
14107 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14108 optional tree argument.
14109 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14111 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14112 but continue to use the current value as a fallback.
14113 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14114 to compare the updates.
14115 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14116 (get_load_store_type): Use it when handling a strided access.
14117 (vect_get_strided_load_store_ops): New function.
14118 (vect_get_data_ptr_increment): Likewise.
14119 (vectorizable_load): Handle strided gather loads. Always pass
14120 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14123 Alan Hayward <alan.hayward@arm.com>
14124 David Sherwood <david.sherwood@arm.com>
14126 * doc/md.texi (gather_load@var{m}): Document.
14127 (mask_gather_load@var{m}): Likewise.
14128 * genopinit.c (main): Add supports_vec_gather_load and
14129 supports_vec_gather_load_cached to target_optabs.
14130 * optabs-tree.c (init_tree_optimization_optabs): Use
14131 ggc_cleared_alloc to allocate target_optabs.
14132 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14133 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14135 * internal-fn.h (internal_load_fn_p): Declare.
14136 (internal_gather_scatter_fn_p): Likewise.
14137 (internal_fn_mask_index): Likewise.
14138 (internal_gather_scatter_fn_supported_p): Likewise.
14139 * internal-fn.c (gather_load_direct): New macro.
14140 (expand_gather_load_optab_fn): New function.
14141 (direct_gather_load_optab_supported_p): New macro.
14142 (direct_internal_fn_optab): New function.
14143 (internal_load_fn_p): Likewise.
14144 (internal_gather_scatter_fn_p): Likewise.
14145 (internal_fn_mask_index): Likewise.
14146 (internal_gather_scatter_fn_supported_p): Likewise.
14147 * optabs-query.c (supports_at_least_one_mode_p): New function.
14148 (supports_vec_gather_load_p): Likewise.
14149 * optabs-query.h (supports_vec_gather_load_p): Declare.
14150 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14151 and memory_type field.
14152 (NUM_PATTERNS): Bump to 15.
14153 * tree-vect-data-refs.c: Include internal-fn.h.
14154 (vect_gather_scatter_fn_p): New function.
14155 (vect_describe_gather_scatter_call): Likewise.
14156 (vect_check_gather_scatter): Try using internal functions for
14157 gather loads. Recognize existing calls to a gather load function.
14158 (vect_analyze_data_refs): Consider using gather loads if
14159 supports_vec_gather_load_p.
14160 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14161 (vect_get_gather_scatter_offset_type): Likewise.
14162 (vect_convert_mask_for_vectype): Likewise.
14163 (vect_add_conversion_to_patterm): Likewise.
14164 (vect_try_gather_scatter_pattern): Likewise.
14165 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14166 (vect_vect_recog_func_ptrs): Add it.
14167 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14168 internal_fn_mask_index and internal_gather_scatter_fn_p.
14169 (check_load_store_masking): Take the gather_scatter_info as an
14170 argument and handle gather loads.
14171 (vect_get_gather_scatter_ops): New function.
14172 (vectorizable_call): Check internal_load_fn_p.
14173 (vectorizable_load): Likewise. Handle gather load internal
14175 (vectorizable_store): Update call to check_load_store_masking.
14176 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14177 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14178 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14179 (aarch64_gather_scale_operand_d): New predicates.
14180 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14181 (mask_gather_load<mode>): New insns.
14183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14184 Alan Hayward <alan.hayward@arm.com>
14185 David Sherwood <david.sherwood@arm.com>
14187 * optabs.def (fold_left_plus_optab): New optab.
14188 * doc/md.texi (fold_left_plus_@var{m}): Document.
14189 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14190 * internal-fn.c (fold_left_direct): Define.
14191 (expand_fold_left_optab_fn): Likewise.
14192 (direct_fold_left_optab_supported_p): Likewise.
14193 * fold-const-call.c (fold_const_fold_left): New function.
14194 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14195 * tree-parloops.c (valid_reduction_p): New function.
14196 (gather_scalar_reductions): Use it.
14197 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14198 (vect_finish_replace_stmt): Declare.
14199 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14200 (needs_fold_left_reduction_p): New function, split out from...
14201 (vect_is_simple_reduction): ...here. Accept reductions that
14202 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14203 (vect_force_simple_reduction): Also store the reduction type in
14204 the assignment's STMT_VINFO_REDUC_TYPE.
14205 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14206 (merge_with_identity): New function.
14207 (vect_expand_fold_left): Likewise.
14208 (vectorize_fold_left_reduction): Likewise.
14209 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14210 scalar phi in place for it. Check for target support and reject
14211 cases that would reassociate the operation. Defer the transform
14212 phase to vectorize_fold_left_reduction.
14213 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14214 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14215 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14219 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14220 call to ifc_temp_var.
14222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14223 Alan Hayward <alan.hayward@arm.com>
14224 David Sherwood <david.sherwood@arm.com>
14226 * target.def (legitimize_address_displacement): Take the original
14227 offset as a poly_int.
14228 * targhooks.h (default_legitimize_address_displacement): Update
14230 * targhooks.c (default_legitimize_address_displacement): Likewise.
14231 * doc/tm.texi: Regenerate.
14232 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14233 as an argument, moving assert of ad->disp == ad->disp_term to...
14234 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14235 Try calling targetm.legitimize_address_displacement before expanding
14236 the address rather than afterwards, and adjust for the new interface.
14237 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14238 Match the new hook interface. Handle SVE addresses.
14239 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14240 new hook interface.
14242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14244 * Makefile.in (OBJS): Add early-remat.o.
14245 * target.def (select_early_remat_modes): New hook.
14246 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14247 * doc/tm.texi: Regenerate.
14248 * targhooks.h (default_select_early_remat_modes): Declare.
14249 * targhooks.c (default_select_early_remat_modes): New function.
14250 * timevar.def (TV_EARLY_REMAT): New timevar.
14251 * passes.def (pass_early_remat): New pass.
14252 * tree-pass.h (make_pass_early_remat): Declare.
14253 * early-remat.c: New file.
14254 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14256 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14259 Alan Hayward <alan.hayward@arm.com>
14260 David Sherwood <david.sherwood@arm.com>
14262 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14263 vfm1 with a bound_epilog parameter.
14264 (vect_do_peeling): Update calls accordingly, and move the prologue
14265 call earlier in the function. Treat the base bound_epilog as 0 for
14266 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14267 this base when peeling for gaps.
14268 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14269 with fully-masked loops.
14270 (vect_estimate_min_profitable_iters): Handle the single peeled
14271 iteration in that case.
14273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14274 Alan Hayward <alan.hayward@arm.com>
14275 David Sherwood <david.sherwood@arm.com>
14277 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14278 single-element interleaving even if the size is not a power of 2.
14279 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14280 accesses for single-element interleaving if the group size is
14283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14284 Alan Hayward <alan.hayward@arm.com>
14285 David Sherwood <david.sherwood@arm.com>
14287 * doc/md.texi (fold_extract_last_@var{m}): Document.
14288 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14289 * optabs.def (fold_extract_last_optab): New optab.
14290 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14291 * internal-fn.c (fold_extract_direct): New macro.
14292 (expand_fold_extract_optab_fn): Likewise.
14293 (direct_fold_extract_optab_supported_p): Likewise.
14294 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14295 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14296 EXTRACT_LAST_REDUCTION.
14297 (get_initial_def_for_reduction): Do not create an initial vector
14298 for EXTRACT_LAST_REDUCTION reductions.
14299 (vectorizable_reduction): Leave the scalar phi in place for
14300 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14301 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14302 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14303 transform phase to vectorizable_condition.
14304 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14306 (vect_finish_stmt_generation): ...here.
14307 (vect_finish_replace_stmt): New function.
14308 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14309 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14311 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14314 Alan Hayward <alan.hayward@arm.com>
14315 David Sherwood <david.sherwood@arm.com>
14317 * doc/md.texi (extract_last_@var{m}): Document.
14318 * optabs.def (extract_last_optab): New optab.
14319 * internal-fn.def (EXTRACT_LAST): New internal function.
14320 * internal-fn.c (cond_unary_direct): New macro.
14321 (expand_cond_unary_optab_fn): Likewise.
14322 (direct_cond_unary_optab_supported_p): Likewise.
14323 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14324 loops using EXTRACT_LAST.
14325 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14326 (extract_last_<mode>): ...this optab.
14327 (vec_extract<mode><Vel>): Update accordingly.
14329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14330 Alan Hayward <alan.hayward@arm.com>
14331 David Sherwood <david.sherwood@arm.com>
14333 * target.def (empty_mask_is_expensive): New hook.
14334 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14335 * doc/tm.texi: Regenerate.
14336 * targhooks.h (default_empty_mask_is_expensive): Declare.
14337 * targhooks.c (default_empty_mask_is_expensive): New function.
14338 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14339 if the target says that empty masks are expensive.
14340 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14342 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14345 Alan Hayward <alan.hayward@arm.com>
14346 David Sherwood <david.sherwood@arm.com>
14348 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14349 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14350 (vect_use_loop_mask_for_alignment_p): New function.
14351 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14352 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14353 niters_skip argument. Make sure that the first niters_skip elements
14354 of the first iteration are inactive.
14355 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14356 Update call to vect_set_loop_masks_directly.
14357 (get_misalign_in_elems): New function, split out from...
14358 (vect_gen_prolog_loop_niters): ...here.
14359 (vect_update_init_of_dr): Take a code argument that specifies whether
14360 the adjustment should be added or subtracted.
14361 (vect_update_init_of_drs): Likewise.
14362 (vect_prepare_for_masked_peels): New function.
14363 (vect_do_peeling): Skip prologue peeling if we're using a mask
14364 instead. Update call to vect_update_inits_of_drs.
14365 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14367 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14368 alignment. Do not include the number of peeled iterations in
14369 the minimum threshold in that case.
14370 (vectorizable_induction): Adjust the start value down by
14371 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14372 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14373 Take the number of skipped iterations into account when calculating
14375 * tree-vect-stmts.c (vect_gen_while_not): New function.
14377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14378 Alan Hayward <alan.hayward@arm.com>
14379 David Sherwood <david.sherwood@arm.com>
14381 * doc/sourcebuild.texi (vect_fully_masked): Document.
14382 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14383 default value to 0.
14384 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14386 (vect_analyze_loop_2): ...here. Don't check the vectorization
14387 factor against the number of loop iterations if the loop is
14390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14391 Alan Hayward <alan.hayward@arm.com>
14392 David Sherwood <david.sherwood@arm.com>
14394 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14395 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14396 (dump_groups): Update accordingly.
14397 (iv_use::mem_type): New member variable.
14398 (address_p): New function.
14399 (record_use): Add a mem_type argument and initialize the new
14401 (record_group_use): Add a mem_type argument. Use address_p.
14402 Remove obsolete null checks of base_object. Update call to record_use.
14403 (find_interesting_uses_op): Update call to record_group_use.
14404 (find_interesting_uses_cond): Likewise.
14405 (find_interesting_uses_address): Likewise.
14406 (get_mem_type_for_internal_fn): New function.
14407 (find_address_like_use): Likewise.
14408 (find_interesting_uses_stmt): Try find_address_like_use before
14409 calling find_interesting_uses_op.
14410 (addr_offset_valid_p): Use the iv mem_type field as the type
14411 of the addressed memory.
14412 (add_autoinc_candidates): Likewise.
14413 (get_address_cost): Likewise.
14414 (split_small_address_groups_p): Use address_p.
14415 (split_address_groups): Likewise.
14416 (add_iv_candidate_for_use): Likewise.
14417 (autoinc_possible_for_pair): Likewise.
14418 (rewrite_groups): Likewise.
14419 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14420 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14421 (get_alias_ptr_type_for_ptr_address): New function.
14422 (rewrite_use_address): Rewrite address uses in calls that were
14423 identified by find_address_like_use.
14425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14426 Alan Hayward <alan.hayward@arm.com>
14427 David Sherwood <david.sherwood@arm.com>
14429 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14431 * gimple-expr.h (is_gimple_addressable: Likewise.
14432 * gimple-expr.c (is_gimple_address): Likewise.
14433 * internal-fn.c (expand_call_mem_ref): New function.
14434 (expand_mask_load_optab_fn): Use it.
14435 (expand_mask_store_optab_fn): Likewise.
14437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14438 Alan Hayward <alan.hayward@arm.com>
14439 David Sherwood <david.sherwood@arm.com>
14441 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14442 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14443 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14444 (cond_umax@var{mode}): Document.
14445 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14446 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14447 (cond_umin_optab, cond_umax_optab): New optabs.
14448 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14449 (COND_IOR, COND_XOR): New internal functions.
14450 * internal-fn.h (get_conditional_internal_fn): Declare.
14451 * internal-fn.c (cond_binary_direct): New macro.
14452 (expand_cond_binary_optab_fn): Likewise.
14453 (direct_cond_binary_optab_supported_p): Likewise.
14454 (get_conditional_internal_fn): New function.
14455 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14456 Cope with reduction statements that are vectorized as calls rather
14458 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14459 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14460 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14461 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14462 (UNSPEC_COND_EOR): New unspecs.
14463 (optab): Add mappings for them.
14464 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14465 (sve_int_op, sve_fp_op): New int attributes.
14467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14468 Alan Hayward <alan.hayward@arm.com>
14469 David Sherwood <david.sherwood@arm.com>
14471 * optabs.def (while_ult_optab): New optab.
14472 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14473 * internal-fn.def (WHILE_ULT): New internal function.
14474 * internal-fn.h (direct_internal_fn_supported_p): New override
14475 that takes two types as argument.
14476 * internal-fn.c (while_direct): New macro.
14477 (expand_while_optab_fn): New function.
14478 (convert_optab_supported_p): Likewise.
14479 (direct_while_optab_supported_p): New macro.
14480 * wide-int.h (wi::udiv_ceil): New function.
14481 * tree-vectorizer.h (rgroup_masks): New structure.
14482 (vec_loop_masks): New typedef.
14483 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14484 and fully_masked_p.
14485 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14486 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14487 (vect_max_vf): New function.
14488 (slpeel_make_loop_iterate_ntimes): Delete.
14489 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14490 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14491 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14492 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14493 internal-fn.h, stor-layout.h and optabs-query.h.
14494 (vect_set_loop_mask): New function.
14495 (add_preheader_seq): Likewise.
14496 (add_header_seq): Likewise.
14497 (interleave_supported_p): Likewise.
14498 (vect_maybe_permute_loop_masks): Likewise.
14499 (vect_set_loop_masks_directly): Likewise.
14500 (vect_set_loop_condition_masked): Likewise.
14501 (vect_set_loop_condition_unmasked): New function, split out from
14502 slpeel_make_loop_iterate_ntimes.
14503 (slpeel_make_loop_iterate_ntimes): Rename to..
14504 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14505 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14506 (vect_do_peeling): Update call accordingly.
14507 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14509 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14510 mask_compare_type, can_fully_mask_p and fully_masked_p.
14511 (release_vec_loop_masks): New function.
14512 (_loop_vec_info): Use it to free the loop masks.
14513 (can_produce_all_loop_masks_p): New function.
14514 (vect_get_max_nscalars_per_iter): Likewise.
14515 (vect_verify_full_masking): Likewise.
14516 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14517 retries, and free the mask rgroups before retrying. Check loop-wide
14518 reasons for disallowing fully-masked loops. Make the final decision
14519 about whether use a fully-masked loop or not.
14520 (vect_estimate_min_profitable_iters): Do not assume that peeling
14521 for the number of iterations will be needed for fully-masked loops.
14522 (vectorizable_reduction): Disable fully-masked loops.
14523 (vectorizable_live_operation): Likewise.
14524 (vect_halve_mask_nunits): New function.
14525 (vect_double_mask_nunits): Likewise.
14526 (vect_record_loop_mask): Likewise.
14527 (vect_get_loop_mask): Likewise.
14528 (vect_transform_loop): Handle the case in which the final loop
14529 iteration might handle a partial vector. Call vect_set_loop_condition
14530 instead of slpeel_make_loop_iterate_ntimes.
14531 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14532 (check_load_store_masking): New function.
14533 (prepare_load_store_mask): Likewise.
14534 (vectorizable_store): Handle fully-masked loops.
14535 (vectorizable_load): Likewise.
14536 (supportable_widening_operation): Use vect_halve_mask_nunits for
14538 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14539 (vect_gen_while): New function.
14540 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14541 (aarch64_uqdec<mode>): New insn.
14543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14544 Alan Hayward <alan.hayward@arm.com>
14545 David Sherwood <david.sherwood@arm.com>
14547 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14548 (reduc_xor_scal_optab): New optabs.
14549 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14550 (reduc_xor_scal_@var{m}): Document.
14551 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14552 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14553 internal functions.
14554 * fold-const-call.c (fold_const_call): Handle them.
14555 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14556 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14557 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14558 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14559 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14560 (UNSPEC_XORV): New unspecs.
14561 (optab): Add entries for them.
14562 (BITWISEV): New int iterator.
14563 (bit_reduc_op): New int attributes.
14565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14566 Alan Hayward <alan.hayward@arm.com>
14567 David Sherwood <david.sherwood@arm.com>
14569 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14570 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14571 * optabs.def (vec_shl_insert_optab): New optab.
14572 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14573 (duplicate_and_interleave): Likewise.
14574 * tree-vect-loop.c: Include internal-fn.h.
14575 (neutral_op_for_slp_reduction): New function, split out from
14576 get_initial_defs_for_reduction.
14577 (get_initial_def_for_reduction): Handle option 2 for variable-length
14578 vectors by loading the neutral value into a vector and then shifting
14579 the initial value into element 0.
14580 (get_initial_defs_for_reduction): Replace the code argument with
14581 the neutral value calculated by neutral_op_for_slp_reduction.
14582 Use gimple_build_vector for constant-length vectors.
14583 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14584 but the first group_size elements have a neutral value.
14585 Use duplicate_and_interleave otherwise.
14586 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14587 Update call to get_initial_defs_for_reduction. Handle SLP
14588 reductions for variable-length vectors by creating one vector
14589 result for each scalar result, with the elements associated
14590 with other scalar results stubbed out with the neutral value.
14591 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14592 Require IFN_VEC_SHL_INSERT for double reductions on
14593 variable-length vectors, or SLP reductions that have
14594 a neutral value. Require can_duplicate_and_interleave_p
14595 support for variable-length unchained SLP reductions if there
14596 is no neutral value, such as for MIN/MAX reductions. Also require
14597 the number of vector elements to be a multiple of the number of
14598 SLP statements when doing variable-length unchained SLP reductions.
14599 Update call to vect_create_epilog_for_reduction.
14600 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14601 and remove initial values.
14602 (duplicate_and_interleave): Make public.
14603 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14604 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14607 Alan Hayward <alan.hayward@arm.com>
14608 David Sherwood <david.sherwood@arm.com>
14610 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14611 (can_duplicate_and_interleave_p): New function.
14612 (vect_get_and_check_slp_defs): Take the vector of statements
14613 rather than just the current one. Remove excess parentheses.
14614 Restriction rejectinon of vect_constant_def and vect_external_def
14615 for variable-length vectors to boolean types, or types for which
14616 can_duplicate_and_interleave_p is false.
14617 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14618 (duplicate_and_interleave): New function.
14619 (vect_get_constant_vectors): Use gimple_build_vector for
14620 constant-length vectors and suitable variable-length constant
14621 vectors. Use duplicate_and_interleave for other variable-length
14622 vectors. Don't defer the update when inserting new statements.
14624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14625 Alan Hayward <alan.hayward@arm.com>
14626 David Sherwood <david.sherwood@arm.com>
14628 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14629 min_profitable_iters doesn't go negative.
14631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14632 Alan Hayward <alan.hayward@arm.com>
14633 David Sherwood <david.sherwood@arm.com>
14635 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14636 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14637 * optabs.def (vec_mask_load_lanes_optab): New optab.
14638 (vec_mask_store_lanes_optab): Likewise.
14639 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14640 (MASK_STORE_LANES): Likewise.
14641 * internal-fn.c (mask_load_lanes_direct): New macro.
14642 (mask_store_lanes_direct): Likewise.
14643 (expand_mask_load_optab_fn): Handle masked operations.
14644 (expand_mask_load_lanes_optab_fn): New macro.
14645 (expand_mask_store_optab_fn): Handle masked operations.
14646 (expand_mask_store_lanes_optab_fn): New macro.
14647 (direct_mask_load_lanes_optab_supported_p): Likewise.
14648 (direct_mask_store_lanes_optab_supported_p): Likewise.
14649 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14651 (vect_load_lanes_supported): Likewise.
14652 * tree-vect-data-refs.c (strip_conversion): New function.
14653 (can_group_stmts_p): Likewise.
14654 (vect_analyze_data_ref_accesses): Use it instead of checking
14655 for a pair of assignments.
14656 (vect_store_lanes_supported): Take a masked_p parameter.
14657 (vect_load_lanes_supported): Likewise.
14658 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14659 vect_store_lanes_supported and vect_load_lanes_supported.
14660 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14661 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14662 parameter. Don't allow gaps for masked accesses.
14663 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14664 and vect_load_lanes_supported.
14665 (get_load_store_type): Take a masked_p parameter and update
14666 call to get_group_load_store_type.
14667 (vectorizable_store): Update call to get_load_store_type.
14668 Handle IFN_MASK_STORE_LANES.
14669 (vectorizable_load): Update call to get_load_store_type.
14670 Handle IFN_MASK_LOAD_LANES.
14672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14673 Alan Hayward <alan.hayward@arm.com>
14674 David Sherwood <david.sherwood@arm.com>
14676 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14678 * config/aarch64/aarch64-protos.h
14679 (aarch64_sve_struct_memory_operand_p): Declare.
14680 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14681 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14682 (VPRED, vpred): Handle SVE structure modes.
14683 * config/aarch64/constraints.md (Utx): New constraint.
14684 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14685 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14686 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14687 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14688 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14689 structure modes. Split into pieces after RA.
14690 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14691 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14693 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14694 SVE structure modes.
14695 (aarch64_classify_address): Likewise.
14696 (sizetochar): Move earlier in file.
14697 (aarch64_print_operand): Handle SVE register lists.
14698 (aarch64_array_mode): New function.
14699 (aarch64_sve_struct_memory_operand_p): Likewise.
14700 (TARGET_ARRAY_MODE): Redefine.
14702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14703 Alan Hayward <alan.hayward@arm.com>
14704 David Sherwood <david.sherwood@arm.com>
14706 * target.def (array_mode): New target hook.
14707 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14708 * doc/tm.texi: Regenerate.
14709 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14710 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14711 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14712 targetm.array_mode.
14713 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14717 Alan Hayward <alan.hayward@arm.com>
14718 David Sherwood <david.sherwood@arm.com>
14720 * fold-const.c (fold_binary_loc): Check the argument types
14721 rather than the result type when testing for a vector operation.
14723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14725 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14726 * doc/tm.texi: Regenerate.
14728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14729 Alan Hayward <alan.hayward@arm.com>
14730 David Sherwood <david.sherwood@arm.com>
14732 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14733 (sve): Document new AArch64 extension.
14734 * doc/md.texi (w): Extend the description of the AArch64
14735 constraint to include SVE vectors.
14736 (Upl, Upa): Document new AArch64 predicate constraints.
14737 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14739 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14740 (msve-vector-bits=): New option.
14741 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14742 SVE when these are disabled.
14743 (sve): New extension.
14744 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14745 modes. Adjust their number of units based on aarch64_sve_vg.
14746 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14747 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14748 aarch64_addr_query_type.
14749 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14750 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14751 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14752 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14753 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14754 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14755 (aarch64_simd_imm_zero_p): Delete.
14756 (aarch64_check_zero_based_sve_index_immediate): Declare.
14757 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14758 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14759 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14760 (aarch64_sve_float_mul_immediate_p): Likewise.
14761 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14762 rather than an rtx.
14763 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14764 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14765 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14766 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14767 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14768 (aarch64_regmode_natural_size): Likewise.
14769 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14770 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14772 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14773 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14774 for VG and the SVE predicate registers.
14775 (V_ALIASES): Add a "z"-prefixed alias.
14776 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14777 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14778 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14779 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14780 (REG_CLASS_NAMES): Add entries for them.
14781 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14782 and the predicate registers.
14783 (aarch64_sve_vg): Declare.
14784 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14785 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14786 (REGMODE_NATURAL_SIZE): Define.
14787 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14789 * config/aarch64/aarch64.c: Include cfgrtl.h.
14790 (simd_immediate_info): Add a constructor for series vectors,
14791 and an associated step field.
14792 (aarch64_sve_vg): New variable.
14793 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14794 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14795 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14796 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14797 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14798 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14799 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14800 (aarch64_get_mask_mode): New functions.
14801 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14802 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14803 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14804 predicate modes and predicate registers. Explicitly restrict
14805 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14806 to store a vector mode if it is recognized by
14807 aarch64_classify_vector_mode.
14808 (aarch64_regmode_natural_size): New function.
14809 (aarch64_hard_regno_caller_save_mode): Return the original mode
14811 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14812 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14813 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14814 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14816 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14817 does not overlap dest if the function is frame-related. Handle
14819 (aarch64_split_add_offset): New function.
14820 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14821 them aarch64_add_offset.
14822 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14823 and update call to aarch64_sub_sp.
14824 (aarch64_add_cfa_expression): New function.
14825 (aarch64_expand_prologue): Pass extra temporary registers to the
14826 functions above. Handle the case in which we need to emit new
14827 DW_CFA_expressions for registers that were originally saved
14828 relative to the stack pointer, but now have to be expressed
14829 relative to the frame pointer.
14830 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14832 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14833 IP0 and IP1 values for SVE frames.
14834 (aarch64_expand_vec_series): New function.
14835 (aarch64_expand_sve_widened_duplicate): Likewise.
14836 (aarch64_expand_sve_const_vector): Likewise.
14837 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14838 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14839 into the register, rather than emitting a SET directly.
14840 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14841 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14842 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14843 (offset_9bit_signed_scaled_p): New functions.
14844 (aarch64_replicate_bitmask_imm): New function.
14845 (aarch64_bitmask_imm): Use it.
14846 (aarch64_cannot_force_const_mem): Reject expressions involving
14847 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14848 (aarch64_classify_index): Handle SVE indices, by requiring
14849 a plain register index with a scale that matches the element size.
14850 (aarch64_classify_address): Handle SVE addresses. Assert that
14851 the mode of the address is VOIDmode or an integer mode.
14852 Update call to aarch64_classify_symbol.
14853 (aarch64_classify_symbolic_expression): Update call to
14854 aarch64_classify_symbol.
14855 (aarch64_const_vec_all_in_range_p): New function.
14856 (aarch64_print_vector_float_operand): Likewise.
14857 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14858 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14859 and the FP immediates 1.0 and 0.5.
14860 (aarch64_print_address_internal): Handle SVE addresses.
14861 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14862 (aarch64_regno_regclass): Handle predicate registers.
14863 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14865 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14866 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14867 (aarch64_convert_sve_vector_bits): New function.
14868 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14869 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14870 rather than an rtx.
14871 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14872 Handle SVE vector and predicate modes. Accept VL-based constants
14873 that need only one temporary register, and VL offsets that require
14874 no temporary registers.
14875 (aarch64_conditional_register_usage): Mark the predicate registers
14876 as fixed if SVE isn't available.
14877 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14878 Return true for SVE vector and predicate modes.
14879 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14880 rather than an unsigned int. Handle SVE modes.
14881 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14883 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14885 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14886 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14887 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14888 (aarch64_sve_float_mul_immediate_p): New functions.
14889 (aarch64_sve_valid_immediate): New function.
14890 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14891 Explicitly reject structure modes. Check for INDEX constants.
14892 Handle PTRUE and PFALSE constants.
14893 (aarch64_check_zero_based_sve_index_immediate): New function.
14894 (aarch64_simd_imm_zero_p): Delete.
14895 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14896 vector modes. Accept constants in the range of CNT[BHWD].
14897 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14898 ask for an Advanced SIMD mode.
14899 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14900 (aarch64_simd_vector_alignment): Handle SVE predicates.
14901 (aarch64_vectorize_preferred_vector_alignment): New function.
14902 (aarch64_simd_vector_alignment_reachable): Use it instead of
14904 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14905 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14907 (MAX_VECT_LEN): Delete.
14908 (expand_vec_perm_d): Add a vec_flags field.
14909 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14910 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14911 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14913 (aarch64_evpc_rev): Rename to...
14914 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14915 (aarch64_evpc_rev_global): New function.
14916 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14917 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14919 (aarch64_evpc_sve_tbl): New function.
14920 (aarch64_expand_vec_perm_const_1): Update after rename of
14921 aarch64_evpc_rev. Handle SVE permutes too, trying
14922 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14923 than aarch64_evpc_tbl.
14924 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14925 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14926 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14927 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14928 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14929 (aarch64_expand_sve_vcond): New functions.
14930 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14931 of aarch64_vector_mode_p.
14932 (aarch64_dwarf_poly_indeterminate_value): New function.
14933 (aarch64_compute_pressure_classes): Likewise.
14934 (aarch64_can_change_mode_class): Likewise.
14935 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14936 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14937 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14938 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14939 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14940 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14941 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14942 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14944 (Dn, Dl, Dr): Accept const as well as const_vector.
14945 (Dz): Likewise. Compare against CONST0_RTX.
14946 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14947 of "vector" where appropriate.
14948 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14949 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14950 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14951 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14952 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14953 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14954 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14955 (v_int_equiv): Extend to SVE modes.
14956 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14958 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14959 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14960 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14961 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14962 (SVE_COND_FP_CMP): New int iterators.
14963 (perm_hilo): Handle the new unpack unspecs.
14964 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14966 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14967 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14968 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14969 (aarch64_equality_operator, aarch64_constant_vector_operand)
14970 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14971 (aarch64_sve_nonimmediate_operand): Likewise.
14972 (aarch64_sve_general_operand): Likewise.
14973 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14974 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14975 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14976 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14977 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14978 (aarch64_sve_float_arith_immediate): Likewise.
14979 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14980 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14981 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14982 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14983 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14984 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14985 (aarch64_sve_float_arith_operand): Likewise.
14986 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14987 (aarch64_sve_float_mul_operand): Likewise.
14988 (aarch64_sve_vec_perm_operand): Likewise.
14989 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14990 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14991 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14992 as well as const_vector.
14993 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14994 in file. Use CONST0_RTX and CONSTM1_RTX.
14995 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14996 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14997 Use aarch64_simd_imm_zero.
14998 * config/aarch64/aarch64-sve.md: New file.
14999 * config/aarch64/aarch64.md: Include it.
15000 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15001 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15002 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15003 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15004 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15005 (sve): New attribute.
15006 (enabled): Disable instructions with the sve attribute unless
15008 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15009 aarch64_expand_mov_immediate.
15010 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15011 CNT[BHSD] immediates.
15012 (movti): Split CONST_POLY_INT moves into two halves.
15013 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15014 Split additions that need a temporary here if the destination
15015 is the stack pointer.
15016 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15017 (*add<mode>3_poly_1): New instruction.
15018 (set_clobber_cc): New expander.
15020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15022 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15023 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15024 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15025 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15026 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15027 Change innermode from fixed_mode_size to machine_mode.
15028 (simplify_subreg): Update call accordingly. Handle a constant-sized
15029 subreg of a variable-length CONST_VECTOR.
15031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15032 Alan Hayward <alan.hayward@arm.com>
15033 David Sherwood <david.sherwood@arm.com>
15035 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15036 (add_offset_to_base): New function, split out from...
15037 (create_mem_ref): ...here. When handling a scale other than 1,
15038 check first whether the address is valid without the offset.
15039 Add it into the base if so, leaving the index and scale as-is.
15041 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15044 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15045 fold_for_warn before checking if arg2 is INTEGER_CST.
15047 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15049 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15050 (store_multiple_operation): Delete.
15051 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15052 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15053 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15054 guarded by TARGET_STRING.
15055 (rs6000_output_load_multiple): Delete.
15056 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15057 OPTION_MASK_STRING / TARGET_STRING handling.
15058 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15059 (const rs6000_opt_masks) <"string">: Change mask to 0.
15060 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15061 (MASK_STRING): Delete.
15062 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15064 (load_multiple): Delete.
15071 (store_multiple): Delete.
15078 (movmemsi_8reg): Delete.
15079 (corresponding unnamed define_insn): Delete.
15080 (movmemsi_6reg): Delete.
15081 (corresponding unnamed define_insn): Delete.
15082 (movmemsi_4reg): Delete.
15083 (corresponding unnamed define_insn): Delete.
15084 (movmemsi_2reg): Delete.
15085 (corresponding unnamed define_insn): Delete.
15086 (movmemsi_1reg): Delete.
15087 (corresponding unnamed define_insn): Delete.
15088 * config/rs6000/rs6000.opt (mno-string): New.
15089 (mstring): Replace by deprecation warning stub.
15090 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15092 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15094 * regrename.c (regrename_do_replace): If replacing the same
15095 reg multiple times, try to reuse last created gen_raw_REG.
15098 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15099 main to workaround a bug in GDB.
15101 2018-01-12 Tom de Vries <tom@codesourcery.com>
15104 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15106 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15108 PR rtl-optimization/80481
15109 * ira-color.c (get_cap_member): New function.
15110 (allocnos_conflict_by_live_ranges_p): Use it.
15111 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15112 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15114 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15117 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15118 (*saddl_se_1): Ditto.
15119 (*ssubsi_1): Ditto.
15120 (*ssubl_se_1): Ditto.
15122 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15124 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15125 rather than wi::to_widest for DR_INITs.
15126 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15127 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15128 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15130 (vect_analyze_group_access_1): Note that here.
15132 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15134 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15135 polynomial type sizes.
15137 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15139 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15140 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15141 (gimple_add_tmp_var): Likewise.
15143 2018-01-12 Martin Liska <mliska@suse.cz>
15145 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15146 (gimple_alloc_sizes): Likewise.
15147 (dump_gimple_statistics): Use PRIu64 in printf format.
15148 * gimple.h: Change uint64_t to int.
15150 2018-01-12 Martin Liska <mliska@suse.cz>
15152 * tree-core.h: Use uint64_t instead of int.
15153 * tree.c (tree_node_counts): Likewise.
15154 (tree_node_sizes): Likewise.
15155 (dump_tree_statistics): Use PRIu64 in printf format.
15157 2018-01-12 Martin Liska <mliska@suse.cz>
15159 * Makefile.in: As qsort_chk is implemented in vec.c, add
15160 vec.o to linkage of gencfn-macros.
15161 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15162 passing the info to record_node_allocation_statistics.
15163 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15165 * ggc-common.c (struct ggc_usage): Add operator== and use
15166 it in operator< and compare function.
15167 * mem-stats.h (struct mem_usage): Likewise.
15168 * vec.c (struct vec_usage): Remove operator< and compare
15169 function. Can be simply inherited.
15171 2018-01-12 Martin Jambor <mjambor@suse.cz>
15174 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15175 * tree-ssa-math-opts.c: Include domwalk.h.
15176 (convert_mult_to_fma_1): New function.
15177 (fma_transformation_info): New type.
15178 (fma_deferring_state): Likewise.
15179 (cancel_fma_deferring): New function.
15180 (result_of_phi): Likewise.
15181 (last_fma_candidate_feeds_initial_phi): Likewise.
15182 (convert_mult_to_fma): Added deferring logic, split actual
15183 transformation to convert_mult_to_fma_1.
15184 (math_opts_dom_walker): New type.
15185 (math_opts_dom_walker::after_dom_children): New method, body moved
15186 here from pass_optimize_widening_mul::execute, added deferring logic
15188 (pass_optimize_widening_mul::execute): Moved most of code to
15189 math_opts_dom_walker::after_dom_children.
15190 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15191 * config/i386/i386.c (ix86_option_override_internal): Added
15192 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15194 2018-01-12 Richard Biener <rguenther@suse.de>
15197 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15198 inline instance vars.
15200 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15203 * config/rx/rx.c (rx_is_restricted_memory_address):
15204 Handle SUBREG case.
15206 2018-01-12 Richard Biener <rguenther@suse.de>
15208 PR tree-optimization/80846
15209 * target.def (split_reduction): New target hook.
15210 * targhooks.c (default_split_reduction): New function.
15211 * targhooks.h (default_split_reduction): Declare.
15212 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15213 target requests first reduce vectors by combining low and high
15215 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15216 (get_vectype_for_scalar_type_and_size): Export.
15217 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15218 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15219 * doc/tm.texi: Regenerate.
15220 * config/i386/i386.c (ix86_split_reduction): Implement
15221 TARGET_VECTORIZE_SPLIT_REDUCTION.
15223 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15226 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15227 in PIC mode except for TARGET_VXWORKS_RTP.
15228 * config/sparc/sparc.c: Include cfgrtl.h.
15229 (TARGET_INIT_PIC_REG): Define.
15230 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15231 (sparc_pic_register_p): New predicate.
15232 (sparc_legitimate_address_p): Use it.
15233 (sparc_legitimize_pic_address): Likewise.
15234 (sparc_delegitimize_address): Likewise.
15235 (sparc_mode_dependent_address_p): Likewise.
15236 (gen_load_pcrel_sym): Remove 4th parameter.
15237 (load_got_register): Adjust call to above. Remove obsolete stuff.
15238 (sparc_expand_prologue): Do not call load_got_register here.
15239 (sparc_flat_expand_prologue): Likewise.
15240 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15241 (sparc_use_pseudo_pic_reg): New function.
15242 (sparc_init_pic_reg): Likewise.
15243 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15244 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15246 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15248 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15249 Add item for branch_cost.
15251 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15253 PR rtl-optimization/83565
15254 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15255 not extend the result to a larger mode for rotate operations.
15256 (num_sign_bit_copies1): Likewise.
15258 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15261 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15263 Use values-Xc.o for -pedantic.
15264 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15266 2018-01-12 Martin Liska <mliska@suse.cz>
15269 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15271 (possible_polymorphic_call_targets): Use it.
15272 (ipa_devirt): Likewise.
15274 2018-01-12 Martin Liska <mliska@suse.cz>
15276 * profile-count.h (enum profile_quality): Use 0 as invalid
15277 enum value of profile_quality.
15279 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15281 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15282 -mext-string options.
15284 2018-01-12 Richard Biener <rguenther@suse.de>
15286 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15287 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15288 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15290 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15292 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15294 * configure.ac (--with-long-double-format): Add support for the
15295 configuration option to change the default long double format on
15297 * config.gcc (powerpc*-linux*-*): Likewise.
15298 * configure: Regenerate.
15299 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15300 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15301 used without modification.
15303 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15305 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15306 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15307 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15308 MISC_BUILTIN_SPEC_BARRIER.
15309 (rs6000_init_builtins): Likewise.
15310 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15312 (speculation_barrier): New define_insn.
15313 * doc/extend.texi: Document __builtin_speculation_barrier.
15315 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15318 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15319 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15320 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15322 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15323 integral modes instead of "ss" and "sd".
15324 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15325 vectors with 32-bit and 64-bit elements.
15326 (vecdupssescalarmodesuffix): New mode attribute.
15327 (vec_dup<mode>): Use it.
15329 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15332 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15333 frame if argument is passed on stack.
15335 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15338 * ree.c (combine_reaching_defs): Optimize also
15339 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15340 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15342 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15344 PR middle-end/83189
15345 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15347 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15349 PR middle-end/83718
15350 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15351 after they are computed.
15353 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15355 PR tree-optimization/83695
15356 * gimple-loop-linterchange.cc
15357 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15358 reset cached scev information after interchange.
15359 (pass_linterchange::execute): Remove call to scev_reset_htab.
15361 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15363 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15364 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15365 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15366 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15367 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15368 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15369 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15370 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15371 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15372 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15373 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15374 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15375 (V_lane_reg): Likewise.
15376 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15378 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15379 (vfmal_lane_low<mode>_intrinsic,
15380 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15381 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15382 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15383 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15384 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15385 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15387 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15389 * config/arm/arm-cpus.in (fp16fml): New feature.
15390 (ALL_SIMD): Add fp16fml.
15391 (armv8.2-a): Add fp16fml as an option.
15392 (armv8.3-a): Likewise.
15393 (armv8.4-a): Add fp16fml as part of fp16.
15394 * config/arm/arm.h (TARGET_FP16FML): Define.
15395 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15397 * config/arm/arm-modes.def (V2HF): Define.
15398 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15399 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15400 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15401 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15402 vfmsl_low, vfmsl_high): New set of builtins.
15403 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15404 (vfml_op): New code attribute.
15405 (VFMLHALVES): New int iterator.
15406 (VFML, VFMLSEL): New mode attributes.
15407 (V_reg): Define mapping for V2HF.
15408 (V_hi, V_lo): New mode attributes.
15409 (VF_constraint): Likewise.
15410 (vfml_half, vfml_half_selector): New int attributes.
15411 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15413 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15414 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15416 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15417 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15418 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15419 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15421 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15422 Document new effective target and option set.
15424 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15426 * config/arm/arm-cpus.in (armv8_4): New feature.
15427 (ARMv8_4a): New fgroup.
15428 (armv8.4-a): New arch.
15429 * config/arm/arm-tables.opt: Regenerate.
15430 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15431 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15432 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15433 Add matching rules for -march=armv8.4-a and extensions.
15434 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15436 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15439 * config/rx/rx.md (BW): New mode attribute.
15440 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15442 2018-01-11 Richard Biener <rguenther@suse.de>
15444 PR tree-optimization/83435
15445 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15446 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15447 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15449 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15450 Alan Hayward <alan.hayward@arm.com>
15451 David Sherwood <david.sherwood@arm.com>
15453 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15455 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15456 (aarch64_print_address_internal): Use it to check for a zero offset.
15458 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15459 Alan Hayward <alan.hayward@arm.com>
15460 David Sherwood <david.sherwood@arm.com>
15462 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15463 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15464 Return a poly_int64 rather than a HOST_WIDE_INT.
15465 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15466 rather than a HOST_WIDE_INT.
15467 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15468 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15469 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15470 final_offset from HOST_WIDE_INT to poly_int64.
15471 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15472 to_constant when getting the number of units in an Advanced SIMD
15474 (aarch64_builtin_vectorized_function): Check for a constant number
15476 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15478 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15479 attribute instead of GET_MODE_NUNITS.
15480 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15481 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15482 GET_MODE_SIZE for fixed-size registers.
15483 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15484 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15485 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15486 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15487 (aarch64_print_operand, aarch64_print_address_internal)
15488 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15489 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15490 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15491 Handle polynomial GET_MODE_SIZE.
15492 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15493 wider than SImode without modification.
15494 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15495 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15496 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15497 passing and returning SVE modes.
15498 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15499 rather than GEN_INT.
15500 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15501 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15502 (aarch64_allocate_and_probe_stack_space): Likewise.
15503 (aarch64_layout_frame): Cope with polynomial offsets.
15504 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15505 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15506 polynomial offsets.
15507 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15508 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15509 poly_int64 rather than a HOST_WIDE_INT.
15510 (aarch64_get_separate_components, aarch64_process_components)
15511 (aarch64_expand_prologue, aarch64_expand_epilogue)
15512 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15513 (aarch64_anchor_offset): New function, split out from...
15514 (aarch64_legitimize_address): ...here.
15515 (aarch64_builtin_vectorization_cost): Handle polynomial
15516 TYPE_VECTOR_SUBPARTS.
15517 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15519 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15520 number of elements from the PARALLEL rather than the mode.
15521 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15522 rather than GET_MODE_BITSIZE.
15523 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15524 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15525 (aarch64_expand_vec_perm_const_1): Handle polynomial
15526 d->perm.length () and d->perm elements.
15527 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15528 Apply to_constant to d->perm elements.
15529 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15530 polynomial CONST_VECTOR_NUNITS.
15531 (aarch64_move_pointer): Take amount as a poly_int64 rather
15533 (aarch64_progress_pointer): Avoid temporary variable.
15534 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15535 the mode attribute instead of GET_MODE.
15537 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15538 Alan Hayward <alan.hayward@arm.com>
15539 David Sherwood <david.sherwood@arm.com>
15541 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15542 x exists before using it.
15543 (aarch64_add_constant_internal): Rename to...
15544 (aarch64_add_offset_1): ...this. Replace regnum with separate
15545 src and dest rtxes. Handle the case in which they're different,
15546 including when the offset is zero. Replace scratchreg with an rtx.
15547 Use 2 additions if there is no spare register into which we can
15548 move a 16-bit constant.
15549 (aarch64_add_constant): Delete.
15550 (aarch64_add_offset): Replace reg with separate src and dest
15551 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15552 Use aarch64_add_offset_1.
15553 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15554 an rtx rather than an int. Take the delta as a poly_int64
15555 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15556 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15557 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15558 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15559 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15560 and aarch64_add_sp.
15561 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15562 aarch64_add_constant.
15564 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15566 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15567 Use scalar_float_mode.
15569 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15571 * config/aarch64/aarch64-simd.md
15572 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15573 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15574 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15575 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15576 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15577 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15578 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15579 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15580 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15581 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15583 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15586 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15587 targ_options->x_arm_arch_string is non NULL.
15589 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15591 * config/aarch64/aarch64.h
15592 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15594 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15597 * expmed.c (emit_store_flag_force): Swap if const op0
15598 and change VOIDmode to mode of op0.
15600 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15602 PR rtl-optimization/83761
15603 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15604 than bytes to mode_for_size.
15606 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15608 PR middle-end/83189
15609 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15610 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15613 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15615 PR middle-end/83575
15616 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15617 when in layout mode.
15618 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15619 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15622 2018-01-10 Michael Collison <michael.collison@arm.com>
15624 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15625 * config/aarch64/aarch64-option-extension.def: Add
15626 AARCH64_OPT_EXTENSION of 'fp16fml'.
15627 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15628 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15629 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15630 * config/aarch64/constraints.md (Ui7): New constraint.
15631 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15632 (VFMLA_SEL_W): Ditto.
15635 (VFMLA16_LOW): New int iterator.
15636 (VFMLA16_HIGH): Ditto.
15637 (UNSPEC_FMLAL): New unspec.
15638 (UNSPEC_FMLSL): Ditto.
15639 (UNSPEC_FMLAL2): Ditto.
15640 (UNSPEC_FMLSL2): Ditto.
15641 (f16mac): New code attribute.
15642 * config/aarch64/aarch64-simd-builtins.def
15643 (aarch64_fmlal_lowv2sf): Ditto.
15644 (aarch64_fmlsl_lowv2sf): Ditto.
15645 (aarch64_fmlalq_lowv4sf): Ditto.
15646 (aarch64_fmlslq_lowv4sf): Ditto.
15647 (aarch64_fmlal_highv2sf): Ditto.
15648 (aarch64_fmlsl_highv2sf): Ditto.
15649 (aarch64_fmlalq_highv4sf): Ditto.
15650 (aarch64_fmlslq_highv4sf): Ditto.
15651 (aarch64_fmlal_lane_lowv2sf): Ditto.
15652 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15653 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15654 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15655 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15656 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15657 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15658 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15659 (aarch64_fmlal_lane_highv2sf): Ditto.
15660 (aarch64_fmlsl_lane_highv2sf): Ditto.
15661 (aarch64_fmlal_laneq_highv2sf): Ditto.
15662 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15663 (aarch64_fmlalq_lane_highv4sf): Ditto.
15664 (aarch64_fmlsl_lane_highv4sf): Ditto.
15665 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15666 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15667 * config/aarch64/aarch64-simd.md:
15668 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15669 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15670 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15671 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15672 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15673 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15674 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15675 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15676 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15677 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15678 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15679 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15680 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15681 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15682 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15683 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15684 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15685 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15686 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15687 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15688 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15689 (vfmlsl_low_u32): Ditto.
15690 (vfmlalq_low_u32): Ditto.
15691 (vfmlslq_low_u32): Ditto.
15692 (vfmlal_high_u32): Ditto.
15693 (vfmlsl_high_u32): Ditto.
15694 (vfmlalq_high_u32): Ditto.
15695 (vfmlslq_high_u32): Ditto.
15696 (vfmlal_lane_low_u32): Ditto.
15697 (vfmlsl_lane_low_u32): Ditto.
15698 (vfmlal_laneq_low_u32): Ditto.
15699 (vfmlsl_laneq_low_u32): Ditto.
15700 (vfmlalq_lane_low_u32): Ditto.
15701 (vfmlslq_lane_low_u32): Ditto.
15702 (vfmlalq_laneq_low_u32): Ditto.
15703 (vfmlslq_laneq_low_u32): Ditto.
15704 (vfmlal_lane_high_u32): Ditto.
15705 (vfmlsl_lane_high_u32): Ditto.
15706 (vfmlal_laneq_high_u32): Ditto.
15707 (vfmlsl_laneq_high_u32): Ditto.
15708 (vfmlalq_lane_high_u32): Ditto.
15709 (vfmlslq_lane_high_u32): Ditto.
15710 (vfmlalq_laneq_high_u32): Ditto.
15711 (vfmlslq_laneq_high_u32): Ditto.
15712 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15713 (AARCH64_FL_FOR_ARCH8_4): New.
15714 (AARCH64_ISA_F16FML): New ISA flag.
15715 (TARGET_F16FML): New feature flag for fp16fml.
15716 (doc/invoke.texi): Document new fp16fml option.
15718 2018-01-10 Michael Collison <michael.collison@arm.com>
15720 * config/aarch64/aarch64-builtins.c:
15721 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15722 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15723 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15724 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15725 (AARCH64_ISA_SHA3): New ISA flag.
15726 (TARGET_SHA3): New feature flag for sha3.
15727 * config/aarch64/iterators.md (sha512_op): New int attribute.
15728 (CRYPTO_SHA512): New int iterator.
15729 (UNSPEC_SHA512H): New unspec.
15730 (UNSPEC_SHA512H2): Ditto.
15731 (UNSPEC_SHA512SU0): Ditto.
15732 (UNSPEC_SHA512SU1): Ditto.
15733 * config/aarch64/aarch64-simd-builtins.def
15734 (aarch64_crypto_sha512hqv2di): New builtin.
15735 (aarch64_crypto_sha512h2qv2di): Ditto.
15736 (aarch64_crypto_sha512su0qv2di): Ditto.
15737 (aarch64_crypto_sha512su1qv2di): Ditto.
15738 (aarch64_eor3qv8hi): Ditto.
15739 (aarch64_rax1qv2di): Ditto.
15740 (aarch64_xarqv2di): Ditto.
15741 (aarch64_bcaxqv8hi): Ditto.
15742 * config/aarch64/aarch64-simd.md:
15743 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15744 (aarch64_crypto_sha512su0qv2di): Ditto.
15745 (aarch64_crypto_sha512su1qv2di): Ditto.
15746 (aarch64_eor3qv8hi): Ditto.
15747 (aarch64_rax1qv2di): Ditto.
15748 (aarch64_xarqv2di): Ditto.
15749 (aarch64_bcaxqv8hi): Ditto.
15750 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15751 (vsha512h2q_u64): Ditto.
15752 (vsha512su0q_u64): Ditto.
15753 (vsha512su1q_u64): Ditto.
15754 (veor3q_u16): Ditto.
15755 (vrax1q_u64): Ditto.
15756 (vxarq_u64): Ditto.
15757 (vbcaxq_u16): Ditto.
15758 * config/arm/types.md (crypto_sha512): New type attribute.
15759 (crypto_sha3): Ditto.
15760 (doc/invoke.texi): Document new sha3 option.
15762 2018-01-10 Michael Collison <michael.collison@arm.com>
15764 * config/aarch64/aarch64-builtins.c:
15765 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15766 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15767 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15768 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15769 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15770 (AARCH64_ISA_SM4): New ISA flag.
15771 (TARGET_SM4): New feature flag for sm4.
15772 * config/aarch64/aarch64-simd-builtins.def
15773 (aarch64_sm3ss1qv4si): Ditto.
15774 (aarch64_sm3tt1aq4si): Ditto.
15775 (aarch64_sm3tt1bq4si): Ditto.
15776 (aarch64_sm3tt2aq4si): Ditto.
15777 (aarch64_sm3tt2bq4si): Ditto.
15778 (aarch64_sm3partw1qv4si): Ditto.
15779 (aarch64_sm3partw2qv4si): Ditto.
15780 (aarch64_sm4eqv4si): Ditto.
15781 (aarch64_sm4ekeyqv4si): Ditto.
15782 * config/aarch64/aarch64-simd.md:
15783 (aarch64_sm3ss1qv4si): Ditto.
15784 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15785 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15786 (aarch64_sm4eqv4si): Ditto.
15787 (aarch64_sm4ekeyqv4si): Ditto.
15788 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15789 (sm3part_op): Ditto.
15790 (CRYPTO_SM3TT): Ditto.
15791 (CRYPTO_SM3PART): Ditto.
15792 (UNSPEC_SM3SS1): New unspec.
15793 (UNSPEC_SM3TT1A): Ditto.
15794 (UNSPEC_SM3TT1B): Ditto.
15795 (UNSPEC_SM3TT2A): Ditto.
15796 (UNSPEC_SM3TT2B): Ditto.
15797 (UNSPEC_SM3PARTW1): Ditto.
15798 (UNSPEC_SM3PARTW2): Ditto.
15799 (UNSPEC_SM4E): Ditto.
15800 (UNSPEC_SM4EKEY): Ditto.
15801 * config/aarch64/constraints.md (Ui2): New constraint.
15802 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15803 * config/arm/types.md (crypto_sm3): New type attribute.
15804 (crypto_sm4): Ditto.
15805 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15806 (vsm3tt1aq_u32): Ditto.
15807 (vsm3tt1bq_u32): Ditto.
15808 (vsm3tt2aq_u32): Ditto.
15809 (vsm3tt2bq_u32): Ditto.
15810 (vsm3partw1q_u32): Ditto.
15811 (vsm3partw2q_u32): Ditto.
15812 (vsm4eq_u32): Ditto.
15813 (vsm4ekeyq_u32): Ditto.
15814 (doc/invoke.texi): Document new sm4 option.
15816 2018-01-10 Michael Collison <michael.collison@arm.com>
15818 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15819 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15820 (AARCH64_FL_FOR_ARCH8_4): New.
15821 (AARCH64_FL_V8_4): New flag.
15822 (doc/invoke.texi): Document new armv8.4-a option.
15824 2018-01-10 Michael Collison <michael.collison@arm.com>
15826 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15827 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15828 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15829 * config/aarch64/aarch64-option-extension.def: Add
15830 AARCH64_OPT_EXTENSION of 'sha2'.
15831 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15832 (crypto): Disable sha2 and aes if crypto disabled.
15833 (crypto): Enable aes and sha2 if enabled.
15834 (simd): Disable sha2 and aes if simd disabled.
15835 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15837 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15838 (TARGET_SHA2): New feature flag for sha2.
15839 (TARGET_AES): New feature flag for aes.
15840 * config/aarch64/aarch64-simd.md:
15841 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15842 conditional on TARGET_AES.
15843 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15844 (aarch64_crypto_sha1hsi): Make pattern conditional
15846 (aarch64_crypto_sha1hv4si): Ditto.
15847 (aarch64_be_crypto_sha1hv4si): Ditto.
15848 (aarch64_crypto_sha1su1v4si): Ditto.
15849 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15850 (aarch64_crypto_sha1su0v4si): Ditto.
15851 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15852 (aarch64_crypto_sha256su0v4si): Ditto.
15853 (aarch64_crypto_sha256su1v4si): Ditto.
15854 (doc/invoke.texi): Document new aes and sha2 options.
15856 2018-01-10 Martin Sebor <msebor@redhat.com>
15858 PR tree-optimization/83781
15859 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15862 2018-01-11 Martin Sebor <msebor@gmail.com>
15863 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15865 PR tree-optimization/83501
15866 PR tree-optimization/81703
15868 * tree-ssa-strlen.c (get_string_cst): Rename...
15869 (get_string_len): ...to this. Handle global constants.
15870 (handle_char_store): Adjust.
15872 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15873 Jim Wilson <jimw@sifive.com>
15875 * config/riscv/riscv-protos.h (riscv_output_return): New.
15876 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15877 (riscv_attribute_table, riscv_output_return),
15878 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15879 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15880 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15882 (riscv_expand_prologue): Add early return for naked function.
15883 (riscv_expand_epilogue): Likewise.
15884 (riscv_function_ok_for_sibcall): Return false for naked function.
15885 (riscv_set_current_function): New.
15886 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15887 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15888 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15889 * doc/extend.texi (RISC-V Function Attributes): New.
15891 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15893 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15894 check for 128-bit long double before checking TCmode.
15895 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15896 128-bit long doubles before checking TFmode or TCmode.
15897 (FLOAT128_IBM_P): Likewise.
15899 2018-01-10 Martin Sebor <msebor@redhat.com>
15901 PR tree-optimization/83671
15902 * builtins.c (c_strlen): Unconditionally return zero for the empty
15904 Use -Warray-bounds for warnings.
15905 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15906 for non-constant array indices with COMPONENT_REF, arrays of
15907 arrays, and pointers to arrays.
15908 (gimple_fold_builtin_strlen): Determine and set length range for
15909 non-constant character arrays.
15911 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15913 PR middle-end/81897
15914 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15917 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15919 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15921 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15924 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15925 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15926 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15927 indexed_or_indirect_operand predicate.
15928 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15929 (*vsx_le_perm_load_v8hi): Likewise.
15930 (*vsx_le_perm_load_v16qi): Likewise.
15931 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15932 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15933 (*vsx_le_perm_store_v8hi): Likewise.
15934 (*vsx_le_perm_store_v16qi): Likewise.
15935 (eight unnamed splitters): Likewise.
15937 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15939 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15940 * config/rs6000/emmintrin.h: Likewise.
15941 * config/rs6000/mmintrin.h: Likewise.
15942 * config/rs6000/xmmintrin.h: Likewise.
15944 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15947 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15949 * tree.c (tree_nop_conversion): Return true for location wrapper
15951 (maybe_wrap_with_location): New function.
15952 (selftest::check_strip_nops): New function.
15953 (selftest::test_location_wrappers): New function.
15954 (selftest::tree_c_tests): Call it.
15955 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15956 (maybe_wrap_with_location): New decl.
15957 (EXPR_LOCATION_WRAPPER_P): New macro.
15958 (location_wrapper_p): New inline function.
15959 (tree_strip_any_location_wrapper): New inline function.
15961 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15964 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15965 stack_realign_offset for the largest alignment of stack slot
15967 (ix86_find_max_used_stack_alignment): New function.
15968 (ix86_finalize_stack_frame_flags): Use it. Set
15969 max_used_stack_alignment if we don't realign stack.
15970 * config/i386/i386.h (machine_function): Add
15971 max_used_stack_alignment.
15973 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15975 * config/arm/arm.opt (-mbranch-cost): New option.
15976 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15979 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15982 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15983 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15985 2018-01-10 Richard Biener <rguenther@suse.de>
15988 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15989 early out so it also covers the case where we have a non-NULL
15992 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15994 PR tree-optimization/83753
15995 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15996 for non-strided grouped accesses if the number of elements is 1.
15998 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16001 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16002 * i386.h (TARGET_USE_GATHER): Define.
16003 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16005 2018-01-10 Martin Liska <mliska@suse.cz>
16008 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16009 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16011 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16012 CLEANUP_NO_PARTITIONING is not set.
16014 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16016 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16017 for vectors, as a partial revert of r254296.
16018 * rtl.h (const_vec_p): Delete.
16019 (const_vec_duplicate_p): Don't test for vector CONSTs.
16020 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16021 * expmed.c (make_tree): Likewise.
16024 * common.md (E, F): Use CONSTANT_P instead of checking for
16026 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16027 checking for CONST_VECTOR.
16029 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16031 PR middle-end/83575
16032 * predict.c (force_edge_cold): Handle in more sane way edges
16033 with no prediction.
16035 2018-01-09 Carl Love <cel@us.ibm.com>
16037 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16039 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16040 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16041 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16042 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16043 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16044 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16045 * config/rs6000/rs6000-protos.h: Add extern defition for
16046 rs6000_generate_float2_double_code.
16047 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16049 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16050 (float2_v2df): Add define_expand.
16052 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16055 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16056 op_mode in the force_to_mode call.
16058 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16060 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16061 instead of checking each element individually.
16062 (aarch64_evpc_uzp): Likewise.
16063 (aarch64_evpc_zip): Likewise.
16064 (aarch64_evpc_ext): Likewise.
16065 (aarch64_evpc_rev): Likewise.
16066 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16067 instead of checking each element individually. Return true without
16069 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16070 whether all selected elements come from the same input, instead of
16071 checking each element individually. Remove calls to gen_rtx_REG,
16072 start_sequence and end_sequence and instead assert that no rtl is
16075 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16077 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16078 order of HIGH and CONST checks.
16080 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16082 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16083 if the destination isn't an SSA_NAME.
16085 2018-01-09 Richard Biener <rguenther@suse.de>
16087 PR tree-optimization/83668
16088 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16090 (canonicalize_loop_form): ... here, renamed from ...
16091 (canonicalize_loop_closed_ssa_form): ... this and amended to
16092 swap successor edges for loop exit blocks to make us use
16093 the RPO order we need for initial schedule generation.
16095 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16097 PR tree-optimization/64811
16098 * match.pd: When optimizing comparisons with Inf, avoid
16099 introducing or losing exceptions from comparisons with NaN.
16101 2018-01-09 Martin Liska <mliska@suse.cz>
16104 * asan.c (shadow_mem_size): Add gcc_assert.
16106 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16108 Don't save registers in main().
16111 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16112 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16113 * config/avr/avr.c (avr_set_current_function): Don't error if
16114 naked, OS_task or OS_main are specified at the same time.
16115 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16117 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16119 * common/config/avr/avr-common.c (avr_option_optimization_table):
16120 Switch on -mmain-is-OS_task for optimizing compilations.
16122 2018-01-09 Richard Biener <rguenther@suse.de>
16124 PR tree-optimization/83572
16125 * graphite.c: Include cfganal.h.
16126 (graphite_transform_loops): Connect infinite loops to exit
16127 and remove fake edges at the end.
16129 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16131 * ipa-inline.c (edge_badness): Revert accidental checkin.
16133 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16136 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16137 symbols; not inline clones.
16139 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16142 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16143 hard registers. Formatting fixes.
16145 PR preprocessor/83722
16146 * gcc.c (try_generate_repro): Pass
16147 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16148 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16151 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16152 Kito Cheng <kito.cheng@gmail.com>
16154 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16155 (riscv_leaf_function_p): Delete.
16156 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16158 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16160 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16162 (do_ifelse): New function.
16163 (do_isel): New function.
16164 (do_sub3): New function.
16165 (do_add3): New function.
16166 (do_load_mask_compare): New function.
16167 (do_overlap_load_compare): New function.
16168 (expand_compare_loop): New function.
16169 (expand_block_compare): Call expand_compare_loop() when appropriate.
16170 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16171 option description.
16172 (-mblock-compare-inline-loop-limit): New option.
16174 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16177 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16178 Reverse order of second and third operands in first alternative.
16179 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16180 of first and second elements in UNSPEC_VPERMR vector.
16181 (altivec_expand_vec_perm_le): Likewise.
16183 2018-01-08 Jeff Law <law@redhat.com>
16185 PR rtl-optimizatin/81308
16186 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16187 (process_switch): If group_case_labels makes a change, then set
16189 (pass_convert_switch::execute): If a switch is converted, then
16190 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16192 PR rtl-optimization/81308
16193 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16196 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16198 PR target/83663 - Revert r255946
16199 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16200 generation for cases where splatting a value is not useful.
16201 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16202 across a vec_duplicate and a paradoxical subreg forming a vector
16203 mode to a vec_concat.
16205 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16207 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16208 -march=armv8.3-a variants.
16209 * config/arm/t-multilib: Likewise.
16210 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16212 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16214 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16216 (cceq_ior_compare_complement): Give it a name so I can use it, and
16217 change boolean_or_operator predicate to boolean_operator so it can
16218 be used to generate a crand.
16219 (eqne): New code iterator.
16220 (bd/bd_neg): New code_attrs.
16221 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16222 a single define_insn.
16223 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16224 decrement (bdnzt/bdnzf/bdzt/bdzf).
16225 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16226 with the new names of the branch decrement patterns, and added the
16227 names of the branch decrement conditional patterns.
16229 2018-01-08 Richard Biener <rguenther@suse.de>
16231 PR tree-optimization/83563
16232 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16235 2018-01-08 Richard Biener <rguenther@suse.de>
16237 PR middle-end/83713
16238 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16240 2018-01-08 Richard Biener <rguenther@suse.de>
16242 PR tree-optimization/83685
16243 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16244 references to abnormals.
16246 2018-01-08 Richard Biener <rguenther@suse.de>
16249 * dwarf2out.c (output_indirect_strings): Handle empty
16250 skeleton_debug_str_hash.
16251 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16253 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16255 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16256 (emit_store_direct): Likewise.
16257 (arc_trampoline_adjust_address): Likewise.
16258 (arc_asm_trampoline_template): New function.
16259 (arc_initialize_trampoline): Use asm_trampoline_template.
16260 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16261 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16262 * config/arc/arc.md (flush_icache): Delete pattern.
16264 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16266 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16267 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16270 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16273 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16274 by not USED_FOR_TARGET.
16275 (make_pass_resolve_sw_modes): Likewise.
16277 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16279 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16282 2018-01-08 Richard Biener <rguenther@suse.de>
16284 PR middle-end/83580
16285 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16287 2018-01-08 Richard Biener <rguenther@suse.de>
16289 PR middle-end/83517
16290 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16292 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16294 PR middle-end/81897
16295 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16296 basic blocks with a small number of successors.
16297 (convert_control_dep_chain_into_preds): Improve handling of
16299 (dump_predicates): Split apart into...
16300 (dump_pred_chain): ...here...
16301 (dump_pred_info): ...and here.
16302 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16303 (can_chain_union_be_invalidated_p): Improve check for invalidation
16305 (uninit_uses_cannot_happen): Avoid unnecessary if
16306 convert_control_dep_chain_into_preds yielded nothing.
16308 2018-01-06 Martin Sebor <msebor@redhat.com>
16310 PR tree-optimization/83640
16311 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16312 subtracting negative offset from size.
16313 (builtin_access::overlap): Adjust offset bounds of the access to fall
16314 within the size of the object if possible.
16316 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16318 PR rtl-optimization/83699
16319 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16320 extract_bit_field_as_subreg to cases in which the extracted
16321 value is also a vector.
16323 * lra-constraints.c (process_alt_operands): Test for the equivalence
16324 substitutions when detecting a possible reload cycle.
16326 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16329 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16330 by default if flag_selective_schedling{,2}. Formatting fixes.
16332 PR rtl-optimization/83682
16333 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16334 if it has non-VECTOR_MODE element mode.
16335 (vec_duplicate_p): Likewise.
16337 PR middle-end/83694
16338 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16339 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16341 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16344 * config/i386/i386-builtin.def
16345 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16346 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16347 Require also OPTION_MASK_ISA_AVX512F in addition to
16348 OPTION_MASK_ISA_GFNI.
16349 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16350 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16351 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16352 to OPTION_MASK_ISA_GFNI.
16353 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16354 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16355 OPTION_MASK_ISA_AVX512BW.
16356 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16357 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16358 addition to OPTION_MASK_ISA_GFNI.
16359 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16360 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16361 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16362 to OPTION_MASK_ISA_GFNI.
16363 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16364 a requirement for all ISAs rather than any of them with a few
16366 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16368 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16369 bitmasks to be enabled with 3 exceptions, instead of requiring any
16370 enabled ISA with lots of exceptions.
16371 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16372 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16373 Change avx512bw in isa attribute to avx512f.
16374 * config/i386/sgxintrin.h: Add license boilerplate.
16375 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16376 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16377 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16378 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16380 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16381 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16382 temporarily sse2 rather than sse if not enabled already.
16385 * config/i386/sse.md (VI248_VLBW): Rename to ...
16386 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16387 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16388 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16389 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16390 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16391 mode iterator instead of VI248_VLBW.
16393 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16395 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16396 (record_modified): Skip clobbers; add debug output.
16397 (param_change_prob): Use sreal frequencies.
16399 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16401 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16402 punt for user-aligned variables.
16404 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16406 * tree-chrec.c (chrec_contains_symbols): Return true for
16409 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16412 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16413 of (x|y) == x for BICS pattern.
16415 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16417 PR tree-optimization/83605
16418 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16419 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16422 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16424 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16425 * config/epiphany/rtems.h: New file.
16427 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16428 Uros Bizjak <ubizjak@gmail.com>
16431 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16432 QIreg_operand instead of register_operand predicate.
16433 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16434 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16435 comments instead of -fmitigate[-_]rop.
16437 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16440 * cgraphunit.c (symbol_table::compile): Switch to text_section
16441 before calling assembly_start debug hook.
16442 * run-rtl-passes.c (run_rtl_passes): Likewise.
16445 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16447 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16448 range_int_cst_p rather than !symbolic_range_p before calling
16449 extract_range_from_multiplicative_op_1.
16451 2018-01-04 Jeff Law <law@redhat.com>
16453 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16454 redundant test in assertion.
16456 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16458 * doc/rtl.texi: Document machine_mode wrapper classes.
16460 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16462 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16463 using tree_to_uhwi.
16465 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16467 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16468 the VEC_PERM_EXPR fold to fail.
16470 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16473 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16474 to switched_sections.
16476 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16479 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16480 test for d.testing.
16482 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16485 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16486 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16488 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16491 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16492 is BLKmode and bitpos not zero or mode change is needed.
16494 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16497 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16500 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16503 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16504 instead of MULT rtx. Update all corresponding splitters.
16505 (*saddl_se): Ditto.
16506 (*ssub<modesuffix>): Ditto.
16507 (*ssubl_se): Ditto.
16508 (*cmp_sadd_di): Update split patterns.
16509 (*cmp_sadd_si): Ditto.
16510 (*cmp_sadd_sidi): Ditto.
16511 (*cmp_ssub_di): Ditto.
16512 (*cmp_ssub_si): Ditto.
16513 (*cmp_ssub_sidi): Ditto.
16514 * config/alpha/predicates.md (const23_operand): New predicate.
16515 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16516 Look for ASHIFT, not MULT inner operand.
16517 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16519 2018-01-04 Martin Liska <mliska@suse.cz>
16521 PR gcov-profile/83669
16522 * gcov.c (output_intermediate_file): Add version to intermediate
16524 * doc/gcov.texi: Document new field 'version' in intermediate
16525 file format. Fix location of '-k' option of gcov command.
16527 2018-01-04 Martin Liska <mliska@suse.cz>
16530 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16532 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16534 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16536 2018-01-03 Martin Sebor <msebor@redhat.com>
16538 PR tree-optimization/83655
16539 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16540 checking calls with invalid arguments.
16542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16544 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16545 (vectorizable_mask_load_store): Delete.
16546 (vectorizable_call): Return false for masked loads and stores.
16547 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16548 instead of gimple_assign_rhs1.
16549 (vectorizable_load): Handle IFN_MASK_LOAD.
16550 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16554 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16556 (vectorizable_mask_load_store): ...here.
16557 (vectorizable_load): ...and here.
16559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16561 * tree-vect-stmts.c (vect_build_all_ones_mask)
16562 (vect_build_zero_merge_argument): New functions, split out from...
16563 (vectorizable_load): ...here.
16565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16567 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16569 (vectorizable_mask_load_store): ...here.
16570 (vectorizable_store): ...and here.
16572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16574 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16576 (vectorizable_mask_load_store): ...here.
16578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16580 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16581 (vect_model_store_cost): Take a vec_load_store_type instead of a
16583 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16584 (vect_model_store_cost): Take a vec_load_store_type instead of a
16586 (vectorizable_mask_load_store): Update accordingly.
16587 (vectorizable_store): Likewise.
16588 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16592 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16593 IFN_MASK_LOAD calls here rather than...
16594 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
16601 fall back to extract_bit_field_as_subreg if vec_extract
16604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16605 Alan Hayward <alan.hayward@arm.com>
16606 David Sherwood <david.sherwood@arm.com>
16608 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16609 they are variable or constant sized.
16610 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16611 slots for constant-sized data.
16613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16614 Alan Hayward <alan.hayward@arm.com>
16615 David Sherwood <david.sherwood@arm.com>
16617 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16618 handling COND_EXPRs with boolean comparisons, try to find a better
16619 basis for the mask type than the boolean itself.
16621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16623 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16624 is calculated and how it can be overridden.
16625 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16626 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16628 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16632 Alan Hayward <alan.hayward@arm.com>
16633 David Sherwood <david.sherwood@arm.com>
16635 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16636 Remove the mode argument.
16637 (aarch64_simd_valid_immediate): Remove the mode and inverse
16639 * config/aarch64/iterators.md (bitsize): New iterator.
16640 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16641 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16642 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16643 aarch64_simd_valid_immediate.
16644 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16645 (aarch64_reg_or_bic_imm): Likewise.
16646 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16647 with an insn_type enum and msl with a modifier_type enum.
16648 Replace element_width with a scalar_mode. Change the shift
16649 to unsigned int. Add constructors for scalar_float_mode and
16650 scalar_int_mode elements.
16651 (aarch64_vect_float_const_representable_p): Delete.
16652 (aarch64_can_const_movi_rtx_p)
16653 (aarch64_simd_scalar_immediate_valid_for_move)
16654 (aarch64_simd_make_constant): Update call to
16655 aarch64_simd_valid_immediate.
16656 (aarch64_advsimd_valid_immediate_hs): New function.
16657 (aarch64_advsimd_valid_immediate): Likewise.
16658 (aarch64_simd_valid_immediate): Remove mode and inverse
16659 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16660 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16661 and aarch64_float_const_representable_p on the result.
16662 (aarch64_output_simd_mov_immediate): Remove mode argument.
16663 Update call to aarch64_simd_valid_immediate and use of
16664 simd_immediate_info.
16665 (aarch64_output_scalar_simd_mov_immediate): Update call
16668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16669 Alan Hayward <alan.hayward@arm.com>
16670 David Sherwood <david.sherwood@arm.com>
16672 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16673 (mode_nunits): Likewise CONST_MODE_NUNITS.
16674 * machmode.def (ADJUST_NUNITS): Document.
16675 * genmodes.c (mode_data::need_nunits_adj): New field.
16676 (blank_mode): Update accordingly.
16677 (adj_nunits): New variable.
16678 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16680 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16681 listed in adj_nunits.
16682 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16683 listed in adj_nunits. Don't emit case statements for such modes.
16684 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16685 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16686 nothing if adj_nunits is nonnull.
16687 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16688 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16689 (emit_mode_fbit): Update use of print_maybe_const_decl.
16690 (emit_move_size): Likewise. Treat the array as non-const
16692 (emit_mode_adjustments): Handle adj_nunits.
16694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16696 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16697 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16698 (VECTOR_MODES): Use it.
16699 (make_vector_modes): Take the prefix as an argument.
16701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16702 Alan Hayward <alan.hayward@arm.com>
16703 David Sherwood <david.sherwood@arm.com>
16705 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16706 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16707 for MODE_VECTOR_BOOL.
16708 * machmode.def (VECTOR_BOOL_MODE): Document.
16709 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16710 (make_vector_bool_mode): New function.
16711 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16713 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16714 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16716 * stor-layout.c (int_mode_for_mode): Likewise.
16717 * tree.c (build_vector_type_for_mode): Likewise.
16718 * varasm.c (output_constant_pool_2): Likewise.
16719 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16720 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16721 for MODE_VECTOR_BOOL.
16722 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16723 of mode class checks.
16724 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16725 instead of a list of mode class checks.
16726 (expand_vector_scalar_condition): Likewise.
16727 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16730 Alan Hayward <alan.hayward@arm.com>
16731 David Sherwood <david.sherwood@arm.com>
16733 * machmode.h (mode_size): Change from unsigned short to
16735 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16736 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16737 or if measurement_type is not polynomial.
16738 (fixed_size_mode::includes_p): Check for constant-sized modes.
16739 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16740 return a poly_uint16 rather than an unsigned short.
16741 (emit_mode_size): Change the type of mode_size from unsigned short
16742 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16743 (emit_mode_adjustments): Cope with polynomial vector sizes.
16744 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16746 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16748 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16749 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16750 * caller-save.c (setup_save_areas): Likewise.
16751 (replace_reg_with_saved_mem): Likewise.
16752 * calls.c (emit_library_call_value_1): Likewise.
16753 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16754 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16755 (gen_lowpart_for_combine): Likewise.
16756 * convert.c (convert_to_integer_1): Likewise.
16757 * cse.c (equiv_constant, cse_insn): Likewise.
16758 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16759 (cselib_subst_to_values): Likewise.
16760 * dce.c (word_dce_process_block): Likewise.
16761 * df-problems.c (df_word_lr_mark_ref): Likewise.
16762 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16763 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16764 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16765 (rtl_for_decl_location): Likewise.
16766 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16767 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16768 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16769 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16770 (expand_expr_real_1): Likewise.
16771 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16772 (pad_below): Likewise.
16773 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16774 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16775 * ira.c (get_subreg_tracking_sizes): Likewise.
16776 * ira-build.c (ira_create_allocno_objects): Likewise.
16777 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16778 (ira_sort_regnos_for_alter_reg): Likewise.
16779 * ira-costs.c (record_operand_costs): Likewise.
16780 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16781 (resolve_simple_move): Likewise.
16782 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16783 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16784 (lra_constraints): Likewise.
16785 (CONST_POOL_OK_P): Reject variable-sized modes.
16786 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16787 (add_pseudo_to_slot, lra_spill): Likewise.
16788 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16789 * optabs-query.c (get_best_extraction_insn): Likewise.
16790 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16791 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16792 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16793 * recog.c (offsettable_address_addr_space_p): Likewise.
16794 * regcprop.c (maybe_mode_change): Likewise.
16795 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16796 * regrename.c (build_def_use): Likewise.
16797 * regstat.c (dump_reg_info): Likewise.
16798 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16799 (find_reloads, find_reloads_subreg_address): Likewise.
16800 * reload1.c (eliminate_regs_1): Likewise.
16801 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16802 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16803 (simplify_binary_operation_1, simplify_subreg): Likewise.
16804 * targhooks.c (default_function_arg_padding): Likewise.
16805 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16806 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16807 (verify_gimple_assign_ternary): Likewise.
16808 * tree-inline.c (estimate_move_cost): Likewise.
16809 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16810 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16811 (get_address_cost_ainc): Likewise.
16812 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16813 (vect_supportable_dr_alignment): Likewise.
16814 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16815 (vectorizable_reduction): Likewise.
16816 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16817 (vectorizable_operation, vectorizable_load): Likewise.
16818 * tree.c (build_same_sized_truth_vector_type): Likewise.
16819 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16820 * var-tracking.c (emit_note_insn_var_location): Likewise.
16821 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16822 (ADDR_VEC_ALIGN): Likewise.
16824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16825 Alan Hayward <alan.hayward@arm.com>
16826 David Sherwood <david.sherwood@arm.com>
16828 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16830 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16831 or if measurement_type is polynomial.
16832 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16833 * combine.c (make_extraction): Likewise.
16834 * dse.c (find_shift_sequence): Likewise.
16835 * dwarf2out.c (mem_loc_descriptor): Likewise.
16836 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16837 (extract_bit_field, extract_low_bits): Likewise.
16838 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16839 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16840 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16841 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16842 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16843 * reload.c (find_reloads): Likewise.
16844 * reload1.c (alter_reg): Likewise.
16845 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16846 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16847 * tree-if-conv.c (predicate_mem_writes): Likewise.
16848 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16849 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16850 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16851 * valtrack.c (dead_debug_insert_temp): Likewise.
16852 * varasm.c (mergeable_constant_section): Likewise.
16853 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16856 Alan Hayward <alan.hayward@arm.com>
16857 David Sherwood <david.sherwood@arm.com>
16859 * expr.c (expand_assignment): Cope with polynomial mode sizes
16860 when assigning to a CONCAT.
16862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16863 Alan Hayward <alan.hayward@arm.com>
16864 David Sherwood <david.sherwood@arm.com>
16866 * machmode.h (mode_precision): Change from unsigned short to
16868 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16870 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16871 or if measurement_type is not polynomial.
16872 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16873 in which the mode is already known to be a scalar_int_mode.
16874 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16875 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16877 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16878 for GET_MODE_PRECISION.
16879 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16880 for GET_MODE_PRECISION.
16881 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16883 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16884 (expand_field_assignment, make_extraction): Likewise.
16885 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16886 (get_last_value): Likewise.
16887 * convert.c (convert_to_integer_1): Likewise.
16888 * cse.c (cse_insn): Likewise.
16889 * expr.c (expand_expr_real_1): Likewise.
16890 * lra-constraints.c (simplify_operand_subreg): Likewise.
16891 * optabs-query.c (can_atomic_load_p): Likewise.
16892 * optabs.c (expand_atomic_load): Likewise.
16893 (expand_atomic_store): Likewise.
16894 * ree.c (combine_reaching_defs): Likewise.
16895 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16896 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16897 * tree.h (type_has_mode_precision_p): Likewise.
16898 * ubsan.c (instrument_si_overflow): Likewise.
16900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16901 Alan Hayward <alan.hayward@arm.com>
16902 David Sherwood <david.sherwood@arm.com>
16904 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16905 polynomial numbers of units.
16906 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16907 (valid_vector_subparts_p): New function.
16908 (build_vector_type): Remove temporary shim and take the number
16909 of units as a poly_uint64 rather than an int.
16910 (build_opaque_vector_type): Take the number of units as a
16911 poly_uint64 rather than an int.
16912 * tree.c (build_vector_from_ctor): Handle polynomial
16913 TYPE_VECTOR_SUBPARTS.
16914 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16915 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16916 (build_vector_from_val): If the number of units is variable,
16917 use build_vec_duplicate_cst for constant operands and
16918 VEC_DUPLICATE_EXPR otherwise.
16919 (make_vector_type): Remove temporary is_constant ().
16920 (build_vector_type, build_opaque_vector_type): Take the number of
16921 units as a poly_uint64 rather than an int.
16922 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16924 * cfgexpand.c (expand_debug_expr): Likewise.
16925 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16926 (store_constructor, expand_expr_real_1): Likewise.
16927 (const_scalar_mask_from_tree): Likewise.
16928 * fold-const-call.c (fold_const_reduction): Likewise.
16929 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16930 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16931 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16932 (fold_relational_const): Likewise.
16933 (native_interpret_vector): Likewise. Change the size from an
16934 int to an unsigned int.
16935 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16936 TYPE_VECTOR_SUBPARTS.
16937 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16938 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16939 duplicating a non-constant operand into a variable-length vector.
16940 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16941 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16942 * ipa-icf.c (sem_variable::equals): Likewise.
16943 * match.pd: Likewise.
16944 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16945 * print-tree.c (print_node): Likewise.
16946 * stor-layout.c (layout_type): Likewise.
16947 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16948 * tree-cfg.c (verify_gimple_comparison): Likewise.
16949 (verify_gimple_assign_binary): Likewise.
16950 (verify_gimple_assign_ternary): Likewise.
16951 (verify_gimple_assign_single): Likewise.
16952 * tree-pretty-print.c (dump_generic_node): Likewise.
16953 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16954 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16955 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16956 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16957 (vect_shift_permute_load_chain): Likewise.
16958 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16959 (expand_vector_condition, optimize_vector_constructor): Likewise.
16960 (lower_vec_perm, get_compute_type): Likewise.
16961 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16962 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16963 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16964 (vect_recog_mask_conversion_pattern): Likewise.
16965 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16966 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16967 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16968 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16969 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16970 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16971 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16972 (supportable_widening_operation): Likewise.
16973 (supportable_narrowing_operation): Likewise.
16974 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16976 * varasm.c (output_constant): Likewise.
16978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16979 Alan Hayward <alan.hayward@arm.com>
16980 David Sherwood <david.sherwood@arm.com>
16982 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16983 so that both the length == 3 and length != 3 cases set up their
16984 own permute vectors. Add comments explaining why we know the
16985 number of elements is constant.
16986 (vect_permute_load_chain): Likewise.
16988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16989 Alan Hayward <alan.hayward@arm.com>
16990 David Sherwood <david.sherwood@arm.com>
16992 * machmode.h (mode_nunits): Change from unsigned char to
16994 (ONLY_FIXED_SIZE_MODES): New macro.
16995 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16996 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16997 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16999 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17000 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17001 or if measurement_type is not polynomial.
17002 * genmodes.c (ZERO_COEFFS): New macro.
17003 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17005 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17006 Use ZERO_COEFFS when emitting initializers.
17007 * data-streamer.h (bp_pack_poly_value): New function.
17008 (bp_unpack_poly_value): Likewise.
17009 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17010 for GET_MODE_NUNITS.
17011 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17012 for GET_MODE_NUNITS.
17013 * tree.c (make_vector_type): Remove temporary shim and make
17014 the real function take the number of units as a poly_uint64
17015 rather than an int.
17016 (build_vector_type_for_mode): Handle polynomial nunits.
17017 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17018 * emit-rtl.c (const_vec_series_p_1): Likewise.
17019 (gen_rtx_CONST_VECTOR): Likewise.
17020 * fold-const.c (test_vec_duplicate_folding): Likewise.
17021 * genrecog.c (validate_pattern): Likewise.
17022 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17023 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17024 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17025 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17026 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17027 * rtlanal.c (subreg_get_info): Likewise.
17028 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17029 (vect_grouped_load_supported): Likewise.
17030 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17031 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17032 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17033 (simplify_const_unary_operation, simplify_binary_operation_1)
17034 (simplify_const_binary_operation, simplify_ternary_operation)
17035 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17036 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17037 instead of CONST_VECTOR_NUNITS.
17038 * varasm.c (output_constant_pool_2): Likewise.
17039 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17040 explicit-encoded elements in the XVEC for variable-length vectors.
17042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17044 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17047 Alan Hayward <alan.hayward@arm.com>
17048 David Sherwood <david.sherwood@arm.com>
17050 * coretypes.h (fixed_size_mode): Declare.
17051 (fixed_size_mode_pod): New typedef.
17052 * builtins.h (target_builtins::x_apply_args_mode)
17053 (target_builtins::x_apply_result_mode): Change type to
17054 fixed_size_mode_pod.
17055 * builtins.c (apply_args_size, apply_result_size, result_vector)
17056 (expand_builtin_apply_args_1, expand_builtin_apply)
17057 (expand_builtin_return): Update accordingly.
17059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17061 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17062 * cselib.c (cselib_hash_rtx): Likewise.
17063 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17064 CONST_VECTOR encoding.
17066 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17067 Jeff Law <law@redhat.com>
17070 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17071 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17072 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17073 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17076 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17077 explicitly probe *sp in a noreturn function if there were any callee
17078 register saves or frame pointer is needed.
17080 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17083 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17084 BLKmode for ternary, binary or unary expressions.
17087 * var-tracking.c (delete_vta_debug_insn): New inline function.
17088 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17089 insns from get_insns () to NULL instead of each bb separately.
17090 Use delete_vta_debug_insn. No longer static.
17091 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17092 delete_vta_debug_insns callers.
17093 * rtl.h (delete_vta_debug_insns): Declare.
17094 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17095 instead of variable_tracking_main.
17097 2018-01-03 Martin Sebor <msebor@redhat.com>
17099 PR tree-optimization/83603
17100 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17101 arguments past the endof the argument list in functions declared
17102 without a prototype.
17103 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17104 Avoid checking when arguments are null.
17106 2018-01-03 Martin Sebor <msebor@redhat.com>
17109 * doc/extend.texi (attribute const): Fix a typo.
17110 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17111 issuing -Wsuggest-attribute for void functions.
17113 2018-01-03 Martin Sebor <msebor@redhat.com>
17115 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17116 offset_int::from instead of wide_int::to_shwi.
17117 (maybe_diag_overlap): Remove assertion.
17118 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17119 * gimple-ssa-sprintf.c (format_directive): Same.
17120 (parse_directive): Same.
17121 (sprintf_dom_walker::compute_format_length): Same.
17122 (try_substitute_return_value): Same.
17124 2018-01-03 Jeff Law <law@redhat.com>
17126 PR middle-end/83654
17127 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17128 non-constant residual for zero at runtime and avoid probing in
17129 that case. Reorganize code for trailing problem to mirror handling
17132 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17134 PR tree-optimization/83501
17135 * tree-ssa-strlen.c (get_string_cst): New.
17136 (handle_char_store): Call get_string_cst.
17138 2018-01-03 Martin Liska <mliska@suse.cz>
17140 PR tree-optimization/83593
17141 * tree-ssa-strlen.c: Include tree-cfg.h.
17142 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17143 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17144 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17146 (strlen_dom_walker::before_dom_children): Call
17147 gimple_purge_dead_eh_edges. Dump tranformation with details
17149 (strlen_dom_walker::before_dom_children): Update call by adding
17150 new argument cleanup_eh.
17151 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17153 2018-01-03 Martin Liska <mliska@suse.cz>
17156 * cif-code.def (VARIADIC_THUNK): New enum value.
17157 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17160 2018-01-03 Jan Beulich <jbeulich@suse.com>
17162 * sse.md (mov<mode>_internal): Tighten condition for when to use
17163 vmovdqu<ssescalarsize> for TI and OI modes.
17165 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17167 Update copyright years.
17169 2018-01-03 Martin Liska <mliska@suse.cz>
17172 * ipa-visibility.c (function_and_variable_visibility): Skip
17173 functions with noipa attribure.
17175 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17177 * gcc.c (process_command): Update copyright notice dates.
17178 * gcov-dump.c (print_version): Ditto.
17179 * gcov.c (print_version): Ditto.
17180 * gcov-tool.c (print_version): Ditto.
17181 * gengtype.c (create_file): Ditto.
17182 * doc/cpp.texi: Bump @copying's copyright year.
17183 * doc/cppinternals.texi: Ditto.
17184 * doc/gcc.texi: Ditto.
17185 * doc/gccint.texi: Ditto.
17186 * doc/gcov.texi: Ditto.
17187 * doc/install.texi: Ditto.
17188 * doc/invoke.texi: Ditto.
17190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17192 * vector-builder.h (vector_builder::m_full_nelts): Change from
17193 unsigned int to poly_uint64.
17194 (vector_builder::full_nelts): Update prototype accordingly.
17195 (vector_builder::new_vector): Likewise.
17196 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17197 (vector_builder::operator ==): Likewise.
17198 (vector_builder::finalize): Likewise.
17199 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17200 Take the number of elements as a poly_uint64 rather than an
17202 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17203 from unsigned int to poly_uint64.
17204 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17205 (vec_perm_indices::new_vector): Likewise.
17206 (vec_perm_indices::length): Likewise.
17207 (vec_perm_indices::nelts_per_input): Likewise.
17208 (vec_perm_indices::input_nelts): Likewise.
17209 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17210 number of elements per input as a poly_uint64 rather than an
17211 unsigned int. Use the original encoding for variable-length
17212 vectors, rather than clamping each individual element.
17213 For the second and subsequent elements in each pattern,
17214 clamp the step and base before clamping their sum.
17215 (vec_perm_indices::series_p): Handle polynomial element counts.
17216 (vec_perm_indices::all_in_range_p): Likewise.
17217 (vec_perm_indices_to_tree): Likewise.
17218 (vec_perm_indices_to_rtx): Likewise.
17219 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17220 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17221 (tree_vector_builder::new_binary_operation): Handle polynomial
17222 element counts. Return false if we need to know the number
17223 of elements at compile time.
17224 * fold-const.c (fold_vec_perm): Punt if the number of elements
17225 isn't known at compile time.
17227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17229 * vec-perm-indices.h (vec_perm_builder): Change element type
17230 from HOST_WIDE_INT to poly_int64.
17231 (vec_perm_indices::element_type): Update accordingly.
17232 (vec_perm_indices::clamp): Handle polynomial element_types.
17233 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17234 (vec_perm_indices::all_in_range_p): Likewise.
17235 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17237 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17238 polynomial vec_perm_indices element types.
17239 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17240 * fold-const.c (fold_vec_perm): Likewise.
17241 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17242 * tree-vect-generic.c (lower_vec_perm): Likewise.
17243 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17244 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17245 element type to HOST_WIDE_INT.
17247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17248 Alan Hayward <alan.hayward@arm.com>
17249 David Sherwood <david.sherwood@arm.com>
17251 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17252 rather than an int. Use plus_constant.
17253 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17254 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17257 Alan Hayward <alan.hayward@arm.com>
17258 David Sherwood <david.sherwood@arm.com>
17260 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17261 a HOST_WIDE_INT to a poly_int64.
17263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17264 Alan Hayward <alan.hayward@arm.com>
17265 David Sherwood <david.sherwood@arm.com>
17267 * calls.c (load_register_parameters): Cope with polynomial
17268 mode sizes. Require a constant size for BLKmode parameters
17269 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17270 forces a parameter to be padded at the lsb end in order to
17271 fill a complete number of words, require the parameter size
17272 to be ordered wrt UNITS_PER_WORD.
17274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17275 Alan Hayward <alan.hayward@arm.com>
17276 David Sherwood <david.sherwood@arm.com>
17278 * reload1.c (spill_stack_slot_width): Change element type
17279 from unsigned int to poly_uint64_pod.
17280 (alter_reg): Treat mode sizes as polynomial.
17282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17283 Alan Hayward <alan.hayward@arm.com>
17284 David Sherwood <david.sherwood@arm.com>
17286 * reload.c (complex_word_subreg_p): New function.
17287 (reload_inner_reg_of_subreg, push_reload): Use it.
17289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17290 Alan Hayward <alan.hayward@arm.com>
17291 David Sherwood <david.sherwood@arm.com>
17293 * lra-constraints.c (process_alt_operands): Reject matched
17294 operands whose sizes aren't ordered.
17295 (match_reload): Refer to this check here.
17297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17298 Alan Hayward <alan.hayward@arm.com>
17299 David Sherwood <david.sherwood@arm.com>
17301 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17302 that the mode size is in the set {1, 2, 4, 8, 16}.
17304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17305 Alan Hayward <alan.hayward@arm.com>
17306 David Sherwood <david.sherwood@arm.com>
17308 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17309 Use plus_constant instead of gen_rtx_PLUS.
17311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17312 Alan Hayward <alan.hayward@arm.com>
17313 David Sherwood <david.sherwood@arm.com>
17315 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17316 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17317 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17318 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17319 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17320 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17321 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17322 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17323 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17324 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17326 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17327 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17328 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17329 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17330 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17331 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17332 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17333 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17334 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17335 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17337 * expr.c (emit_move_resolve_push): Treat the input and result
17338 of PUSH_ROUNDING as a poly_int64.
17339 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17340 (emit_push_insn): Likewise.
17341 * lra-eliminations.c (mark_not_eliminable): Likewise.
17342 * recog.c (push_operand): Likewise.
17343 * reload1.c (elimination_effects): Likewise.
17344 * rtlanal.c (nonzero_bits1): Likewise.
17345 * calls.c (store_one_arg): Likewise. Require the padding to be
17346 known at compile time.
17348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17349 Alan Hayward <alan.hayward@arm.com>
17350 David Sherwood <david.sherwood@arm.com>
17352 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17353 Use plus_constant instead of gen_rtx_PLUS.
17355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17356 Alan Hayward <alan.hayward@arm.com>
17357 David Sherwood <david.sherwood@arm.com>
17359 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17360 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17367 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17368 via stack temporaries. Treat the mode size as polynomial too.
17370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17371 Alan Hayward <alan.hayward@arm.com>
17372 David Sherwood <david.sherwood@arm.com>
17374 * expr.c (expand_expr_real_2): When handling conversions involving
17375 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17376 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17377 as a poly_uint64 too.
17379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17380 Alan Hayward <alan.hayward@arm.com>
17381 David Sherwood <david.sherwood@arm.com>
17383 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17386 Alan Hayward <alan.hayward@arm.com>
17387 David Sherwood <david.sherwood@arm.com>
17389 * combine.c (can_change_dest_mode): Handle polynomial
17390 REGMODE_NATURAL_SIZE.
17391 * expmed.c (store_bit_field_1): Likewise.
17392 * expr.c (store_constructor): Likewise.
17393 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17394 and polynomial REGMODE_NATURAL_SIZE.
17395 (gen_lowpart_common): Likewise.
17396 * reginfo.c (record_subregs_of_mode): Likewise.
17397 * rtlanal.c (read_modify_subreg_p): Likewise.
17399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17400 Alan Hayward <alan.hayward@arm.com>
17401 David Sherwood <david.sherwood@arm.com>
17403 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17404 numbers of elements.
17406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17407 Alan Hayward <alan.hayward@arm.com>
17408 David Sherwood <david.sherwood@arm.com>
17410 * match.pd: Cope with polynomial numbers of vector elements.
17412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17413 Alan Hayward <alan.hayward@arm.com>
17414 David Sherwood <david.sherwood@arm.com>
17416 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17417 in a POINTER_PLUS_EXPR.
17419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17420 Alan Hayward <alan.hayward@arm.com>
17421 David Sherwood <david.sherwood@arm.com>
17423 * omp-simd-clone.c (simd_clone_subparts): New function.
17424 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17425 (ipa_simd_modify_function_body): Likewise.
17427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17428 Alan Hayward <alan.hayward@arm.com>
17429 David Sherwood <david.sherwood@arm.com>
17431 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17432 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17433 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17434 (expand_vector_condition, vector_element): Likewise.
17435 (subparts_gt): New function.
17436 (get_compute_type): Use subparts_gt.
17437 (count_type_subparts): Delete.
17438 (expand_vector_operations_1): Use subparts_gt instead of
17439 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
17446 (vect_compile_time_alias): ...this new function. Do the calculation
17447 on poly_ints rather than trees.
17448 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
17456 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
17463 constant and extern definitions for variable-length vectors.
17464 (vect_get_constant_vectors): Note that the number of units
17465 is known to be constant.
17467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17468 Alan Hayward <alan.hayward@arm.com>
17469 David Sherwood <david.sherwood@arm.com>
17471 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17472 of units as polynomial. Choose between WIDE and NARROW based
17475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17476 Alan Hayward <alan.hayward@arm.com>
17477 David Sherwood <david.sherwood@arm.com>
17479 * tree-vect-stmts.c (simd_clone_subparts): New function.
17480 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17483 Alan Hayward <alan.hayward@arm.com>
17484 David Sherwood <david.sherwood@arm.com>
17486 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17487 vectors as polynomial. Use build_index_vector for
17488 IFN_GOMP_SIMD_LANE.
17490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17491 Alan Hayward <alan.hayward@arm.com>
17492 David Sherwood <david.sherwood@arm.com>
17494 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17495 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17496 for variable-length vectors.
17497 (vectorizable_mask_load_store): Treat the number of units as
17498 polynomial, asserting that it is constant if the condition has
17499 already been enforced.
17500 (vectorizable_store, vectorizable_load): Likewise.
17502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17503 Alan Hayward <alan.hayward@arm.com>
17504 David Sherwood <david.sherwood@arm.com>
17506 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17507 of units as polynomial. Punt if we can't tell at compile time
17508 which vector contains the final result.
17510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17511 Alan Hayward <alan.hayward@arm.com>
17512 David Sherwood <david.sherwood@arm.com>
17514 * tree-vect-loop.c (vectorizable_induction): Treat the number
17515 of units as polynomial. Punt on SLP inductions. Use an integer
17516 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17517 cast of such a series for variable-length floating-point
17520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17521 Alan Hayward <alan.hayward@arm.com>
17522 David Sherwood <david.sherwood@arm.com>
17524 * tree.h (build_index_vector): Declare.
17525 * tree.c (build_index_vector): New function.
17526 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17527 of units as polynomial, forcibly converting it to a constant if
17528 vectorizable_reduction has already enforced the condition.
17529 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17530 to create a {1,2,3,...} vector.
17531 (vectorizable_reduction): Treat the number of units as polynomial.
17532 Choose vectype_in based on the largest scalar element size rather
17533 than the smallest number of units. Enforce the restrictions
17536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17537 Alan Hayward <alan.hayward@arm.com>
17538 David Sherwood <david.sherwood@arm.com>
17540 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17541 number of units as polynomial.
17543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17544 Alan Hayward <alan.hayward@arm.com>
17545 David Sherwood <david.sherwood@arm.com>
17547 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17548 * target.def (autovectorize_vector_sizes): Return the vector sizes
17549 by pointer, using vector_sizes rather than a bitmask.
17550 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17551 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17552 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17554 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17555 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17556 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17557 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17558 * omp-general.c (omp_max_vf): Likewise.
17559 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17560 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17561 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17562 * tree-vect-slp.c (vect_slp_bb): Likewise.
17563 * doc/tm.texi: Regenerate.
17564 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17566 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17567 the vector size as a poly_uint64 rather than an unsigned int.
17568 (current_vector_size): Change from an unsigned int to a poly_uint64.
17569 (get_vectype_for_scalar_type): Update accordingly.
17570 * tree.h (build_truth_vector_type): Take the size and number of
17571 units as a poly_uint64 rather than an unsigned int.
17572 (build_vector_type): Add a temporary overload that takes
17573 the number of units as a poly_uint64 rather than an unsigned int.
17574 * tree.c (make_vector_type): Likewise.
17575 (build_truth_vector_type): Take the number of units as a poly_uint64
17576 rather than an unsigned int.
17578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17579 Alan Hayward <alan.hayward@arm.com>
17580 David Sherwood <david.sherwood@arm.com>
17582 * target.def (get_mask_mode): Take the number of units and length
17583 as poly_uint64s rather than unsigned ints.
17584 * targhooks.h (default_get_mask_mode): Update accordingly.
17585 * targhooks.c (default_get_mask_mode): Likewise.
17586 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17587 * doc/tm.texi: Regenerate.
17589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17590 Alan Hayward <alan.hayward@arm.com>
17591 David Sherwood <david.sherwood@arm.com>
17593 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17594 * omp-general.c (omp_max_vf): Likewise.
17595 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17596 (expand_omp_simd): Handle polynomial safelen.
17597 * omp-low.c (omplow_simd_context): Add a default constructor.
17598 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17599 (lower_rec_simd_input_clauses): Update accordingly.
17600 (lower_rec_input_clauses): Likewise.
17602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17603 Alan Hayward <alan.hayward@arm.com>
17604 David Sherwood <david.sherwood@arm.com>
17606 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17607 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17608 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17609 (vect_analyze_slp_cost): Likewise.
17610 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17611 (vect_model_load_cost): Likewise.
17613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17614 Alan Hayward <alan.hayward@arm.com>
17615 David Sherwood <david.sherwood@arm.com>
17617 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17618 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17619 from an unsigned int * to a poly_uint64_pod *.
17620 (calculate_unrolling_factor): New function.
17621 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17624 Alan Hayward <alan.hayward@arm.com>
17625 David Sherwood <david.sherwood@arm.com>
17627 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17628 from an unsigned int to a poly_uint64.
17629 (_loop_vec_info::slp_unrolling_factor): Likewise.
17630 (_loop_vec_info::vectorization_factor): Change from an int
17632 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17633 (vect_get_num_vectors): New function.
17634 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17635 (vect_get_num_copies): Use vect_get_num_vectors.
17636 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17637 to an unsigned int *.
17638 (vect_analyze_data_refs): Change min_vf from an int * to a
17640 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17641 than an unsigned HOST_WIDE_INT.
17642 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17643 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17644 to an unsigned int *.
17645 (vect_analyze_data_ref_dependences): Likewise.
17646 (vect_compute_data_ref_alignment): Handle polynomial vf.
17647 (vect_enhance_data_refs_alignment): Likewise.
17648 (vect_prune_runtime_alias_test_list): Likewise.
17649 (vect_shift_permute_load_chain): Likewise.
17650 (vect_supportable_dr_alignment): Likewise.
17651 (dependence_distance_ge_vf): Take the vectorization factor as a
17652 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17653 (vect_analyze_data_refs): Change min_vf from an int * to a
17655 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17656 vfm1 as a poly_uint64 rather than an int. Make the same change
17657 for the returned bound_scalar.
17658 (vect_gen_vector_loop_niters): Handle polynomial vf.
17659 (vect_do_peeling): Likewise. Update call to
17660 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17661 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17663 * tree-vect-loop.c (vect_determine_vectorization_factor)
17664 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17665 (vect_get_known_peeling_cost): Likewise.
17666 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17667 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17668 (vect_transform_loop): Likewise. Use the lowest possible VF when
17669 updating the upper bounds of the loop.
17670 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17671 rather than an int.
17672 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17673 polynomial unroll factors.
17674 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17675 (vect_make_slp_decision): Likewise.
17676 (vect_supported_load_permutation_p): Likewise, and polynomial
17678 (vect_analyze_slp_cost): Handle polynomial vf.
17679 (vect_slp_analyze_node_operations): Likewise.
17680 (vect_slp_analyze_bb_1): Likewise.
17681 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17682 than an unsigned HOST_WIDE_INT.
17683 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17684 (vectorizable_load): Handle polynomial vf.
17685 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17687 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17690 Alan Hayward <alan.hayward@arm.com>
17691 David Sherwood <david.sherwood@arm.com>
17693 * match.pd: Handle bit operations involving three constants
17694 and try to fold one pair.
17696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17698 * tree-vect-loop-manip.c: Include gimple-fold.h.
17699 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17700 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17701 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17702 Add a path that uses a step of VF instead of 1, but disable it
17704 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17705 and niters_no_overflow parameters. Update calls to
17706 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17707 Create a new SSA name if the latter choses to use a ste other
17708 than zero, and return it via niters_vector_mult_vf_var.
17709 * tree-vect-loop.c (vect_transform_loop): Update calls to
17710 vect_do_peeling, vect_gen_vector_loop_niters and
17711 slpeel_make_loop_iterate_ntimes.
17712 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17713 (vect_gen_vector_loop_niters): Update declarations after above changes.
17715 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17717 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17718 128-bit round to integer instructions.
17719 (ceil<mode>2): Likewise.
17720 (btrunc<mode>2): Likewise.
17721 (round<mode>2): Likewise.
17723 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17725 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17726 unaligned VSX load/store on P8/P9.
17727 (expand_block_clear): Allow the use of unaligned VSX
17728 load/store on P8/P9.
17730 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17732 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17734 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17735 swap associated with both a load and a store.
17737 2018-01-02 Andrew Waterman <andrew@sifive.com>
17739 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17740 * config/riscv/riscv.md (clear_cache): Use it.
17742 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17744 * web.c: Remove out-of-date comment.
17746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17748 * expr.c (fixup_args_size_notes): Check that any existing
17749 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17750 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17751 (emit_single_push_insn): ...here.
17753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17755 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17756 (const_vector_encoded_nelts): New function.
17757 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17758 (const_vector_int_elt, const_vector_elt): Declare.
17759 * emit-rtl.c (const_vector_int_elt_1): New function.
17760 (const_vector_elt): Likewise.
17761 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17762 of CONST_VECTOR_ELT.
17764 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17766 * expr.c: Include rtx-vector-builder.h.
17767 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17768 directly on the tree encoding.
17769 (const_vector_from_tree): Likewise.
17770 * optabs.c: Include rtx-vector-builder.h.
17771 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17772 sequence of "u" values.
17773 * vec-perm-indices.c: Include rtx-vector-builder.h.
17774 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17775 directly on the vec_perm_indices encoding.
17777 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17779 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17780 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17781 * rtx-vector-builder.h: New file.
17782 * rtx-vector-builder.c: Likewise.
17783 * rtl.h (rtx_def::u2): Add a const_vector field.
17784 (CONST_VECTOR_NPATTERNS): New macro.
17785 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17786 (CONST_VECTOR_DUPLICATE_P): Likewise.
17787 (CONST_VECTOR_STEPPED_P): Likewise.
17788 (CONST_VECTOR_ENCODED_ELT): Likewise.
17789 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17790 (unwrap_const_vec_duplicate): Likewise.
17791 (const_vec_series_p): Check for a non-duplicated vector encoding.
17792 Say that the function only returns true for integer vectors.
17793 * emit-rtl.c: Include rtx-vector-builder.h.
17794 (gen_const_vec_duplicate_1): Delete.
17795 (gen_const_vector): Call gen_const_vec_duplicate instead of
17796 gen_const_vec_duplicate_1.
17797 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17798 (gen_const_vec_duplicate): Use rtx_vector_builder.
17799 (gen_const_vec_series): Likewise.
17800 (gen_rtx_CONST_VECTOR): Likewise.
17801 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17802 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17803 Build a new vector rather than modifying a CONST_VECTOR in-place.
17804 (handle_special_swappables): Update call accordingly.
17805 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17806 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17807 Build a new vector rather than modifying a CONST_VECTOR in-place.
17808 (handle_special_swappables): Update call accordingly.
17810 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17812 * simplify-rtx.c (simplify_const_binary_operation): Use
17813 CONST_VECTOR_ELT instead of XVECEXP.
17815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17817 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17818 the selector elements to be different from the data elements
17819 if the selector is a VECTOR_CST.
17820 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17821 ssizetype for the selector.
17823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17825 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17826 before testing each element individually.
17827 * tree-vect-generic.c (lower_vec_perm): Likewise.
17829 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17831 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17832 * selftest-run-tests.c (selftest::run_tests): Call it.
17833 * vector-builder.h (vector_builder::operator ==): New function.
17834 (vector_builder::operator !=): Likewise.
17835 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17836 (vec_perm_indices::all_from_input_p): New function.
17837 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17838 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17839 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17840 instead of reading the VECTOR_CST directly. Detect whether both
17841 vector inputs are the same before constructing the vec_perm_indices,
17842 and update the number of inputs argument accordingly. Use the
17843 utility functions added above. Only construct sel2 if we need to.
17845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17847 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17848 the broadcast of the low byte.
17849 (expand_mult_highpart): Use an explicit encoding for the permutes.
17850 * optabs-query.c (can_mult_highpart_p): Likewise.
17851 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17852 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17853 (vectorizable_bswap): Likewise.
17854 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17855 explicit encoding for the power-of-2 permutes.
17856 (vect_permute_store_chain): Likewise.
17857 (vect_grouped_load_supported): Likewise.
17858 (vect_permute_load_chain): Likewise.
17860 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17862 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17863 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17864 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17865 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17866 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17867 (vect_gen_perm_mask_any): Likewise.
17869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17871 * int-vector-builder.h: New file.
17872 * vec-perm-indices.h: Include int-vector-builder.h.
17873 (vec_perm_indices): Redefine as an int_vector_builder.
17874 (auto_vec_perm_indices): Delete.
17875 (vec_perm_builder): Redefine as a stand-alone class.
17876 (vec_perm_indices::vec_perm_indices): New function.
17877 (vec_perm_indices::clamp): Likewise.
17878 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17879 (vec_perm_indices::new_vector): New function.
17880 (vec_perm_indices::new_expanded_vector): Update for new
17881 vec_perm_indices class.
17882 (vec_perm_indices::rotate_inputs): New function.
17883 (vec_perm_indices::all_in_range_p): Operate directly on the
17884 encoded form, without computing elided elements.
17885 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17886 encoding. Update for new vec_perm_indices class.
17887 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17888 the given vec_perm_builder.
17889 (expand_vec_perm_var): Update vec_perm_builder constructor.
17890 (expand_mult_highpart): Use vec_perm_builder instead of
17891 auto_vec_perm_indices.
17892 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17893 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17894 or double series encoding as appropriate.
17895 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17896 vec_perm_indices instead of auto_vec_perm_indices.
17897 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17898 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17899 (vect_permute_store_chain): Likewise.
17900 (vect_grouped_load_supported): Likewise.
17901 (vect_permute_load_chain): Likewise.
17902 (vect_shift_permute_load_chain): Likewise.
17903 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17904 (vect_transform_slp_perm_load): Likewise.
17905 (vect_schedule_slp_instance): Likewise.
17906 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17907 (vectorizable_mask_load_store): Likewise.
17908 (vectorizable_bswap): Likewise.
17909 (vectorizable_store): Likewise.
17910 (vectorizable_load): Likewise.
17911 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17912 vec_perm_indices instead of auto_vec_perm_indices. Use
17913 tree_to_vec_perm_builder to read the vector from a tree.
17914 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17915 vec_perm_builder instead of a vec_perm_indices.
17916 (have_whole_vector_shift): Use vec_perm_builder and
17917 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17918 truncation to calc_vec_perm_mask_for_shift.
17919 (vect_create_epilog_for_reduction): Likewise.
17920 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17921 from auto_vec_perm_indices to vec_perm_indices.
17922 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17923 instead of changing individual elements.
17924 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17925 the vector in d.perm.
17926 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17927 from auto_vec_perm_indices to vec_perm_indices.
17928 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17929 instead of changing individual elements.
17930 (arm_vectorize_vec_perm_const): Use new_vector to install
17931 the vector in d.perm.
17932 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17933 Update vec_perm_builder constructor.
17934 (rs6000_expand_interleave): Likewise.
17935 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17936 (rs6000_expand_interleave): Likewise.
17938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17940 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17941 to qimode could truncate the indices.
17942 * optabs.c (expand_vec_perm_var): Likewise.
17944 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17946 * Makefile.in (OBJS): Add vec-perm-indices.o.
17947 * vec-perm-indices.h: New file.
17948 * vec-perm-indices.c: Likewise.
17949 * target.h (vec_perm_indices): Replace with a forward class
17951 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17952 * optabs.h: Include vec-perm-indices.h.
17953 (expand_vec_perm): Delete.
17954 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17955 (expand_vec_perm_const): Declare.
17956 * target.def (vec_perm_const_ok): Replace with...
17957 (vec_perm_const): ...this new hook.
17958 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17959 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17960 * doc/tm.texi: Regenerate.
17961 * optabs.def (vec_perm_const): Delete.
17962 * doc/md.texi (vec_perm_const): Likewise.
17963 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17964 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17965 expand_vec_perm for constant permutation vectors. Assert that
17966 the mode of variable permutation vectors is the integer equivalent
17967 of the mode that is being permuted.
17968 * optabs-query.h (selector_fits_mode_p): Declare.
17969 * optabs-query.c: Include vec-perm-indices.h.
17970 (selector_fits_mode_p): New function.
17971 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17972 is defined, instead of checking whether the vec_perm_const_optab
17973 exists. Use targetm.vectorize.vec_perm_const instead of
17974 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17975 fit in the vector mode before using a variable permute.
17976 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17977 vec_perm_indices instead of an rtx.
17978 (expand_vec_perm): Replace with...
17979 (expand_vec_perm_const): ...this new function. Take the selector
17980 as a vec_perm_indices rather than an rtx. Also take the mode of
17981 the selector. Update call to shift_amt_for_vec_perm_mask.
17982 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17983 Use vec_perm_indices::new_expanded_vector to expand the original
17984 selector into bytes. Check whether the indices fit in the vector
17985 mode before using a variable permute.
17986 (expand_vec_perm_var): Make global.
17987 (expand_mult_highpart): Use expand_vec_perm_const.
17988 * fold-const.c: Includes vec-perm-indices.h.
17989 * tree-ssa-forwprop.c: Likewise.
17990 * tree-vect-data-refs.c: Likewise.
17991 * tree-vect-generic.c: Likewise.
17992 * tree-vect-loop.c: Likewise.
17993 * tree-vect-slp.c: Likewise.
17994 * tree-vect-stmts.c: Likewise.
17995 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17997 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17998 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17999 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18000 (aarch64_vectorize_vec_perm_const): ...this new function.
18001 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18002 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18003 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18004 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18005 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18006 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18007 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18009 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18010 check for NEON modes.
18011 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18012 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18013 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18014 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18016 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18017 the old VEC_PERM_CONST conditions.
18018 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18019 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18020 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18021 (ia64_vectorize_vec_perm_const_ok): Merge into...
18022 (ia64_vectorize_vec_perm_const): ...this new function.
18023 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18024 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18025 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18026 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18027 * config/mips/mips.c (mips_expand_vec_perm_const)
18028 (mips_vectorize_vec_perm_const_ok): Merge into...
18029 (mips_vectorize_vec_perm_const): ...this new function.
18030 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18031 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18032 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18033 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18034 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18035 (rs6000_expand_vec_perm_const): Delete.
18036 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18038 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18039 (altivec_expand_vec_perm_const_le): Take each operand individually.
18040 Operate on constant selectors rather than rtxes.
18041 (altivec_expand_vec_perm_const): Likewise. Update call to
18042 altivec_expand_vec_perm_const_le.
18043 (rs6000_expand_vec_perm_const): Delete.
18044 (rs6000_vectorize_vec_perm_const_ok): Delete.
18045 (rs6000_vectorize_vec_perm_const): New function.
18046 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18047 an element count and rtx array.
18048 (rs6000_expand_extract_even): Update call accordingly.
18049 (rs6000_expand_interleave): Likewise.
18050 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18051 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18052 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18053 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18054 (rs6000_expand_vec_perm_const): Delete.
18055 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18056 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18057 (altivec_expand_vec_perm_const_le): Take each operand individually.
18058 Operate on constant selectors rather than rtxes.
18059 (altivec_expand_vec_perm_const): Likewise. Update call to
18060 altivec_expand_vec_perm_const_le.
18061 (rs6000_expand_vec_perm_const): Delete.
18062 (rs6000_vectorize_vec_perm_const_ok): Delete.
18063 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18064 reference to the SPE evmerge intructions.
18065 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18066 an element count and rtx array.
18067 (rs6000_expand_extract_even): Update call accordingly.
18068 (rs6000_expand_interleave): Likewise.
18069 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18070 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18072 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18076 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18077 vector mode and that that mode matches the mode of the data
18079 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18080 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18081 directly using expand_vec_perm_1 when forcing selectors into
18083 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18085 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18087 * optabs-query.h (can_vec_perm_p): Delete.
18088 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18089 * optabs-query.c (can_vec_perm_p): Split into...
18090 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18091 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18092 particular selector is valid.
18093 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18094 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18095 (vect_grouped_load_supported): Likewise.
18096 (vect_shift_permute_load_chain): Likewise.
18097 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18098 (vect_transform_slp_perm_load): Likewise.
18099 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18100 (vectorizable_bswap): Likewise.
18101 (vect_gen_perm_mask_checked): Likewise.
18102 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18103 implementations of variable permutation vectors into account
18104 when deciding which selector to use.
18105 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18106 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18107 with a false third argument.
18108 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18109 to test whether the constant selector is valid and can_vec_perm_var_p
18110 to test whether a variable selector is valid.
18112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18114 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18115 * optabs-query.c (can_vec_perm_p): Likewise.
18116 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18117 instead of vec_perm_indices.
18118 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18119 (vect_gen_perm_mask_checked): Likewise,
18120 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18121 (vect_gen_perm_mask_checked): Likewise,
18123 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18125 * optabs-query.h (qimode_for_vec_perm): Declare.
18126 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18127 (qimode_for_vec_perm): ...this new function.
18128 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18130 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18132 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18133 does not have a conditional at the top.
18135 2018-01-02 Richard Biener <rguenther@suse.de>
18137 * ipa-inline.c (big_speedup_p): Fix expression.
18139 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18142 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18145 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18149 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18150 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18151 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18152 cond_taken_branch_cost 3->4.
18154 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18156 PR tree-optimization/83581
18157 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18158 TODO_cleanup_cfg if any changes have been made.
18160 PR middle-end/83608
18161 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18162 convert_modes if target mode has the right side, but different mode
18165 PR middle-end/83609
18166 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18167 last argument when extracting from CONCAT. If either from_real or
18168 from_imag is NULL, use expansion through memory. If result is not
18169 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18170 the parts directly to inner mode, if even that fails, use expansion
18173 PR middle-end/83623
18174 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18175 check for bswap in mode rather than HImode and use that in expand_unop
18178 Copyright (C) 2018 Free Software Foundation, Inc.
18180 Copying and distribution of this file, with or without modification,
18181 are permitted in any medium without royalty provided the copyright
18182 notice and this notice are preserved.