1 2018-06-28 Martin Liska <mliska@suse.cz>
3 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4 Add new checking assert to catch invalid state.
5 (jump_table_cluster::can_be_handled): Handle single case
7 (jump_table_cluster::is_beneficial): Bail out for such case.
8 (bit_test_cluster::find_bit_tests):
9 Add new checking assert to catch invalid state.
10 (bit_test_cluster::can_be_handled): Handle single case
12 (bit_test_cluster::is_beneficial): Bail out for such case.
13 (switch_decision_tree::analyze_switch_statement):
16 2018-06-28 Martin Liska <mliska@suse.cz>
18 * common.opt: Introduce -completion option.
19 * gcc.c (driver_handle_option): Handle it.
20 (driver::main): Print completions if completion
22 * opt-suggestions.c (option_proposer::get_completions):
24 (option_proposer::suggest_completion): Likewise.
25 (option_proposer::find_param_completions): Likewise.
26 (verify_autocompletions): Likewise.
27 (test_completion_valid_options): Likewise.
28 (test_completion_valid_params): Likewise.
29 (in_completion_p): Likewise.
30 (empty_completion_p): Likewise.
31 (test_completion_partial_match): Likewise.
32 (test_completion_garbage): Likewise.
33 (opt_proposer_c_tests): Likewise.
34 * opt-suggestions.h: Declare new functions.
35 * opts.c (common_handle_option): Handle OPT__completion_.
36 * selftest-run-tests.c (selftest::run_tests): Add
38 * selftest.c (assert_str_startswith): New.
39 * selftest.h (assert_str_startswith): Likewise.
40 (opt_proposer_c_tests): New.
41 (ASSERT_STR_STARTSWITH): Likewise.
43 2018-06-28 Martin Liska <mliska@suse.cz>
45 * Makefile.in: Add opt-suggestions.o.
46 * gcc-main.c: Include opt-suggestions.h.
47 * gcc.c (driver::driver): Likewise.
48 (driver::~driver): Remove m_option_suggestions.
49 (driver::build_option_suggestions): Moved to option_proposer.
50 (driver::suggest_option): Likewise.
51 (driver::handle_unrecognized_options): Use option_proposer.
52 * gcc.h (class driver): Add new memver m_option_proposer.
53 * opt-suggestions.c: New file.
54 * opt-suggestions.h: New file.
56 2018-06-28 Martin Liska <mliska@suse.cz>
58 * vec.h (class auto_string_vec): New (moved from auto_argvec).
59 (auto_string_vec::~auto_string_vec): Likewise.
61 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
63 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
64 statement if id->reset_location is true.
65 (copy_edges_for_bb): Do not set goto_locus on the new edges if
66 id->reset_location is true.
67 (copy_phis_for_bb): Force input_location on the arguments if
68 id->reset_location is true.
69 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
70 is set on the function to be inlined.
71 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
72 prevent_decl_creation_for_types fields up and add reset_location field.
74 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
76 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
78 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
80 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
82 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
83 hard registers for the clobbered pseudo.
85 2018-06-27 Paul Koning <ni1d@arrl.net>
87 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
88 mutually exclusive options.
89 * config/pdp11/constraints.md (h): New constraint.
90 (O): Update definition to match shift code generation.
92 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
94 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
96 (output_jump): Change arguments.
97 (pdp11_fixed_cc_regs): New function.
98 (pdp11_cc_mode): Ditto.
99 (pdp11_expand_shift): Ditto.
100 (pdp11_assemble_shift): Ditto.
101 (pdp11_small_shift): Ditto.
102 (pdp11_branch_cost): Remove.
103 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
105 (pdp11_register_move_cost): Update for CC registers.
106 (pdp11_rtx_costs): Add case for LSHIFTRT.
107 (pdp11_output_jump): Add CCNZ mode conditional branches.
108 (notice_update_cc_on_set): Remove.
109 (pdp11_cc_mode): New function.
110 (simple_memory_operand): Correct pre/post decrement case.
111 (no_side_effect_operand): New function.
112 (pdp11_regno_reg_class): Add CC_REGS class.
113 (pdp11_fixed_cc_regs): New function.
114 (pdp11_small_shift): New function.
115 (pdp11_expand_shift): New function to expand shift insns.
116 (pdp11_assemble_shift): New function to output shifts.
117 (pdp11_branch_cost): Remove.
118 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
119 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
121 (PTRDIFF_TYPE): Ditto.
122 (ADJUST_INSN_LENGTH): New macro.
123 (FIXED_REGISTERS): Add CC registers.
124 (CALL_USED_REGISTERS): Ditto.
126 (REG_CLASS_NAMES): Ditto.
127 (REG_CLASS_CONTENTS): Ditto.
128 (SELECT_CC_MODE): Use new function.
129 (TARGET_FLAGS_REGNUM): New macro.
130 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
131 (cc0_reg_rtx): Remove.
132 (CC_STATUS_MDEP): Remove.
133 (CC_STATUS_MDEFP_INIT): Remove.
135 (NOTICE_UPDATE_CC): Remove.
136 (REGISTER_NAMES): Add CC registers.
137 (BRANCH_COST): Change to constant 1.
138 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
140 * config/pdp11/pdp11.opt (mbcopy): Remove.
141 (mbcopy-builtin): Remove.
142 (mbranch-cheap): Remove.
143 (mbranch-expensive): Remove.
144 * config/pdp11/predicates.md (expand_shift_operand): Update to
145 match shift code generation.
146 (ccnz_operator): New predicate.
147 * doc/invoke.texi (PDP-11 Options): Remove deleted options
148 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
149 Remove non-existent option -mabshi, -mno-abshi. Document mutually
151 * doc/md.texi (PDP-11): Document new D and h constraints. Update
152 description of O constraint.
154 2018-06-27 Jeff Law <law@redhat.com>
155 Austin Law <austinklaw@gmail.com>
157 * config/v850/v850.md (addsi3_set_flags): New pattern.
158 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
159 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
160 (zero_extendhisi2_v850_set_flags): Likewise.
161 (zero_extendqisi2_v850_set_flags): Likewise.
162 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
163 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
164 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
166 * config/v850/v850-protos.h (notice_update_cc): Remove.
167 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
168 (v850_print_operand): Handle 'D' and "d".
169 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
170 Add handling of arithmetic/logical operations compared against zero.
171 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
172 Do not look at v850_compare_op, instead get mode from last argument.
173 (v850_gen_compare): Remove
174 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
175 after reload for prologue insns.
176 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
178 (construct_save_jarl): Likewise.
179 (TARGET_FLAGS_REGNUM): Define.
180 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
181 (NOTICE_UPDATE_CC): Remove.
182 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
183 than cc0. Conditionalize on reload_completed.
184 (cmpsi_insn, setfcc_insn): Likewise.
185 (tst1 splitter): Turn into define_and_split which sets the flags
187 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
188 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
189 (cstoresf4, cstoredf4): Clobber the flags.
190 (cmpsi, cmpsf, cmpdf): Remove expanders.
191 (setf_insn): Remove pattern.
192 (addsi3): Turn into define_and_split which clobbers the flags after
193 reload and a suitable pattern (addsi3_clobber_flags) for use after
195 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
196 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
197 (ashrsi3, ashrsi3_v850e2): Likewise.
198 (bins): Clobber the flags.
199 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
200 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
201 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
202 (call_value_internal_short, call_value_internal_long): Likewise.
203 (callt_save_interrupt, callt_return_interrupt): Likewise.
204 (save_interrupt, return_interrupt): Likewise.
205 (callt_save_all_interrupt, save_all_interrupt): Likewise.
206 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
207 (restore_all_interrupt, _restore_all_interrupt): Likewise.
208 (All FP comparisons): Only allow after reload has completed.
210 (divh, divhu): Tweak output template.
211 (branch_z_normal, branch_z_invert): Remove
212 (branch_nz_normal, branch_nz_invert): Likewise.
213 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
215 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
216 * config/v850/v850.c (notice_update_cc): Remove.
217 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
218 (CC_NO_CARRY): Likewise.
219 (NOTICE_UPDATE_CC): Define to nothing.
220 * config/v850/v850.md: Remove block comment on cc0 handling
221 Remove "cc" attribute from all patterns. Remove cc_status handling
222 from all patterns. Minor formatting fixes.
224 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
226 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
227 (cortex-a76.cortex-a55): Likewise.
228 * config/aarch64/aarch64-tune.md: Regenerate.
229 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
230 cortex-a76.cortex-a55.
232 2018-06-27 Jeff Law <law@redhat.com>
234 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
235 (MULTILIB_DIRNAMES): Similarly.
237 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
239 * gimple.h (gimple_return_retbnd): Delete.
240 (gimple_return_set_retbnd): Likewise.
241 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
242 gimple_return_set_retbnd.
243 * gimple-pretty-print.c (dump_gimple_return): Remove call to
244 gimple_return_retbnd and adjust.
245 * tree-inline.h (struct copy_body_data): Remove retbnd field.
246 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
247 Explicitly return NULL in a couple more cases. Move assertion
248 on debug statements and remove unreachable code.
249 (reset_debug_binding): Do not test id->retbnd.
250 (expand_call_inline): Do not set it.
252 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
254 * configure.ac: Add --disable-gcov option.
255 * configure: Regenerate.
256 * Makefile.in: Honour @enable_gcov@.
257 * doc/install.texi: Document --disable-gcov.
259 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
261 * config/arm/arm-cpus.in (cortex-a76): New entry.
262 (cortex-a76.cortex-a55): Likewise.
263 * config/arm/arm-tables.opt: Regenerate.
264 * config/arm/arm-tune.md: Likewise.
265 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
266 * doc/invoke.texi (ARM Options): Document cortex-a76 and
267 cortex-a76.cortex-a55.
269 2018-06-27 Tamar Christina <tamar.christina@arm.com>
272 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
274 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
276 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
278 (EPILOGUE_USES): Likewise.
280 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
282 * tree-inline.c (remap_location): New function extracted from...
283 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
284 (copy_phis_for_bb): ...here. Call remap_location.
285 (copy_cfg_body): Adjust call to copy_edges_for_bb.
287 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
289 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
290 unaligned vsx for 16B memset.
292 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
295 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
296 ieee128_float_type_node to long_double_type_node unless
297 TARGET_LONG_DOUBLE_128 is set.
299 2018-06-26 David Malcolm <dmalcolm@redhat.com>
301 * cfgloop.c (get_loop_location): Convert return type from
302 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
303 by implicit construction from rtx_insn *, and using
304 dump_user_location_t::from_function_decl for the fallback case.
305 * cfgloop.h (get_loop_location): Convert return type from
306 location_t to dump_user_location_t.
307 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
308 dump_printf_loc to pass in a dump_location_t rather than a
309 location_t, via the gimple stmt.
310 * coverage.c (get_coverage_counts): Update calls to
311 dump_printf_loc to pass in dump_location_t rather than a
313 * doc/optinfo.texi (Dump types): Convert example of
314 dump_printf_loc from taking "locus" to taking "insn". Update
315 description of the "_loc" calls to cover dump_location_t.
316 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
318 (dump_user_location_t::dump_user_location_t): New constructors,
319 from gimple *stmt and rtx_insn *.
320 (dump_user_location_t::from_function_decl): New function.
321 (dump_loc): Make static.
322 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
323 const dump_location_t &.
324 (dump_generic_expr_loc): Delete.
325 (dump_printf_loc): Convert param "loc" from location_t to
326 const dump_location_t &.
327 (selftest::test_impl_location): New function.
328 (selftest::dumpfile_c_tests): New function.
329 * dumpfile.h: Include "profile-count.h".
330 (class dump_user_location_t): New class.
331 (struct dump_impl_location_t): New struct.
332 (class dump_location_t): New class.
333 (dump_printf_loc): Convert 2nd param from source_location to
334 const dump_location_t &.
335 (dump_generic_expr_loc): Delete.
336 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
337 const dump_location_t &.
338 * gimple-fold.c (fold_gimple_assign): Update call to
339 dump_printf_loc to pass in a dump_location_t rather than a
340 location_t, via the gimple stmt.
341 (gimple_fold_call): Likewise.
342 * gimple-loop-interchange.cc
343 (loop_cand::analyze_iloop_reduction_var): Update for change to
344 check_reduction_path.
345 (tree_loop_interchange::interchange): Update for change to
347 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
348 change in return-type of find_loop_location.
349 (graphite_regenerate_ast_isl): Likewise.
350 * graphite-optimize-isl.c (optimize_isl): Likewise.
351 * graphite.c (graphite_transform_loops): Likewise.
352 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
353 pass in a dump_location_t rather than a location_t, via the
355 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
356 * ipa.c (walk_polymorphic_call_targets): Likewise.
357 * loop-unroll.c (report_unroll): Convert "locus" param from
358 location_t to dump_location_t.
359 (decide_unrolling): Update for change to get_loop_location's
361 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
362 location_t to dump_user_location_t.
363 (grid_find_single_omp_among_assignments_1): Updates calls to
364 dump_printf_loc to pass in a dump_location_t rather than a
365 location_t, via the gimple stmt.
366 (grid_parallel_clauses_gridifiable): Convert "tloc" from
367 location_t to dump_location_t. Updates calls to dump_printf_loc
368 to pass in a dump_location_t rather than a location_t, via the
370 (grid_inner_loop_gridifiable_p): Likewise.
371 (grid_dist_follows_simple_pattern): Likewise.
372 (grid_gfor_follows_tiling_pattern): Likewise.
373 (grid_target_follows_gridifiable_pattern): Likewise.
374 (grid_attempt_target_gridification): Convert initialization
375 of local "grid" from memset to zero-initialization; FIXME: does
376 this require C++11? Update call to dump_printf_loc to pass in a
377 optinfo_location rather than a location_t, via the gimple stmt.
378 * profile.c (read_profile_edge_counts): Updates call to
379 dump_printf_loc to pass in a dump_location_t rather than a
381 (compute_branch_probabilities): Likewise.
382 * selftest-run-tests.c (selftest::run_tests): Call
384 * selftest.h (dumpfile_c_tests): New decl.
385 * tree-loop-distribution.c (pass_loop_distribution::execute):
386 Update for change in return type of find_loop_location.
387 * tree-parloops.c (parallelize_loops): Likewise.
388 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
389 "locus" from location_t to dump_user_location_t.
390 (canonicalize_loop_induction_variables): Likewise.
391 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
392 for change in return type of find_loop_location.
393 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
394 to dump_printf_loc to pass in a dump_location_t rather than a
395 location_t, via the stmt.
396 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
398 * tree-vect-loop-manip.c (find_loop_location): Convert return
399 type from source_location to dump_user_location_t.
400 (vect_do_peeling): Update for above change.
401 (vect_loop_versioning): Update for change in type of
403 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
404 from location_t to dump_user_location_t.
405 (vect_estimate_min_profitable_iters): Update for change in type
407 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
408 location_t to dump_location_t.
409 (vect_slp_bb): Update for change in type of vect_location.
410 * tree-vectorizer.c (vect_location): Convert from source_location
411 to dump_user_location_t.
412 (try_vectorize_loop_1): Update for change in vect_location's type.
413 (vectorize_loops): Likewise.
414 (increase_alignment): Likewise.
415 * tree-vectorizer.h (vect_location): Convert from source_location
416 to dump_user_location_t.
417 (find_loop_location): Convert return type from source_location to
418 dump_user_location_t.
419 (check_reduction_path): Convert 1st param from location_t to
420 dump_user_location_t.
421 * value-prof.c (check_counter): Update call to dump_printf_loc to
422 pass in a dump_user_location_t rather than a location_t; update
423 call to error_at for change in type of "locus".
424 (check_ic_target): Update call to dump_printf_loc to
425 pass in a dump_user_location_t rather than a location_t, via the
428 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
430 * config/s390/s390.h (enum processor_flags): Do not use
431 default tune parameter when -march was specified.
433 2018-06-26 Jakub Jelinek <jakub@redhat.com>
436 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
437 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
440 2018-06-26 Richard Biener <rguenther@suse.de>
442 PR tree-optimization/86287
444 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
445 (vect_analyze_loop): Initialize n_stmts.
447 2018-06-26 Richard Biener <rguenther@suse.de>
450 * fold-const.c (fold_convertible_p): Pointer extension
453 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
456 * dwarf2out.c (loc_list_has_views): Adjust comments.
457 (dw_loc_list): Split single cross-partition range with
460 2018-06-25 Jeff Law <law@redhat.com>
462 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
463 on -mbig-switch by default.
465 * config/v850/predicates.md (const_float_1_operand): Fix match_code
467 (const_float_0_operand): Remove unused predicate.
468 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
469 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
470 (recipsf2): New expander. Original pattern now called
472 (recipdf2, recipdf2_insn): Similarly.
473 (rsqrtsf2, rsqrtsf2_insn): Similarly
474 (rsqrtdf2, rsqrtdf2_insn): Similarly
476 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
478 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
479 Simplify logic for FreeBSD (twice).
481 2018-06-25 Martin Sebor <msebor@redhat.com>
483 PR tree-optimization/86204
484 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
485 a strnlen result if it's less than the length of the string.
487 2018-06-25 Martin Sebor <msebor@redhat.com>
489 PR tree-optimization/85700
490 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
491 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
492 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
494 2018-06-25 Martin Sebor <msebor@redhat.com>
496 * doc/extend.texi (Zero-length arrays): Update and clarify.
498 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
500 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
501 added IEEE/IBM long double multilib support on PowerPC little
502 endian Linux systems.
503 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
504 (MULTILIB_DEFAULTS): Likewise.
505 * config/rs6000/rs6000.c (rs6000_option_override_internal):
507 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
508 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
509 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
511 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
514 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
515 (REORDER_45): Likewise.
517 2018-06-25 Jeff Law <law@redhat.com>
519 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
520 dividend to 32 bits. Adjust length.
521 (udivmodhi4): Cleanup output template. Fix length.
523 2018-06-25 Carl Love <cel@us.ibm.com>
525 * config/rs6000/vsx.md: Change word selector to prefered location.
527 2018-06-25 Richard Biener <rguenther@suse.de>
529 PR tree-optimization/86304
530 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
531 epilogue-if-converted loops as well.
533 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
535 * lto-section-out.c (lto_begin_section): Do not print section
536 name for noaddr and unnumbered dumps.
538 2018-06-25 Richard Biener <rguenther@suse.de>
540 * tree-vectorizer.h (struct vec_info_shared): New structure
541 with parts split out from struct vec_info and loop_nest from
542 struct _loop_vec_info.
543 (struct vec_info): Adjust accordingly.
544 (struct _loop_vec_info): Likewise.
545 (LOOP_VINFO_LOOP_NEST): Adjust.
546 (LOOP_VINFO_DATAREFS): Likewise.
547 (LOOP_VINFO_DDRS): Likewise.
548 (struct _bb_vec_info): Likewise.
549 (BB_VINFO_DATAREFS): Likewise.
550 (BB_VINFO_DDRS): Likewise.
551 (struct _stmt_vec_info): Add dr_aux member.
552 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
553 (DR_MISALIGNMENT_UNINITIALIZED): New.
554 (set_dr_misalignment): Adjust.
555 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
556 (vect_analyze_loop): Adjust prototype.
557 (vect_analyze_loop_form): Likewise.
558 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
559 Compute dependences lazily.
560 (vect_record_base_alignments): Use shared datarefs/ddrs.
561 (vect_verify_datarefs_alignment): Likewise.
562 (vect_analyze_data_refs_alignment): Likewise.
563 (vect_analyze_data_ref_accesses): Likewise.
564 (vect_analyze_data_refs): Likewise.
565 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
566 constructor parameter for shared part.
567 (vect_analyze_loop_form): Pass in shared part and adjust.
568 (vect_analyze_loop_2): Pass in storage for the number of
569 stmts. Move loop nest finding to the caller. Compute
571 (vect_analyze_loop): Pass in shared part.
572 (vect_transform_loop): Verify shared datarefs are unchanged.
573 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
574 constructor parameter for shared part.
575 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
576 (vect_slp_bb): Verify shared datarefs are unchanged before
578 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
580 (new_stmt_vec_info): Initialize DR_AUX misalignment to
581 DR_MISALIGNMENT_UNINITIALIZED.
582 * tree-vectorizer.c (vec_info::vec_info): Add constructor
583 parameter for shared part.
584 (vec_info::~vec_info): Adjust.
585 (vec_info_shared::vec_info_shared): New.
586 (vec_info_shared::~vec_info_shared): Likewise.
587 (vec_info_shared::save_datarefs): Likewise.
588 (vec_info_shared::check_datarefs): Likewise.
589 (try_vectorize_loop_1): Construct shared part live for analyses
590 of a single loop for multiple vector sizes.
591 * tree-parloops.c (gather_scalar_reductions): Adjust.
593 2018-06-25 Richard Biener <rguenther@suse.de>
595 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
596 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
597 (vect_analyze_data_refs): Remove similar code from here and
598 simplify accordingly.
600 2018-06-25 Richard Biener <rguenther@suse.de>
602 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
603 for reverse storage order accesses rather than asserting
604 they cannot happen here.
606 2018-06-25 Tom de Vries <tdevries@suse.de>
609 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
610 Use data16 instead of .byte for insn prefix.
612 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
615 * parser.c (make_char_string_pack): Pass this literal chars
616 through cpp_interpret_string.
617 (cp_parser_userdef_numeric_literal): Check the result of
618 make_char_string_pack.
620 2018-06-24 Maya Rashish <coypu@sdf.org>
622 * ginclude/stddef.h: Simplify conditions around avoiding
623 re-definition of __size_t.
625 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
627 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
628 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
630 2018-06-22 Maya Rashish <coypu@sdf.org>
632 * doc/invoke.texi (mno-fancy-math-387): Update for changes
633 made to OpenBSD and NetBSD through the years.
635 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
637 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
638 behavior of vec_pack (vector double, vector double) to match
639 behavior of vec_float2 (vector double, vector double).
641 2018-06-22 Olivier Hainque <hainque@adacore.com>
643 * gimplify.c (gimplify_function_tree): Prevent creation
644 of a trampoline for the address of the current function
645 passed to entry/exit instrumentation hooks.
647 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
650 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
653 2018-06-22 Martin Liska <mliska@suse.cz>
655 PR tree-optimization/86263
656 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
657 Bail out if is_enabled is false.
658 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
660 (jump_table_cluster::is_enabled): New function.
662 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
664 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
665 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
666 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
667 (lto_input_ts_binfo_tree_pointers): Likewise.
668 * tree-streamer-out.c (streamer_write_tree_bitfields,
669 write_ts_binfo_tree_pointers): Likewise.
670 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
672 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
674 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
676 2018-06-22 Martin Liska <mliska@suse.cz>
678 * symbol-summary.h (get): Make it pure and inline move
679 functionality from ::get function.
680 (get): Remove and inline into ::get and ::get_create.
681 (get_create): Move code from ::get function.
683 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
686 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
687 -x assembler-with-cpp.
689 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
691 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
692 _FILE_OFFSET_BITS=64 for C++.
694 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
696 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
697 conversion insn that shows up when pr85657-3.c is compiled using
698 IEEE 128-bit long double.
699 (neg<mode>2_internal): Use the correct mode to check whether the
700 mode is IBM extended.
701 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
702 multiply and divide external functions from being created more
705 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
707 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
709 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
710 the edge can be forwarded.
711 (cfg_layout_merge_blocks): Likewise.
713 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
715 * except.c (finish_eh_generation): Commit edge insertions only after
716 the EH edges have been redirected from post-landing to landing pads.
718 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
720 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
721 create_tmp_var_for to create the FRAME decl.
722 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
724 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
726 * tree-inline.c (copy_edges_for_bb): Minor tweak.
727 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
728 debug statement when resetting its value.
729 (expand_call_inline): Copy the locus of the call onto the assignment
730 of the return value, if any. Use local variable in more cases.
732 2018-06-21 Martin Liska <mliska@suse.cz>
734 * ipa-pure-const.c (propagate_nothrow): Use
735 funct_state_summaries->get.
736 (dump_malloc_lattice): Likewise.
737 (propagate_malloc): Likewise.
739 2018-06-21 Richard Biener <rguenther@suse.de>
741 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
742 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
743 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
745 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
746 BLOCK_ABSTRACT_ORIGIN unconditionally.
748 2018-06-21 David Malcolm <dmalcolm@redhat.com>
750 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
752 * ipa-reference.c (ipa_reference_c_finalize): Delete
753 ipa_ref_opt_sum_summaries and set it to NULL.
755 2018-06-21 Tom de Vries <tdevries@suse.de>
757 PR tree-optimization/85859
758 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
759 test with comment from bb_no_side_effects_p.
761 2018-06-21 Richard Biener <rguenther@suse.de>
763 PR tree-optimization/86232
764 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
765 max for constant niter.
767 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
769 * config/aarch64/aarch64-simd.md
770 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
772 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
774 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
775 Make opernads of the unspec commutative.
777 2018-06-21 Richard Biener <rguenther@suse.de>
779 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
780 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
781 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
782 (vect_analyze_data_ref_dependence): Re-order checks to deal with
784 (vect_record_base_alignments): Do not record base alignment
785 for gathers or scatters.
786 (vect_compute_data_ref_alignment): Drop return value that is always
787 true. Bail out early for gathers or scatters.
788 (vect_enhance_data_refs_alignment): Bail out early for gathers
790 (vect_find_same_alignment_drs): Likewise.
791 (vect_analyze_data_refs_alignment): Remove dead code.
792 (vect_slp_analyze_and_verify_node_alignment): Likewise.
793 (vect_analyze_data_refs): For possible gathers or scatters do
794 not create an alternate DR, just check their possible validity
795 and mark them. Adjust DECL_NONALIASED handling to not rely
797 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
798 update inits of gathers or scatters.
799 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
800 Also copy gather/scatter flag to pattern vinfo.
802 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
804 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
805 behavior of vec_packsu (vector unsigned long long, vector unsigned
806 long long) to match behavior of vec_packs with same signature.
808 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
809 Thomas Schwinge <thomas@codesourcery.com>
810 Cesar Philippidis <cesar@codesourcery.com>
812 * gimplify.c (gimplify_scan_omp_clauses): Add support for
813 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
814 (gimplify_adjust_omp_clauses): Likewise.
815 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
816 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
817 (gimplify_omp_target_update): Update handling of acc update and
819 * omp-low.c (install_var_field): Remove unused parameter
820 base_pointers_restrict.
821 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
822 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
824 (omp_target_base_pointers_restrict_p): Delete.
825 (scan_omp_target): Update call to scan_sharing_clauses.
826 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
828 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
829 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
830 (convert_local_omp_clauses): Likewise.
831 * tree-pretty-print.c (dump_omp_clause): Likewise.
832 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
834 (omp_clause_code_name): Likewise.
836 2018-06-20 Jakub Jelinek <jakub@redhat.com>
839 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
842 PR tree-optimization/86231
843 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
844 anti-range don't overwrite *vr0min before using it to compute *vr0max.
846 2018-06-20 Tom de Vries <tdevries@suse.de>
848 PR tree-optimization/86097
849 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
850 iv type if signedness of iv type is not the same as that of *nit.
852 2018-06-20 Jakub Jelinek <jakub@redhat.com>
854 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
855 EDGE_EH edges, verify they are all EDGE_EH.
857 2018-06-20 Maya Rashish <coypu@sdf.org>
859 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
861 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
863 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
864 * config/aarch64/aarch64.c (xgene1_tunings): Add
865 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
866 (aarch64_mode_valid_for_sched_fusion_p):
868 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
869 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
871 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
873 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
874 * config/aarch64/iterators.md (VQ2): New mode iterator.
876 2018-06-20 Martin Liska <mliska@suse.cz>
878 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
879 Change default ratio from 10 to 8.
881 2018-06-20 Martin Liska <mliska@suse.cz>
883 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
885 (bit_test_cluster::find_bit_tests): Likewise.
886 (switch_decision_tree::analyze_switch_statement): Find clusters.
887 * tree-switch-conversion.h (struct jump_table_cluster): Document
890 2018-06-20 Martin Liska <mliska@suse.cz>
892 * tree-switch-conversion.c (switch_conversion::collect):
893 Record m_uniq property.
894 (switch_conversion::expand): Bail out for special conditions.
895 (group_cluster::~group_cluster): New.
896 (group_cluster::group_cluster): Likewise.
897 (group_cluster::dump): Likewise.
898 (jump_table_cluster::emit): New.
899 (switch_decision_tree::fix_phi_operands_for_edges): New.
900 (struct case_node): Remove struct.
901 (jump_table_cluster::can_be_handled): New.
902 (case_values_threshold): Moved to header.
903 (reset_out_edges_aux): Likewise.
904 (jump_table_cluster::is_beneficial): New.
905 (bit_test_cluster::can_be_handled): Likewise.
906 (add_case_node): Remove.
907 (bit_test_cluster::is_beneficial): New.
908 (case_bit_test::cmp): New.
909 (bit_test_cluster::emit): New.
910 (expand_switch_as_decision_tree_p): Remove.
911 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
912 (fix_phi_operands_for_edge): Likewise.
913 (switch_decision_tree::analyze_switch_statement): New.
914 (compute_cases_per_edge): Move ...
915 (switch_decision_tree::compute_cases_per_edge): ... here.
916 (try_switch_expansion): Likewise.
917 (switch_decision_tree::try_switch_expansion): Likewise.
918 (record_phi_operand_mapping): Likewise.
919 (switch_decision_tree::record_phi_operand_mapping): Likewise.
920 (emit_case_decision_tree): Likewise.
921 (switch_decision_tree::emit): Likewise.
922 (balance_case_nodes): Likewise.
923 (switch_decision_tree::balance_case_nodes): Likewise.
924 (dump_case_nodes): Likewise.
925 (switch_decision_tree::dump_case_nodes): Likewise.
926 (emit_jump): Likewise.
927 (switch_decision_tree::emit_jump): Likewise.
928 (emit_cmp_and_jump_insns): Likewise.
929 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
930 (emit_case_nodes): Likewise.
931 (switch_decision_tree::emit_case_nodes): Likewise.
932 (conditional_probability): Remove.
933 * tree-switch-conversion.h (enum cluster_type): New.
935 (struct cluster): Likewise.
936 (cluster::cluster): Likewise.
937 (struct simple_cluster): Likewise.
938 (simple_cluster::simple_cluster): Likewise.
939 (struct group_cluster): Likewise.
940 (struct jump_table_cluster): Likewise.
941 (struct bit_test_cluster): Likewise.
942 (struct min_cluster_item): Likewise.
943 (struct case_tree_node): Likewise.
944 (case_tree_node::case_tree_node): Likewise.
945 (jump_table_cluster::case_values_threshold): Likewise.
946 (struct case_bit_test): Likewise.
947 (struct switch_decision_tree): Likewise.
948 (struct switch_conversion): Likewise.
949 (switch_decision_tree::reset_out_edges_aux): Likewise.
951 2018-06-20 Martin Liska <mliska@suse.cz>
953 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
954 (hoist_edge_and_branch_if_true): Likewise.
955 (expand_switch_using_bit_tests_p): Likewise.
956 (struct case_bit_test): Likewise.
957 (case_bit_test_cmp): Likewise.
958 (emit_case_bit_tests): Likewise.
959 (switch_conversion::switch_conversion): New class.
960 (struct switch_conv_info): Remove old struct.
961 (collect_switch_conv_info): More to ...
962 (switch_conversion::collect): ... this.
963 (check_range): Likewise.
964 (switch_conversion::check_range): Likewise.
965 (check_all_empty_except_final): Likewise.
966 (switch_conversion::check_all_empty_except_final): Likewise.
967 (check_final_bb): Likewise.
968 (switch_conversion::check_final_bb): Likewise.
969 (create_temp_arrays): Likewise.
970 (switch_conversion::create_temp_arrays): Likewise.
971 (free_temp_arrays): Likewise.
972 (gather_default_values): Likewise.
973 (switch_conversion::gather_default_values): Likewise.
974 (build_constructors): Likewise.
975 (switch_conversion::build_constructors): Likewise.
976 (constructor_contains_same_values_p): Likewise.
977 (switch_conversion::contains_same_values_p): Likewise.
978 (array_value_type): Likewise.
979 (switch_conversion::array_value_type): Likewise.
980 (build_one_array): Likewise.
981 (switch_conversion::build_one_array): Likewise.
982 (build_arrays): Likewise.
983 (switch_conversion::build_arrays): Likewise.
984 (gen_def_assigns): Likewise.
985 (switch_conversion::gen_def_assigns): Likewise.
986 (prune_bbs): Likewise.
987 (switch_conversion::prune_bbs): Likewise.
988 (fix_phi_nodes): Likewise.
989 (switch_conversion::fix_phi_nodes): Likewise.
990 (gen_inbound_check): Likewise.
991 (switch_conversion::gen_inbound_check): Likewise.
992 (process_switch): Use the newly created class.
993 (switch_conversion::expand): New.
994 (switch_conversion::~switch_conversion): New.
995 * tree-switch-conversion.h: New file.
997 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
999 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1000 tree-vect-patterns.c.
1001 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1002 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1003 (vect_recog_sad_pattern): Likewise.
1004 (vect_recog_widen_sum_pattern): Likewise.
1005 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1006 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1007 (vect_recog_rotate_pattern): Likewise.
1008 (vect_recog_mult_pattern): Likewise.
1009 (vect_recog_vector_vector_shift_pattern): Likewise.
1010 (vect_recog_divmod_pattern): Likewise.
1011 (vect_recog_mixed_size_cond_pattern): Likewise.
1012 (vect_recog_bool_pattern): Likewise.
1013 (vect_recog_mask_conversion_pattern): Likewise.
1014 (vect_try_gather_scatter_pattern): Likewise.
1015 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1016 (vect_recog_over_widening_pattern): Likewise.
1017 (vect_recog_gather_scatter_pattern): Likewise.
1018 (vect_recog_func_ptr): Move from tree-vectorizer.h
1019 (vect_vect_recog_func_ptrs): Move further down the file.
1020 (vect_recog_func): Likewise. Remove the third argument.
1021 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1022 (vect_pattern_recog_1): Expect the pattern function to do any
1023 necessary target tests. Also expect it to provide a vector type.
1024 Remove the type_in handling.
1026 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1028 * tree-vect-patterns.c (vect_pattern_detected): New function.
1029 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1030 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1031 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1032 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1033 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1034 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1035 (vect_recog_mask_conversion_pattern)
1036 (vect_try_gather_scatter_pattern): Likewise.
1038 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1040 * tree-vect-patterns.c (vect_get_internal_def): New function.
1041 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1042 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1043 (search_type_for_mask_1): Use it.
1045 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1047 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1048 redundant WIDEN_SUM_EXPR handling.
1049 (vect_recog_sad_pattern): Likewise.
1051 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1053 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1054 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1055 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1056 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1057 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1059 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1061 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1062 use the stmt_vec_info of the original bb statement for the
1063 new zero assignment, even if the call is part of a pattern.
1065 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1067 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1068 that the sequence is attached to the original statement rather
1069 than the pattern statement.
1070 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1071 PATTERN_DEF_SEQ from the original statement rather than
1072 the main pattern statement.
1073 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1074 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1075 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1077 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1079 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1080 definition statements before the early exit for statements that aren't
1082 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1084 (vect_transform_loop): ...here. Process pattern definition
1085 statements without first checking whether the main pattern
1086 statement is live or relevant.
1088 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1090 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1091 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1093 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1095 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1096 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1097 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1098 (expand_block_compare): Change select_block_compare_mode call.
1099 (expand_strncmp_align_check): Use new functions, fix comment.
1100 (emit_final_str_compare_gpr): New function.
1101 (expand_strn_compare): Refactor and clean up code.
1102 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1104 2018-06-19 Tony Reix <tony.reix@atos.com>
1105 Damien Bergamini <damien.bergamini@atos.com>
1106 David Edelsohn <dje.gcc@gmail.com>
1108 * collect2.c (static_obj): New variable.
1109 (static_libs): New variable.
1110 (is_in_list): Uncomment declaration.
1111 (main): Track AIX libraries linked statically.
1112 (is_in_list): Uncomment definition.
1113 (scan_prog_file): Don't add AIX shared libraries initializer
1114 to constructor list if linking statically.
1116 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1118 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1120 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1122 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1124 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1127 2018-06-19 Martin Liska <mliska@suse.cz>
1129 * config/i386/i386.c (ix86_can_inline_p): Do not use
1130 ipa_fn_summaries::get_create.
1131 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1133 (devirtualization_time_bonus): Likewise.
1134 (ipcp_propagate_stage): Likewise.
1135 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1136 (edge_set_predicate): Likewise.
1137 (evaluate_conditions_for_known_args): Likewise.
1138 (evaluate_properties_for_edge): Likewise.
1139 (ipa_call_summary::reset): Tranform to ...
1140 (ipa_call_summary::~ipa_call_summary): ... this.
1141 (ipa_fn_summary::reset): Transform to ...
1142 (ipa_fn_summary::~ipa_fn_summary): ... this.
1143 (ipa_fn_summary_t::remove): Rename to ...
1144 (ipa_fn_summary_t::remove_callees): ... this.
1145 (ipa_fn_summary_t::duplicate): Use placement new
1146 instead of memory copy.
1147 (ipa_call_summary_t::duplicate): Likewise.
1148 (ipa_call_summary_t::remove): Remove.
1149 (dump_ipa_call_summary): Change get_create to get.
1150 (ipa_dump_fn_summary): Dump only when summary exists.
1151 (analyze_function_body): Use symbol_summary::get instead
1153 (compute_fn_summary): Likewise.
1154 (estimate_edge_devirt_benefit): Likewise.
1155 (estimate_edge_size_and_time): Likewise.
1156 (inline_update_callee_summaries): Likewise.
1157 (remap_edge_change_prob): Likewise.
1158 (remap_edge_summaries): Likewise.
1159 (ipa_merge_fn_summary_after_inlining): Likewise.
1160 (write_ipa_call_summary): Likewise.
1161 (ipa_fn_summary_write): Likewise.
1162 (ipa_free_fn_summary): Likewise.
1163 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1164 (struct ipa_call_summary): Likewise.
1165 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1167 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1168 (estimate_size_after_inlining): Likewise.
1169 (estimate_growth): Likewise.
1170 (growth_likely_positive): Likewise.
1171 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1172 (inline_call): Likewise.
1173 * ipa-inline.c (caller_growth_limits): Likewise.
1174 (can_inline_edge_p): Likewise.
1175 (can_inline_edge_by_limits_p): Likewise.
1176 (compute_uninlined_call_time): Likewise.
1177 (compute_inlined_call_time): Likewise.
1178 (want_inline_small_function_p): Likewise.
1179 (edge_badness): Likewise.
1180 (update_caller_keys): Likewise.
1181 (update_callee_keys): Likewise.
1182 (inline_small_functions): Likewise.
1183 (inline_to_all_callers_1): Likewise.
1184 (dump_overall_stats): Likewise.
1185 (early_inline_small_functions): Likewise.
1186 (early_inliner): Likewise.
1187 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1188 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1189 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1190 * ipa-split.c (execute_split_functions): Likewise.
1191 * symbol-summary.h: Likewise.
1192 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1194 2018-06-19 Richard Biener <rguenther@suse.de>
1196 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1197 (vectorize_loops): ... here. Fix dbgcnt handling.
1198 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1200 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1203 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1204 ieee128 argument takes up only one (vector) register, not two (floating
1207 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1209 * gimplify.c (gimplify_init_constructor): Really never clear for an
1210 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1212 2018-06-19 Richard Biener <rguenther@suse.de>
1214 PR tree-optimization/86179
1215 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1216 after failed recognition.
1218 2018-06-18 Martin Sebor <msebor@redhat.com>
1221 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1222 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1223 Handle integer subtraction.
1224 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1225 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1227 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1229 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1230 param from rtx to rtx_insn *.
1231 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1233 (frv_ifcvt_modify_insn): Likwise.
1234 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1235 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1236 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1237 as_a <rtx_insn *> cast to local "unprotected_region" once
1238 it's been established that it's not NULL or pc_rtx.
1239 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1240 param "sethi" from rtx to rtx_insn *.
1241 (nds32_group_float_insns): Likewise for param "insn".
1242 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1244 (vax_output_int_subtract): Likewise.
1245 * config/vax/vax.c (vax_output_int_add): Likewise for param
1247 (vax_output_int_subtract): Likewise.
1248 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1249 (emit_pattern_after): Likewise for param "after".
1250 (emit_insn_after): Likewise.
1251 (emit_jump_insn_after): Likewise.
1252 (emit_call_insn_after): Likewise.
1253 (emit_debug_insn_after): Likewise.
1254 (emit_pattern_before): Likewise for param "before".
1255 (emit_insn_before): Likewise.
1256 (emit_jump_insn_before): Likewise.
1257 * final.c (get_insn_template): Likewise for param "insn", removing
1259 * output.h (get_insn_template): Likewise for 2nd param.
1260 * rtl.h (emit_insn_before): Likewise.
1261 (emit_jump_insn_before): Likewise.
1262 (emit_debug_insn_before_noloc): Likewise.
1263 (emit_insn_after): Likewise.
1264 (emit_jump_insn_after): Likewise.
1265 (emit_call_insn_after): Likewise.
1266 (emit_debug_insn_after): Likewise.
1267 (set_insn_deleted): Likewise for param.
1269 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1272 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1273 floating point modes, so that IFmode is numerically greater than
1274 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1275 to declare the ordering. This prevents IFmode from being
1276 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1277 machine. Include rs6000-modes.h to share the fractional values
1278 between genmodes* and the rest of the compiler.
1282 * config/rs6000/rs6000-modes.h: New file.
1283 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1284 meaning of rs6000_long_double_size so that 126..128 selects an
1285 appropriate 128-bit floating point type.
1286 (rs6000_option_override_internal): Likewise.
1287 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1288 (TARGET_LONG_DOUBLE_128): Change the meaning of
1289 rs6000_long_double_size so that 126..128 selects an appropriate
1290 128-bit floating point type.
1291 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1292 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1293 source and destination to match the standard usage.
1294 (truncifkf2): Likewise.
1295 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1296 ISA 2.07 to use an explicit clobber, instead of passing in a
1298 (copysign<mode>3_soft): Likewise.
1300 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1302 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1303 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1304 (vect_slp_analyze_instance_dependence): Likewise.
1305 (vect_enhance_data_refs_alignment): Likewise.
1306 (vect_analyze_data_refs_alignment): Likewise.
1307 (vect_slp_analyze_and_verify_instance_alignment
1308 (vect_analyze_data_ref_accesses): Likewise.
1309 (vect_prune_runtime_alias_test_list): Likewise.
1310 (vect_analyze_data_refs): Likewise.
1311 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1312 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1313 (vect_analyze_scalar_cycles_1): Likewise.
1314 (vect_get_loop_niters): Likewise.
1315 (vect_analyze_loop_form_1): Likewise.
1316 (vect_update_vf_for_slp): Likewise.
1317 (vect_analyze_loop_operations): Likewise.
1318 (vect_analyze_loop): Likewise.
1319 (vectorizable_induction): Likewise.
1320 (vect_transform_loop): Likewise.
1321 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1322 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1323 (vect_make_slp_decision): Likewise.
1324 (vect_detect_hybrid_slp): Likewise.
1325 (vect_slp_analyze_operations): Likewise.
1326 (vect_slp_bb): Likewise.
1327 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1328 (vectorizable_bswap): Likewise.
1329 (vectorizable_call): Likewise.
1330 (vectorizable_simd_clone_call): Likewise.
1331 (vectorizable_conversion): Likewise.
1332 (vectorizable_assignment): Likewise.
1333 (vectorizable_shift): Likewise.
1334 (vectorizable_operation): Likewise.
1335 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1337 2018-06-18 Martin Sebor <msebor@redhat.com>
1339 PR tree-optimization/81384
1340 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1341 * builtins.c (expand_builtin_strnlen): New function.
1342 (expand_builtin): Call it.
1343 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1344 * builtins.def (BUILT_IN_STRNLEN): New.
1345 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1346 Warn for bounds in excess of maximum object size.
1347 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1348 single-value ranges. Handle strnlen.
1349 (handle_builtin_strlen): Handle strnlen.
1350 (strlen_check_and_optimize_stmt): Same.
1351 * doc/extend.texi (Other Builtins): Document strnlen.
1353 2018-06-18 Maya Rashish <coypu@sdf.org>
1355 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1356 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1357 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1359 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1361 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1363 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1365 * tree.c (escaped_string::escape): Replace cast to char * by
1366 const_cast<char *> (unescaped).
1368 2018-06-18 Nick Clifton <nickc@redhat.com>
1371 * tree.c (escaped_string): New class. Converts an unescaped
1372 string into its escaped equivalent.
1373 (warn_deprecated_use): Use the new class to convert the
1374 deprecation message, if present.
1375 (test_escaped_strings): New self test.
1376 (test_c_tests): Add test_escaped_strings.
1377 * doc/extend.texi (deprecated): Add a note that the
1378 deprecation message is affected by the -fmessage-length
1379 option, and that control characters will be escaped.
1380 (#pragma GCC error): Document this pragma.
1381 (#pragma GCC warning): Likewise.
1382 * doc/invoke.texi (-fmessage-length): Document this option's
1383 effect on the #warning and #error preprocessor directives and
1384 the deprecated attribute.
1386 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1388 * tree.c (decl_value_expr_lookup): Revert latest change.
1389 (decl_value_expr_insert): Likewise.
1391 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1393 * gimplify.c (nonlocal_vlas): Delete.
1394 (nonlocal_vla_vars): Likewise.
1395 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1397 (gimplify_body): Do not create and destroy nonlocal_vlas.
1398 * tree-nested.c: Include diagnostic.h.
1399 (use_pointer_in_frame): Tweak.
1400 (lookup_field_for_decl): Add assertion and declare the transformation.
1401 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1402 internal error when the reference is in a wrong context. Do not
1403 create a debug decl by default.
1404 (note_nonlocal_block_vlas): Delete.
1405 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1406 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1407 create a debug decl by default.
1408 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1409 call is in a wrong context.
1410 (fixup_vla_decls): New function.
1411 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1412 debug variables were created.
1413 * tree.c (decl_value_expr_lookup): Add checking assertion.
1414 (decl_value_expr_insert): Likewise.
1416 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1419 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1420 * tree-scalar-evolution.c (interpret_expr): Likewise.
1421 (expression_expensive_p): Likewise.
1422 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1423 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1424 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1425 (ssa_defined_by_minus_one_stmt_p): New.
1427 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1430 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1431 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1432 * dojump.c (do_jump): Likewise.
1433 * expr.c (expand_expr_real_2): Check operand type's sign.
1434 * fold-const.c (const_unop): Handle ABSU_EXPR.
1435 (fold_abs_const): Likewise.
1436 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1437 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1438 (strip_sign_op_1): Likesise.
1439 * match.pd: Add new pattern to generate ABSU_EXPR.
1440 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1441 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1442 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1443 * tree-inline.c (estimate_operator_cost): Likewise.
1444 * tree-pretty-print.c (dump_generic_node): Likewise.
1445 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1446 * tree.def (ABSU_EXPR): New.
1448 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1451 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1452 documented as preserved for backward compatibility only.
1453 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1455 PR rtl-optimization/86108
1456 * bb-reorder.c (create_forwarder_block): Renamed to ...
1457 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1458 jump from new landing pad to the second part.
1459 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1462 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1465 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1466 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1467 Only call store_expr for halves if the mode is the same.
1470 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1471 Fix up comment formatting.
1473 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1475 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1476 (typed_splay_tree::closure,
1477 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1478 (typed_splay_tree::typed_splay_tree,
1479 typed_splay_tree::operator =): Declared private.
1480 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1481 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1482 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1483 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1484 typed_splay_tree::splay_tree_splay,
1485 typed_splay_tree::splay_tree_foreach_helper,
1486 typed_splay_tree::splay_tree_insert,
1487 typed_splay_tree::splay_tree_remove,
1488 typed_splay_tree::splay_tree_lookup,
1489 typed_splay_tree::splay_tree_predecessor,
1490 typed_splay_tree::splay_tree_successor,
1491 typed_splay_tree::splay_tree_min,
1492 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1493 (typed_splay_tree::root, typed_splay_tree::comp,
1494 typed_splay_tree::delete_key,
1495 typed_splay_tree::delete_value): New data members.
1496 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1497 typed_splay_tree::remove.
1499 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1501 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1502 -mginv and -mno-ginv to the assembler.
1503 * config/mips/mips.opt (-mcrc): New option.
1505 * doc/invoke.text (-mcrc): Document.
1508 2018-06-15 Nick Clifton <nickc@redhat.com>
1511 * tree.c (escaped_string): New class. Converts an unescaped
1512 string into its escaped equivalent.
1513 (warn_deprecated_use): Use the new class to convert the
1514 deprecation message, if present.
1515 (test_escaped_strings): New self test.
1516 (test_c_tests): Add test_escaped_strings.
1517 * doc/extend.texi (deprecated): Add a note that the
1518 deprecation message is affected by the -fmessage-length
1519 option, and that control characters will be escaped.
1520 (#pragma GCC error): Document this pragma.
1521 (#pragma GCC warning): Likewise.
1522 * doc/invoke.texi (-fmessage-length): Document this option's
1523 effect on the #warning and #error preprocessor directives and
1524 the deprecated attribute.
1526 2018-06-15 Richard Biener <rguenther@suse.de>
1528 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1529 here, also noting vector size used.
1530 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1531 size used in MSG_OPTIMIZED_LOCATIONS dump.
1532 (pass_slp_vectorize::execute): Adjust.
1534 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1537 * config/arc/arc.c (arc_return_address_register): Fix
1540 2018-06-15 Richard Biener <rguenther@suse.de>
1543 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1544 leave useless conversion stripping to force_gimple_operand_gsi.
1545 (gimplify_build2): Likewise.
1546 (gimplify_build1): Likewise.
1548 2018-06-15 Richard Biener <rguenther@suse.de>
1551 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1552 before adjusting their block.
1554 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1556 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1557 multilibs for *-*-rtems*.
1558 * config/riscv/t-rtems: New file.
1560 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1563 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1564 unsigned_type_for returns NULL.
1567 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1568 subregs of multi-word pseudos unless the float mode has word size.
1570 2018-06-14 Richard Biener <rguenther@suse.de>
1573 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1574 duplicate and harmful type_hash_canon.
1575 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1577 2018-06-14 Richard Biener <rguenther@suse.de>
1580 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1583 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1585 * config/rtems.h (STDINT_LONG32): Define.
1587 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1588 Prachi Godbole <prachi.godbole@imgtec.com>
1590 * config/mips/mips-cpus.def: Define P6600.
1591 * config/mips/mips-tables.opt: Regenerate.
1592 * config/mips/mips.c (mips_ucbranch_type): New enum.
1593 (mips_rtx_cost_data): Add support for P6600.
1594 (mips_issue_rate): Likewise.
1595 (mips_multipass_dfa_lookahead): Likewise.
1596 (mips_avoid_hazard): Likewise.
1597 (mips_reorg_process_insns): Likewise.
1598 (mips_classify_branch_p6600): New function.
1599 * config/mips/mips.h (TUNE_P6600): New define.
1600 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1601 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1602 * config/mips/mips.md: Include p6600.md.
1603 (processor): Add p6600.
1604 * config/mips/p6600.md: New file.
1605 * doc/invoke.texi: Add p6600 to supported architectures.
1607 2018-06-13 Martin Sebor <msebor@redhat.com>
1609 PR tree-optimization/86114
1610 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1612 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1614 2018-06-13 Richard Biener <rguenther@suse.de>
1616 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1617 Properly set vector type of the intermediate stmt.
1618 * tree-vect-stmts.c (vectorizable_operation): The destination
1619 var always has vectype_out type.
1621 2018-06-13 Jeff Law <law@redhat.com>
1623 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1624 integer 0 for argument to print_rtl_with_bb.
1625 (rl78_reorg): Likewise.
1627 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1629 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1630 from rtx to rtx_insn *.
1631 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1633 (add_sched_insns_for_speculation): Likewise for local "target",
1634 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1635 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1636 from rtx_insn ** to rtx_code_label **.
1637 (reorg_emit_nops): Likewise.
1638 (c6x_reorg): Likewise for local "call_labels".
1639 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1641 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1642 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1643 the loops over LABEL_REFS.
1644 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1646 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1647 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1648 (split_branches): Strengthen local "olabel" from rtx to
1649 rtx_insn *, adding a safe_as_a cast.
1650 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1652 (add_insn_after): Likewise for first two params.
1653 (add_insn_before): Likewise.
1654 (remove_insn): Likewise for param.
1655 (emit_pattern_before_noloc): Likewise for second and third params.
1656 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1657 (emit_call_insn_before_noloc): Likewise.
1658 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1660 (emit_barrier_before): Likewise.
1661 (emit_label_before): Strengthen "label" param from "rtx" to
1662 "rtx_code_label *". Strengthen "before" param from "rtx" to
1664 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1666 (emit_pattern_after_noloc): Likewise.
1667 (emit_insn_after_noloc): Likewise.
1668 (emit_jump_insn_after_noloc): Likewise.
1669 (emit_call_insn_after_noloc): Likewise.
1670 (emit_debug_insn_after_noloc): Likewise.
1671 (emit_barrier_after): Likewise.
1672 (emit_label_after): Likewise for both params.
1673 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1674 "loc" param from "int" to "location_t".
1675 (emit_insn_after_setloc): Likewise.
1676 (emit_jump_insn_after_setloc): Likewise.
1677 (emit_call_insn_after_setloc): Likewise.
1678 (emit_debug_insn_after_setloc): Likewise.
1679 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1680 "loc" param from "int" to "location_t".
1681 (emit_pattern_before): Convert NULL_RTX to NULL.
1682 (emit_insn_before_setloc): Convert "loc" param from "int" to
1684 (emit_jump_insn_before_setloc): Likewise.
1685 (emit_call_insn_before_setloc): Likewise.
1686 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1687 rtx_insn *. Convert "loc" param from "int" to "location_t".
1688 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1689 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1690 Convert 3rd param from "int" to "location_t".
1691 (emit_barrier_before, emit_barrier_after, next_real_insn):
1692 Strengthen param from rtx to rtx_insn *.
1693 (emit_label_before): Strengthen 1st param from "rtx" to
1694 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1696 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1697 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1698 Strengthen 2nd param from "rtx" to "rtx_insn *".
1699 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1700 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1701 Likewise. Convert 3rd param from "int" to "location_t".
1702 (emit_label_after): Strengthen 1st param from "rtx" to
1704 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1706 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1707 from "rtx" to "rtx_insn *".
1709 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1711 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1712 bodies streamed in with -Q.
1713 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1714 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1715 * gimple-streamer-out.c: Include gimple-pretty-print.h
1716 (output_bb): Dump stmts streamed.
1717 * lto-section-out.c: Include print-tree.h
1718 (lto_begin_section): Dump sections created.
1719 (lto_output_decl_index): Dump decl encoded.
1720 * lto-streamer-out.c: Include print-tree.h
1721 (create_output_block): Dump output block created.
1722 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1723 (output_function): Dump function output.
1724 (output_constructor): Dump constructor streamed.
1725 (write_global_stream): Output indexes encoded.
1726 (produce_asm_for_decls): Dump streams encoded.
1727 * lto-streamer.c (streamer_dump_file): New global var.
1728 * lto-streamer.h (streamer_dump_file): Declare.
1729 * passes.c (ipa_write_summaries): Initialize streamer dump.
1730 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1733 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1736 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1737 offsets for register save directives. Emit a second batch of save
1738 directives, if need be, when the function accesses prior frames.
1740 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1742 * config/arc/fpu.md (fmasf4): Force operand to register.
1743 (fnmasf4): Likewise.
1745 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1747 * config/arc/arc-protos.h (arc_pad_return): Remove.
1748 * config/arc/arc.c (machine_function): Remove force_short_suffix
1750 (arc_print_operand): Adjust printing of '&'.
1751 (arc_verify_short): Remove conditional printing of short suffix.
1752 (arc_final_prescan_insn): Remove reference to size_reason.
1753 (pad_return): New function.
1754 (arc_reorg): Call pad_return.
1755 (arc_pad_return): Remove.
1756 (arc_init_machine_status): Remove reference to force_short_suffix.
1757 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1758 (attr length): When attribute iscompact is true force to 2
1759 regardless; in the case of maybe check if we want to force the
1760 instruction to have 4 bytes length.
1761 (nopv): Change it to generate 4 byte long nop as well.
1762 (blockage): New pattern.
1763 (simple_return): Remove call to arc_pad_return.
1764 (p_return_i): Likewise.
1766 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1768 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1770 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1772 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1775 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1777 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1778 for ARC700 and ARCv2.
1780 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1783 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1784 operands[2] instead of operands[1].
1787 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1789 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1790 case, check whether the outer register overlaps an unallocatable
1791 register, not just whether it fits the required class.
1793 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1795 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1797 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1798 (memrefs_conflict_p): Likewise.
1799 (init_alias_analysis): Likewise.
1800 * cfgexpand.c (expand_debug_expr): Likewise.
1801 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1802 * cse.c (fold_rtx): Likewise.
1803 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1804 * expr.c (emit_block_move_hints): Likewise.
1805 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1806 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1807 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1808 (emit_group_store): Likewise.
1809 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1810 to read the PRE/POST_MODIFY increment.
1811 * calls.c (store_one_arg): Use strip_offset.
1812 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1814 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1816 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1817 (simplify_binary_operation_1): Extend CONST_INT handling to
1819 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1820 than a HOST_WIDE_INT.
1821 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1823 (adjust_mems, add_stores): Update accodingly.
1824 (vt_canonicalize_addr): Track polynomial offsets.
1825 (emit_note_insn_var_location): Likewise.
1826 (vt_add_function_parameter): Likewise.
1827 (vt_initialize): Likewise.
1829 2018-06-12 Jeff Law <law@redhat.com>
1831 * config.gcc (alpha*-*-freebsd*): Remove.
1832 * config/alpha/freebsd.h: Remove.
1834 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1837 * spellcheck-tree.c (levenshtein_distance): Rename to...
1838 (get_edit_distance): ...this, and update for underlying renaming.
1839 * spellcheck-tree.h (levenshtein_distance): Rename to...
1840 (get_edit_distance): ...this.
1841 * spellcheck.c (levenshtein_distance): Rename to...
1842 (get_edit_distance): ...this. Convert from Levenshtein distance
1843 to Damerau-Levenshtein distance by supporting transpositions of
1844 adjacent characters. Rename "v1" to "v_next" and "v0" to
1846 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1847 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1848 update for underlying renaming.
1849 (selftest::levenshtein_distance_unit_test): Rename to...
1850 (selftest::test_get_edit_distance_unit): ...this, and update for
1851 underlying renaming.
1852 (selftest::test_find_closest_string): Add example from PR 69968
1853 where transposition helps
1854 (selftest::test_metric_conditions): Update for renaming.
1855 (selftest::test_metric_conditions): Likewise.
1856 (selftest::spellcheck_c_tests): Likewise.
1857 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1858 (get_edit_distance): ...this.
1859 (best_match::consider): Update for renaming.
1861 2018-06-12 Martin Sebor <msebor@redhat.com>
1863 PR tree-optimization/85259
1864 * builtins.c (compute_objsize): Handle constant offsets.
1865 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1866 true iff a warning has been issued.
1867 * gimple.h (gimple_nonartificial_location): New function.
1868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1869 gimple_nonartificial_location and handle -Wno-system-headers.
1870 (handle_builtin_stxncpy): Same.
1872 2018-06-12 Martin Sebor <msebor@redhat.com>
1875 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1877 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1879 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1880 BUILTIN_VEC_XST entries for pointer to double and long long.
1882 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1885 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1887 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1890 2018-06-12 Martin Liska <mliska@suse.cz>
1892 * doc/options.texi: Document IntegerRange.
1894 2018-06-12 Martin Liska <mliska@suse.cz>
1896 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1897 * opt-functions.awk: Handle Deprecated flag.
1898 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1900 (read_cmdline_option): Report warning for a deprecated option.
1901 * opts.h (struct cl_option): Add new field cl_deprecated.
1902 (CL_ERR_DEPRECATED): New.
1904 2018-06-12 Martin Liska <mliska@suse.cz>
1906 * doc/options.texi: Document Deprecated option flag.
1908 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1910 * config/arc/arc-arch.h (arc_extras): New enum.
1911 (arc_cpu_t):Add field extra.
1912 (arc_cpu_types): Consider the extras.
1913 * config/arc/arc-cpus.def: Add extras info.
1914 * config/arc/arc-opts.h (processor_type): Consider extra field.
1915 * config/arc/arc.c (arc_override_options): Handle extra field.
1917 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1919 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1920 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1922 * config/arc/arc.c (arc_init): Update pic warning.
1923 (irq_range): Update irq range parsing warnings.
1924 (arc_override_options): Update various warning messages.
1925 (arc_handle_aux_attribute): Likewise.
1927 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1929 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1931 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1933 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1934 compiler messages directives.
1936 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1938 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1939 * config/mips/mips-tables.opt: Regenerate.
1940 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1942 * doc/invoke.texi: Document -march=i6500.
1944 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1946 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1947 (i6400_gpmul): Add cpu_unit.
1948 (i6400_gpdiv): Likewise.
1949 (i6400_msa_add_d): Update reservations.
1950 (i6400_msa_int_add) Likewise.
1951 (i6400_msa_short_logic3) Likewise.
1952 (i6400_msa_short_logic2) Likewise.
1953 (i6400_msa_short_logic) Likewise.
1954 (i6400_msa_move) Likewise.
1955 (i6400_msa_cmp) Likewise.
1956 (i6400_msa_short_float2) Likewise.
1957 (i6400_msa_div_d) Likewise.
1958 (i6400_msa_long_logic1) Likewise.
1959 (i6400_msa_long_logic2) Likewise.
1960 (i6400_msa_mult) Likewise.
1961 (i6400_msa_long_float2) Likewise.
1962 (i6400_msa_long_float4) Likewise.
1963 (i6400_msa_long_float5) Likewise.
1964 (i6400_msa_long_float8) Likewise.
1965 (i6400_fpu_fadd): Include frint type.
1966 (i6400_fpu_store): New define_insn_reservation.
1967 (i6400_fpu_load): Likewise.
1968 (i6400_fpu_move): Likewise.
1969 (i6400_fpu_fcmp): Likewise.
1970 (i6400_fpu_fmadd): Likewise.
1971 (i6400_int_mult): Include imul3nc type and update reservation.
1972 (i6400_int_div): Include idiv3 type and update reservation.
1973 (i6400_int_load): Update to check type not move_type.
1974 (i6400_int_store): Likewise.
1975 (i6400_int_prefetch): Set zero latency.
1977 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1979 * gcc.c: Document new %@{...} sequence.
1980 (LINK_COMMAND_SPEC): Use it for the -L switches.
1981 (cpp_unique_options): Use it for the -I switches.
1982 (at_file_argbuf): New global variable.
1983 (in_at_file): Likewise.
1984 (alloc_args): Create at_file_argbuf.
1985 (clear_args): Truncate at_file_argbuf.
1986 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1987 (open_at_file): New function.
1988 (close_at_file): Likewise.
1989 (create_at_file): Delete.
1990 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1993 (validate_switches_from_spec): Deal with %@{...} sequence.
1994 (validate_switches): Likewise.
1995 (driver::finalize): Call clear_args.
1997 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1999 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2001 2018-06-11 Martin Sebor <msebor@redhat.com>
2003 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2004 (Wno-class-memaccess): Add @opindex.
2005 (Wno-templates, Wno-multiple-inheritance): Same.
2006 (Wno-virtual-inheritance, Wno-namespaces): Same.
2007 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2008 (Wno-format-overflow, Wno-format-truncation): Same.
2009 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2010 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2011 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2012 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2013 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2014 (Wno-misspelled-isr): Same.
2016 2018-06-11 Martin Sebor <msebor@redhat.com>
2018 * PR tree-optimization/86083
2019 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2021 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2023 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2025 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2028 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2029 on the correct operand.
2030 (*movdi_internal64): Ditto.
2032 2018-06-11 Martin Liska <mliska@suse.cz>
2034 PR tree-optimization/86089
2035 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2037 2018-06-11 Julia Koval <julia.koval@intel.com>
2039 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2040 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2041 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2043 2018-06-11 Olivier Hainque <hainque@adacore.com>
2045 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2046 for Ada with strict dwarf2.
2048 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2051 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2054 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2056 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2058 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2060 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2063 2018-06-08 Martin Liska <mliska@suse.cz>
2065 * tree-cfg.h (debug_function): Fix argument type to match
2068 2018-06-08 Martin Liska <mliska@suse.cz>
2070 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2071 Remove usage of MPX-related (and removed) fields.
2072 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2074 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2076 * cfg.c (debug): Use TDF_NONE rather than 0.
2077 * cfghooks.c (debug): Likewise.
2078 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2079 (struct dump_option_value_info): Convert to...
2080 (struct kv_pair): ...this template type.
2081 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2083 (optinfo_verbosity_options): Likewise.
2084 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2086 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2087 than int for "optgroup_flags" param.
2088 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2090 (dump_dec): Likewise.
2091 (dump_finish): Use TDF_NONE rather than 0.
2092 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2093 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2094 than 0. Update for change to option_ptr.
2095 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2096 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2097 0. Update for changes to optinfo_verbosity_options and
2099 (opt_info_switch_p): Convert optgroup_flags from int to
2101 (dump_basic_block): Use dump_flags_t rather than int
2102 for "dump_kind" param.
2103 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2104 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2105 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2106 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2107 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2108 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2109 TDF_NONE): Convert from macros to...
2110 (enum dump_flag): ...this new enum.
2111 (dump_flags_t): Update to use enum.
2112 (operator|, operator&, operator~, operator|=, operator&=):
2113 Implement for dump_flags_t.
2114 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2115 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2116 Convert from macros to...
2117 (enum optgroup_flag): ...this new enum.
2118 (optgroup_flags_t): New typedef.
2119 (operator|, operator|=): Implement for optgroup_flags_t.
2120 (struct dump_file_info): Convert field "alt_flags" to
2121 dump_flags_t. Convert field "optgroup_flags" to
2123 (dump_basic_block): Use dump_flags_t rather than int for param.
2124 (dump_generic_expr_loc): Likewise.
2125 (dump_dec): Likewise.
2126 (dump_register): Convert param "optgroup_flags" to
2128 (opt_info_enable_passes): Likewise.
2129 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2131 * gimple-pretty-print.c (debug): Likewise.
2132 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2133 (merged_store_group::apply_stores): Likewise.
2134 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2135 * gimple.c (verify_gimple_pp): Likewise.
2136 * graphite-poly.c (print_pbb_body): Likewise.
2137 * passes.c (pass_manager::register_one_dump_file): Convert
2138 local "optgroup_flags" to optgroup_flags_t.
2139 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2141 (debug_body): Likewise.
2142 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2143 to optgroup_flags_t.
2144 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2146 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2147 (convert_mult_to_fma): Likewise.
2148 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2149 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2150 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2151 "dump_kind" to dump_flags_t.
2153 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2155 * config/rs6000/rs6000.c (min, max): Delete.
2157 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2159 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2160 -mabi=spe and -mabi=no-spe.
2162 2018-06-08 Martin Liska <mliska@suse.cz>
2164 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2165 where we expect an existing summary.
2167 2018-06-08 Martin Liska <mliska@suse.cz>
2169 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2170 * ipa-inline.h (estimate_edge_growth): Likewise.
2172 2018-06-08 Martin Liska <mliska@suse.cz>
2174 * cgraph.c (function_version_hasher::hash): Use
2175 cgraph_node::get_uid ().
2176 (function_version_hasher::equal):
2177 * cgraph.h (cgraph_node::get_uid): New method.
2178 * ipa-inline.c (update_caller_keys): Use
2179 cgraph_node::get_uid ().
2180 (update_callee_keys): Likewise.
2181 * ipa-utils.c (searchc): Likewise.
2182 (ipa_reduced_postorder): Likewise.
2183 * lto-cgraph.c (input_node): Likewise.
2184 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2185 * symbol-summary.h (symtab_insertion): Likewise.
2186 (symtab_removal): Likewise.
2187 (symtab_duplication): Likewise.
2188 * tree-pretty-print.c (dump_function_header): Likewise.
2189 * tree-sra.c (convert_callers_for_node): Likewise.
2191 2018-06-08 Martin Liska <mliska@suse.cz>
2193 * cgraph.c (symbol_table::create_edge): Always assign a new
2195 (symbol_table::free_edge): Do not recycle numbers.
2196 * cgraph.h (cgraph_edge::get): New method.
2197 * symbol-summary.h (symtab_removal): Use it.
2198 (symtab_duplication): Likewise.
2199 (call_summary::hashable_uid): Remove.
2201 2018-06-08 Martin Liska <mliska@suse.cz>
2203 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2204 (initialize_growth_caches): Remove.
2205 (free_growth_caches): Likewise.
2206 (do_estimate_edge_time): Use edge_growth_cache.
2207 (do_estimate_edge_size): Likewise.
2208 (do_estimate_edge_hints): Likewise.
2209 * ipa-inline.c (reset_edge_caches): Likewise.
2210 (recursive_inlining): Likewise.
2211 (inline_small_functions): Likewise.
2212 * ipa-inline.h (initialize_growth_caches): Remove.
2213 (estimate_edge_size): Likewise.
2214 (estimate_edge_time): Likewise.
2215 (estimate_edge_hints): Likewise.
2216 (reset_edge_growth_cache): Likewise.
2217 * symbol-summary.h (call_summary::remove): New method.
2219 2018-06-08 Martin Liska <mliska@suse.cz>
2221 * ipa-cp.c (class edge_clone_summary): New summary.
2222 (grow_edge_clone_vectors): Remove.
2223 (ipcp_edge_duplication_hook): Remove.
2224 (class edge_clone_summary_t): New call_summary class.
2225 (ipcp_edge_removal_hook): Remove.
2226 (edge_clone_summary_t::duplicate): New function.
2227 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2228 (create_specialized_node): Likewise.
2229 (ipcp_driver): Initialize edge_clone_summaries and do not
2232 2018-06-08 Martin Liska <mliska@suse.cz>
2234 * symbol-summary.h (get): New function.
2235 (call_summary::m_initialize_when_cloning): New class member.
2237 2018-06-08 Martin Liska <mliska@suse.cz>
2239 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2240 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2241 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2242 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2243 (remove_cgraph_node_from_order): Use the removed_nodes set.
2244 (do_per_function_toporder): Likwise.
2245 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2246 instead of summary_uid.
2247 (symtab_removal): Likewise.
2248 (symtab_duplication): Likewise.
2250 2018-06-08 Martin Liska <mliska@suse.cz>
2252 * ipa-cp.c (ipcp_store_bits_results): Use
2253 ipcp_transformation_sum.
2254 (ipcp_store_vr_results): Likewise.
2255 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2257 (ipcp_transformation_initialize): ... this.
2258 (ipa_set_node_agg_value_chain):
2259 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2260 (write_ipcp_transformation_info): Likewise.
2261 (read_ipcp_transformation_info): Likewise.
2262 (ipcp_update_bits): Likewise.
2263 (ipcp_update_vr): Likewise.
2264 (ipcp_transform_function): Likewise.
2265 * ipa-prop.h: Rename ipcp_transformation_summary to
2266 ipcp_transformation.
2267 (class ipcp_transformation_t): New function summary.
2268 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2269 (ipa_get_agg_replacements_for_node): Likewise.
2271 2018-06-08 Martin Liska <mliska@suse.cz>
2273 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2275 (class funct_state_summary_t): New function_summary class.
2276 (has_function_state): Remove.
2277 (get_function_state): Likewise.
2278 (set_function_state): Likewise.
2279 (add_new_function): Likewise.
2280 (funct_state_summary_t::insert): New function.
2281 (duplicate_node_data): Remove.
2282 (remove_node_data): Remove.
2283 (funct_state_summary_t::duplicate): New function.
2284 (register_hooks): Create new funct_state_summaries.
2285 (pure_const_generate_summary): Use it.
2286 (pure_const_write_summary): Likewise.
2287 (pure_const_read_summary): Likewise.
2288 (propagate_pure_const): Likewise.
2289 (propagate_nothrow): Likewise.
2290 (dump_malloc_lattice): Likewise.
2291 (propagate_malloc): Likewise.
2292 (execute): Do not register hooks, just remove summary
2294 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2297 2018-06-08 Martin Liska <mliska@suse.cz>
2299 * ipa-reference.c (remove_node_data): Remove.
2300 (duplicate_node_data): Likewise.
2301 (class ipa_ref_var_info_summary_t): New class.
2302 (class ipa_ref_opt_summary_t): Likewise.
2303 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2304 (get_reference_optimization_summary): Use
2305 ipa_ref_opt_sum_summaries.
2306 (set_reference_vars_info): Remove.
2307 (set_reference_optimization_summary): Likewise.
2308 (ipa_init): Create summaries.
2309 (init_function_info): Use function summary.
2310 (ipa_ref_opt_summary_t::duplicate): New function.
2311 (ipa_ref_opt_summary_t::remove): New function.
2312 (get_read_write_all_from_node): Fix GNU coding style.
2313 (propagate): Use function summary.
2314 (write_node_summary_p): Fix GNU coding style.
2315 (stream_out_bitmap): Likewise.
2316 (ipa_reference_read_optimization_summary): Use function summary.
2317 (ipa_reference_c_finalize): Do not release hooks.
2319 2018-06-08 Martin Liska <mliska@suse.cz>
2321 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2322 (analyze_function_body): Extract multiple calls of get_create.
2323 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2324 * ipa-inline.c (recursive_inlining): Use ::get method.
2325 * ipa-inline.h (estimate_edge_growth): Likewise.
2327 2018-06-08 Martin Liska <mliska@suse.cz>
2329 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2331 (hsa_function_summary::hsa_function_summary): Use the new enum
2333 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2334 * hsa-gen.c (hsa_get_host_function): Likewise.
2335 (get_brig_function_name): Likewise.
2336 * ipa-hsa.c (process_hsa_functions): Likewise.
2337 (ipa_hsa_write_summary): Likewise.
2338 * symbol-summary.h (symtab_duplication): Use ::get function/
2339 (get): New function.
2341 2018-06-08 Martin Liska <mliska@suse.cz>
2343 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2345 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2346 (hsa_register_kernel): Likewise.
2347 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2348 * hsa-gen.c (hsa_get_host_function): Likewise.
2349 (get_brig_function_name): Likewise.
2350 (generate_hsa): Likewise.
2351 (pass_gen_hsail::execute): Likewise.
2352 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2353 (devirtualization_time_bonus): Likewise.
2354 (ipcp_propagate_stage): Likewise.
2355 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2356 (edge_set_predicate): Likewise.
2357 (evaluate_conditions_for_known_args): Likewise.
2358 (evaluate_properties_for_edge): Likewise.
2359 (ipa_fn_summary::reset): Likewise.
2360 (ipa_fn_summary_t::duplicate): Likewise.
2361 (dump_ipa_call_summary): Likewise.
2362 (ipa_dump_fn_summary): Likewise.
2363 (analyze_function_body): Likewise.
2364 (compute_fn_summary): Likewise.
2365 (estimate_edge_devirt_benefit): Likewise.
2366 (estimate_edge_size_and_time): Likewise.
2367 (estimate_calls_size_and_time): Likewise.
2368 (estimate_node_size_and_time): Likewise.
2369 (inline_update_callee_summaries): Likewise.
2370 (remap_edge_change_prob): Likewise.
2371 (remap_edge_summaries): Likewise.
2372 (ipa_merge_fn_summary_after_inlining): Likewise.
2373 (ipa_update_overall_fn_summary): Likewise.
2374 (read_ipa_call_summary): Likewise.
2375 (inline_read_section): Likewise.
2376 (write_ipa_call_summary): Likewise.
2377 (ipa_fn_summary_write): Likewise.
2378 (ipa_free_fn_summary): Likewise.
2379 * ipa-hsa.c (process_hsa_functions): Likewise.
2380 (ipa_hsa_write_summary): Likewise.
2381 (ipa_hsa_read_section): Likewise.
2382 * ipa-icf.c (sem_function::merge): Likewise.
2383 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2384 (do_estimate_edge_time): Likewise.
2385 (estimate_size_after_inlining): Likewise.
2386 (estimate_growth): Likewise.
2387 (growth_likely_positive): Likewise.
2388 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2389 (inline_call): Likewise.
2390 * ipa-inline.c (caller_growth_limits): Likewise.
2391 (can_inline_edge_p): Likewise.
2392 (can_inline_edge_by_limits_p): Likewise.
2393 (compute_uninlined_call_time): Likewise.
2394 (compute_inlined_call_time): Likewise.
2395 (want_inline_small_function_p): Likewise.
2396 (edge_badness): Likewise.
2397 (update_caller_keys): Likewise.
2398 (update_callee_keys): Likewise.
2399 (recursive_inlining): Likewise.
2400 (inline_small_functions): Likewise.
2401 (inline_to_all_callers_1): Likewise.
2402 (dump_overall_stats): Likewise.
2403 (early_inline_small_functions): Likewise.
2404 (early_inliner): Likewise.
2405 * ipa-inline.h (estimate_edge_growth): Likewise.
2406 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2407 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2408 * ipa-prop.h (IPA_NODE_REF): Likewise.
2409 (IPA_EDGE_REF): Likewise.
2410 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2411 (propagate_malloc): Likewise.
2412 * ipa-split.c (execute_split_functions): Likewise.
2413 * symbol-summary.h: Rename get to get_create.
2415 (get_create): Likewise.
2416 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2418 2018-06-08 Martin Liska <mliska@suse.cz>
2420 * symbol-summary.h (release): Move definition out of class
2422 (symtab_removal): Likewise.
2423 (symtab_duplication): Likewise.
2425 2018-06-08 Martin Liska <mliska@suse.cz>
2427 * symbol-summary.h (function_summary): Move constructor
2428 implementation out of class declaration.
2429 (release): Likewise.
2430 (symtab_insertion): Likewise.
2431 (symtab_removal): Likewise.
2432 (symtab_duplication): Likewise.
2435 2018-06-08 Martin Liska <mliska@suse.cz>
2437 * Makefile.in: Remove support for MPX (macros, related functions,
2438 fields in cgraph_node, ...).
2439 * builtin-types.def (BT_BND): Likewise.
2440 (BT_FN_BND_CONST_PTR): Likewise.
2441 (BT_FN_CONST_PTR_BND): Likewise.
2442 (BT_FN_VOID_PTR_BND): Likewise.
2443 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2444 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2445 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2446 (expand_builtin_mempcpy_with_bounds): Likewise.
2447 (expand_builtin_memset_with_bounds): Likewise.
2448 (expand_builtin_memset_args): Likewise.
2449 (std_expand_builtin_va_start): Likewise.
2450 (expand_builtin): Likewise.
2451 (expand_builtin_with_bounds): Likewise.
2452 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2453 (DEF_LIB_BUILTIN_CHKP): Likewise.
2454 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2455 (DEF_CHKP_BUILTIN): Likewise.
2456 (BUILT_IN_MEMCPY): Likewise.
2457 (BUILT_IN_MEMMOVE): Likewise.
2458 (BUILT_IN_MEMPCPY): Likewise.
2459 (BUILT_IN_MEMSET): Likewise.
2460 (BUILT_IN_STPCPY): Likewise.
2461 (BUILT_IN_STRCAT): Likewise.
2462 (BUILT_IN_STRCHR): Likewise.
2463 (BUILT_IN_STRCPY): Likewise.
2464 (BUILT_IN_STRLEN): Likewise.
2465 (BUILT_IN_MEMCPY_CHK): Likewise.
2466 (BUILT_IN_MEMMOVE_CHK): Likewise.
2467 (BUILT_IN_MEMPCPY_CHK): Likewise.
2468 (BUILT_IN_MEMSET_CHK): Likewise.
2469 (BUILT_IN_STPCPY_CHK): Likewise.
2470 (BUILT_IN_STRCAT_CHK): Likewise.
2471 (BUILT_IN_STRCPY_CHK): Likewise.
2472 * calls.c (store_bounds): Likewise.
2473 (emit_call_1): Likewise.
2474 (special_function_p): Likewise.
2475 (maybe_warn_nonstring_arg): Likewise.
2476 (initialize_argument_information): Likewise.
2477 (finalize_must_preallocate): Likewise.
2478 (compute_argument_addresses): Likewise.
2479 (expand_call): Likewise.
2480 * cfgexpand.c (expand_call_stmt): Likewise.
2481 (expand_return): Likewise.
2482 (expand_gimple_stmt_1): Likewise.
2483 (pass_expand::execute): Likewise.
2484 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2485 (cgraph_node::remove): Likewise.
2486 (cgraph_node::dump): Likewise.
2487 (cgraph_node::verify_node): Likewise.
2488 * cgraph.h (chkp_function_instrumented_p): Likewise.
2489 (symtab_node::get_alias_target): Likewise.
2490 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2491 (cgraph_local_p): Likewise.
2492 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2493 (cgraph_edge::rebuild_references): Likewise.
2494 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2495 (walk_polymorphic_call_targets): Likewise.
2496 (cgraph_node::expand_thunk): Likewise.
2497 (symbol_table::output_weakrefs): Likewise.
2498 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2499 (ix86_handle_option): Likewise.
2500 * config/i386/constraints.md: Likewise.
2501 * config/i386/i386-builtin-types.def (BND): Likewise.
2505 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2506 (BDESC_FIRST): Likewise.
2508 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2509 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2510 * config/i386/i386.c (enum reg_class): Likewise.
2511 (ix86_target_string): Likewise.
2512 (ix86_option_override_internal): Likewise.
2513 (ix86_conditional_register_usage): Likewise.
2514 (ix86_valid_target_attribute_inner_p): Likewise.
2515 (ix86_set_indirect_branch_type): Likewise.
2516 (ix86_set_current_function): Likewise.
2517 (ix86_function_arg_regno_p): Likewise.
2518 (init_cumulative_args): Likewise.
2519 (ix86_function_arg_advance): Likewise.
2520 (ix86_function_arg): Likewise.
2521 (ix86_pass_by_reference): Likewise.
2522 (ix86_function_value_regno_p): Likewise.
2523 (ix86_function_value_1): Likewise.
2524 (ix86_function_value_bounds): Likewise.
2525 (ix86_return_in_memory): Likewise.
2526 (ix86_setup_incoming_vararg_bounds): Likewise.
2527 (ix86_va_start): Likewise.
2528 (indirect_thunk_need_prefix): Likewise.
2529 (print_reg): Likewise.
2530 (ix86_print_operand): Likewise.
2531 (ix86_expand_call): Likewise.
2532 (ix86_output_function_return): Likewise.
2533 (reg_encoded_number): Likewise.
2534 (BDESC_VERIFYS): Likewise.
2535 (ix86_init_mpx_builtins): Likewise.
2536 (ix86_init_builtins): Likewise.
2537 (ix86_emit_cmove): Likewise.
2538 (ix86_emit_move_max): Likewise.
2539 (ix86_expand_builtin): Likewise.
2540 (ix86_builtin_mpx_function): Likewise.
2541 (ix86_get_arg_address_for_bt): Likewise.
2542 (ix86_load_bounds): Likewise.
2543 (ix86_store_bounds): Likewise.
2544 (ix86_load_returned_bounds): Likewise.
2545 (ix86_store_returned_bounds): Likewise.
2546 (ix86_class_likely_spilled_p): Likewise.
2547 (ix86_hard_regno_mode_ok): Likewise.
2548 (x86_order_regs_for_local_alloc): Likewise.
2549 (ix86_mitigate_rop): Likewise.
2550 (ix86_bnd_prefixed_insn_p): Likewise.
2551 (ix86_mpx_bound_mode): Likewise.
2552 (ix86_make_bounds_constant): Likewise.
2553 (ix86_initialize_bounds): Likewise.
2554 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2555 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2556 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2557 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2558 (TARGET_CHKP_BOUND_MODE): Likewise.
2559 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2560 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2561 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2562 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2563 * config/i386/i386.h (TARGET_MPX): Likewise.
2564 (TARGET_MPX_P): Likewise.
2565 (VALID_BND_REG_MODE): Likewise.
2566 (FIRST_BND_REG): Likewise.
2567 (LAST_BND_REG): Likewise.
2568 (enum reg_class): Likewise.
2569 (BND_REG_P): Likewise.
2570 (BND_REGNO_P): Likewise.
2571 (BNDmode): Likewise.
2572 (ADJUST_INSN_LENGTH): Likewise.
2573 * config/i386/i386.md: Likewise.
2574 * config/i386/i386.opt: Likewise.
2575 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2576 (defined): Likewise.
2577 (LINK_MPX): Likewise.
2578 (MPX_SPEC): Likewise.
2579 (LIBMPX_SPEC): Likewise.
2580 (LIBMPXWRAPPERS_SPEC): Likewise.
2581 (CHKP_SPEC): Likewise.
2582 * config/i386/predicates.md: Likewise.
2583 * dbxout.c (dbxout_type): Likewise.
2584 * doc/extend.texi: Likewise.
2585 * doc/invoke.texi: Likewise.
2586 * doc/md.texi: Likewise.
2587 * doc/tm.texi: Likewise.
2588 * doc/tm.texi.in: Likewise.
2589 * dwarf2out.c (is_base_type): Likewise.
2590 (gen_formal_types_die): Likewise.
2591 (gen_subprogram_die): Likewise.
2592 (gen_type_die_with_usage): Likewise.
2593 (gen_decl_die): Likewise.
2594 (dwarf2out_late_global_decl): Likewise.
2595 * expr.c (expand_assignment): Likewise.
2596 (emit_storent_insn): Likewise.
2597 (store_expr_with_bounds): Likewise.
2598 (store_expr): Likewise.
2599 (expand_expr_real_1): Likewise.
2600 * expr.h (store_expr_with_bounds): Likewise.
2601 * function.c (use_register_for_decl): Likewise.
2602 (struct bounds_parm_data): Likewise.
2603 (assign_parms_augmented_arg_list): Likewise.
2604 (assign_parm_find_entry_rtl): Likewise.
2605 (assign_parm_is_stack_parm): Likewise.
2606 (assign_parm_load_bounds): Likewise.
2607 (assign_bounds): Likewise.
2608 (assign_parms): Likewise.
2609 (expand_function_start): Likewise.
2610 * gcc.c (CHKP_SPEC): Likewise.
2611 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2612 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2613 (wrestrict_dom_walker::check_call): Likewise.
2614 * gimple.c (gimple_build_call_from_tree): Likewise.
2615 * gimple.h (enum gf_mask): Likewise.
2616 (gimple_call_with_bounds_p): Likewise.
2617 (gimple_call_set_with_bounds): Likewise.
2618 * gimplify.c (gimplify_init_constructor): Likewise.
2619 * ipa-cp.c (initialize_node_lattices): Likewise.
2620 (propagate_constants_across_call): Likewise.
2621 (find_more_scalar_values_for_callers_subset): Likewise.
2622 * ipa-hsa.c (process_hsa_functions): Likewise.
2623 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2624 * ipa-icf.c (sem_function::merge): Likewise.
2625 * ipa-inline.c (early_inliner): Likewise.
2626 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2627 (warn_function_cold): Likewise.
2628 (propagate_pure_const): Likewise.
2629 * ipa-ref.h (enum GTY): Likewise.
2630 * ipa-split.c (find_retbnd): Likewise.
2631 (consider_split): Likewise.
2632 (split_function): Likewise.
2633 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2634 * ipa.c (walk_polymorphic_call_targets): Likewise.
2635 (symbol_table::remove_unreachable_nodes): Likewise.
2636 (process_references): Likewise.
2637 (cgraph_build_static_cdtor_1): Likewise.
2638 * lto-cgraph.c (lto_output_node): Likewise.
2639 (output_refs): Likewise.
2640 (compute_ltrans_boundary): Likewise.
2641 (input_overwrite_node): Likewise.
2642 (input_node): Likewise.
2643 (input_cgraph_1): Likewise.
2644 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2645 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2646 (class pass_chkp_instrumentation_passes): Likewise.
2647 (make_pass_chkp_instrumentation_passes): Likewise.
2648 * passes.def: Likewise.
2649 * rtl.h (struct GTY): Likewise.
2650 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2651 * stor-layout.c (layout_type): Likewise.
2652 * symtab.c: Likewise.
2653 * target.def: Likewise.
2654 * targhooks.c (default_chkp_bound_type): Likewise.
2655 (default_chkp_bound_mode): Likewise.
2656 (default_builtin_chkp_function): Likewise.
2657 (default_chkp_function_value_bounds): Likewise.
2658 (default_chkp_make_bounds_constant): Likewise.
2659 (default_chkp_initialize_bounds): Likewise.
2660 * targhooks.h (default_chkp_bound_type): Likewise.
2661 (default_chkp_bound_mode): Likewise.
2662 (default_builtin_chkp_function): Likewise.
2663 (default_chkp_function_value_bounds): Likewise.
2664 (default_chkp_make_bounds_constant): Likewise.
2665 (default_chkp_initialize_bounds): Likewise.
2666 * toplev.c (compile_file): Likewise.
2667 (process_options): Likewise.
2668 * tree-core.h (DEF_BUILTIN): Likewise.
2669 (DEF_BUILTIN_CHKP): Likewise.
2670 * tree-inline.c (declare_return_variable): Likewise.
2671 (remap_gimple_stmt): Likewise.
2672 (copy_bb): Likewise.
2673 (initialize_inlined_parameters): Likewise.
2674 (expand_call_inline): Likewise.
2675 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2676 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2677 (make_pass_ipa_chkp_produce_thunks): Likewise.
2678 (make_pass_chkp): Likewise.
2679 (make_pass_chkp_opt): Likewise.
2680 (make_pass_chkp_instrumentation_passes): Likewise.
2681 * tree-pretty-print.c (dump_generic_node): Likewise.
2682 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2683 * tree-ssa-dce.c (propagate_necessity): Likewise.
2684 (eliminate_unnecessary_stmts): Likewise.
2685 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2686 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2687 * tree-ssa-sccvn.h: Likewise.
2688 * tree-ssa-strlen.c (get_string_length): Likewise.
2689 (valid_builtin_call): Likewise.
2690 (adjust_last_stmt): Likewise.
2691 (handle_builtin_strchr): Likewise.
2692 (handle_builtin_strcpy): Likewise.
2693 (handle_builtin_stxncpy): Likewise.
2694 (handle_builtin_memcpy): Likewise.
2695 (handle_builtin_strcat): Likewise.
2696 (strlen_check_and_optimize_stmt): Likewise.
2697 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2698 * tree-streamer-in.c: Likewise.
2699 * tree-streamer.c (record_common_node): Likewise.
2700 * tree.c (tree_code_size): Likewise.
2701 (wide_int_to_tree_1): Likewise.
2702 (type_contains_placeholder_1): Likewise.
2703 (build_common_tree_nodes): Likewise.
2704 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2705 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2706 (POINTER_BOUNDS_P): Likewise.
2707 (BOUNDED_TYPE_P): Likewise.
2708 (BOUNDED_P): Likewise.
2709 (CALL_WITH_BOUNDS_P): Likewise.
2710 (pointer_bounds_type_node): Likewise.
2711 * value-prof.c (gimple_ic): Likewise.
2712 * var-tracking.c (vt_add_function_parameters): Likewise.
2713 * varasm.c (make_decl_rtl): Likewise.
2714 (assemble_start_function): Likewise.
2715 (output_constant): Likewise.
2716 (maybe_assemble_visibility): Likewise.
2717 * varpool.c (ctor_for_folding): Likewise.
2718 * chkp-builtins.def: Remove.
2719 * ipa-chkp.c: Remove.
2720 * ipa-chkp.h: Remove.
2721 * rtl-chkp.c: Remove.
2722 * rtl-chkp.h: Remove.
2723 * tree-chkp-opt.c: Remove.
2724 * tree-chkp.c: Remove.
2725 * tree-chkp.h: Remove.
2727 2018-06-07 Carl Love <cel@us.ibm.com>
2729 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2730 vextract_fp_from_shortl): Add BE support.
2732 2018-06-07 Paul Koning <ni1d@arrl.net>
2734 * compare-elim.c (try_merge_compare): Don't merge compare if
2735 address contains a side effect.
2736 (try_eliminate_compare): Likewise.
2738 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2740 * config.gcc: Support "tremont".
2741 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2742 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2744 * config/i386/i386.c (m_TREMONT): Define.
2745 (processor_target_table): Add "tremont".
2746 (PTA_TREMONT): Define.
2747 (ix86_lea_outperforms): Add TARGET_TREMONT.
2748 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2749 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2750 and M_INTEL_GOLDMONT_PLUS.
2751 (fold_builtin_cpu): Add "tremont".
2752 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2753 (ix86_option_override_internal): Add "tremont".
2754 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2755 (processor_type): Add PROCESSOR_TREMONT.
2756 * config/i386/x86-tune.def: Add m_TREMONT.
2757 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2759 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2761 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2762 symbol defined for msp430i* devices to be lower case.
2764 gcc/testsuite/gcc.target/msp430/
2765 * msp430i-device-symbol.c: New test.
2766 * msp430f-device-symbol.c: New test.
2767 * msp430.h: New test header file.
2769 2018-06-07 Richard Biener <rguenther@suse.de>
2771 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2772 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2773 Properly wrap signed arithmetic if overflow wraps.
2775 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2777 PR tree-optimization/69615
2778 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2779 of a cast from a same precision integral SSA_NAME in a bb dominated
2780 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2781 cast to utype if rhs2 has already a compatible type.
2783 2018-06-07 Richard Biener <rguenther@suse.de>
2785 PR tree-optimization/85935
2786 * graphite-scop-detection.c (find_params_in_bb): Analyze
2787 condition operands with respect to the correct loop. Assert
2788 the analysis doesn't fail.
2790 2018-06-04 Carl Love <cel@us.ibm.com>
2792 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2793 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2794 as it is slightly cheaper.
2795 (first_match_or_eos_index_<mode>):
2796 Calculate index using natural element order.
2797 (first_match_index_<mode>):
2798 Calculate index using natural element order.
2799 (first_match_or_eos_index_<mode>):
2800 Calculate index using natural order.
2801 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2802 for BE and LE modes.
2803 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2804 P9V_BUILTIN_VCLZLSBB_V16QI.
2805 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2808 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2810 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2811 indentation and line wrap for many prototypes. Add missing
2812 @smallexample directives around block of prototypes for vec_xl and
2815 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2817 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2818 track if we pass or return IEEE 128-bit floating point.
2819 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2820 C++ mangling that is compatible with GCC 8.1.
2821 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2822 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2823 floating point types.
2824 (rs6000_function_arg_advance_1): Likewise.
2825 (rs6000_mangle_type): Optionally generate mangled names that match
2826 what GCC 8.1 generated for IEEE 128-bit floating point types.
2827 (rs6000_globalize_decl_name): If we have an external function that
2828 passes or returns IEEE 128-bit types, generate a weak reference
2829 from the mangled name used in GCC 8.1 to the current mangled
2831 (rs6000_init_builtins): Make __ibm128 use the long double type if
2832 long double is IBM extended double. Make __float128 use the long
2833 double type if long double is IEEE 128-bit.
2836 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2837 macro for __ibm128 built-in functions.
2838 (PACK_IF): Add __ibm128 pack/unpack functions.
2839 (UNPACK_IF): Likewise.
2840 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2841 enable long double built-in functions if long double is IEEE
2842 128-bit floating point.
2843 (rs6000_invalid_builtin): Update long double built-in function
2845 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2846 functions, adjust the built-in function to use the long double
2847 built-in function if __ibm128 and long double are the same type.
2848 * doc/extend.texi (PowerPC builtins): Update documention for
2849 __builtin_{,un}pack_longdouble. Add documentation for
2850 __builtin_{,un}pack_ibm128.
2852 2018-06-06 Jim Wilson <jimw@sifive.com>
2854 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2855 (struct machine_function): New field interrupt_mode.
2856 (riscv_handle_type_attribute): New function. Add forward declaration.
2857 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2858 (riscv_expand_epilogue): Check interrupt_mode field.
2859 (riscv_set_current_function): Check interrupt attribute args and
2860 set interrupt_mode field.
2861 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2862 (riscv_sret, riscv_uret): New.
2863 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2864 new arguments to interrupt attribute.
2866 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2869 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2870 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2872 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2874 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2875 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2876 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2877 several redundant entries.
2879 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2881 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2882 type from "rtx" to "rtx_insn *".
2883 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2884 for local "call_insn", removing cast.
2885 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2887 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2889 PR tree-optimization/86066
2890 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2891 for BIT_INSERT_EXPR stores.
2893 2018-06-06 Richard Biener <rguenther@suse.de>
2895 PR tree-optimization/86062
2896 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2897 component refs ontop
2898 of to be offsetted base.
2900 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2902 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2903 to be static and remove check on interrupt attribute name.
2905 gcc/testsuite/gcc.target/msp430/
2906 * function-attributes-4.c: New test.
2907 * static-interrupts.c: New test.
2909 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2911 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2912 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2914 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2917 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2919 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2920 Remove second argument, change how error is called.
2921 (aarch64_layout_arg): Remove second argument from
2922 aarch64_err_no_fpadvsimd call.
2923 (aarch64_init_cumulative_args): Ditto.
2924 (aarch64_gimplify_va_arg_expr): Ditto.
2925 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2927 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2929 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2930 (*simple_return_indirect_internal<mode>): Rename from
2931 simple_return_indirect_internal. Use W mode iterator.
2932 (rstorssp): New expander.
2933 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2934 (clrssbsy): New expander.
2935 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2937 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2939 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2941 (cmse_check_pointed_object): Likewise.
2943 2018-06-05 Martin Liska <mliska@suse.cz>
2945 PR gcov-profile/47618
2946 * doc/invoke.texi: Document how -fprofile-dir format
2949 2018-06-05 Richard Biener <rguenther@suse.de>
2951 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2952 removal pretend DOM info isn't available so we do not update
2953 it and only remove edges, not dominated blocks. Actually free
2954 DOM info in case we removed something. Remove unreachable blocks.
2955 (mfb_keep_latches): Work with either DOM info or marked backedges.
2956 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2957 first. Mark backedges if DOM info isn't available.
2958 (Re-)compute DOM info after cleanup_control_flow_pre.
2960 2018-06-05 Richard Biener <rguenther@suse.de>
2962 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2963 (locus_discrim_hasher::hash): Adjust.
2964 (locus_discrim_hasher::equal): Likewise.
2965 (next_discriminator_for_locus): Work on line directly.
2966 (same_line_p): Pass in expanded locus1 as well.
2967 (assign_discriminators): Avoid redundant location expansions.
2969 2018-06-05 Richard Biener <rguenther@suse.de>
2971 PR tree-optimization/86046
2972 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2973 if required after clearing TREE_ADDRESSABLE.
2975 2018-06-05 Richard Biener <rguenther@suse.de>
2977 PR tree-optimization/86047
2978 * tree-ssa-loop.c (for_each_index): Glob handling of all
2979 decls and constants and really handle all of them.
2981 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2984 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2985 qualifier_void_pointer and qualifier_const_void_pointer.
2986 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2987 (arm_init_builtins): Handle the above.
2988 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2989 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2990 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2993 2018-06-05 Martin Liska <mliska@suse.cz>
2995 * auto-profile.c (read_autofdo_file): Do not use
2996 gcov_ctr_summary struct.
2997 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2998 * coverage.c (struct counts_entry): Likewise.
2999 (read_counts_file): Read just single summary entry.
3000 (get_coverage_counts): Use gcov_summary struct.
3001 * coverage.h (get_coverage_counts): Likewise.
3002 * gcov-dump.c (dump_working_sets): Likewise.
3003 (tag_summary): Dump just single summary.
3004 * gcov-io.c (gcov_write_summary): Write just histogram
3006 (gcov_read_summary): Read just single summary.
3007 (compute_working_sets): Use gcov_summary struct.
3008 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3009 of GCOV_COUNTERS_SUMMABLE.
3010 (GCOV_COUNTERS_SUMMABLE): Remove.
3011 (GCOV_FIRST_VALUE_COUNTER): Replace with
3012 GCOV_COUNTER_V_INTERVAL.
3013 (struct gcov_ctr_summary): Remove.
3014 (struct gcov_summary): Directly use fields of former
3016 (compute_working_sets): Use gcov_summary struct.
3017 * gcov.c (read_count_file): Do not use ctrs fields.
3018 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3020 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3022 * profile.c: Likewise.
3023 * profile.h: Likewise.
3025 2018-06-05 Martin Liska <mliska@suse.cz>
3027 PR gcov-profile/84846
3028 * gcov.c (output_lines): Print working directory only
3029 in intermediate format.
3031 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3033 * config/s390/s390-builtin-types.def: Add void function type.
3034 * config/s390/s390-builtins.def: Use the function type for the
3037 2018-06-04 Jim Wilson <jimw@sifive.com>
3039 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3041 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3042 and maybe_eh_return. Change regno to unsigned int. Use new args to
3043 handle EH_RETURN_DATA_REGNO registers properly.
3044 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3045 (riscv_expand_epilogue): Update comment. Change argument name and
3046 type. Update code to use new name and type. Pass new args to
3047 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3049 * config/riscv/riscv.md (NORMAL_RETURN): New.
3050 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3051 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3052 (eh_return): Call gen_eh_return_internal and emit barrier.
3053 (eh_return_internal): Call riscv_expand_epilogue.
3055 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3057 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3058 bit_insertion field and declare can_be_merged_into method.
3059 (merged_store_group::can_be_merged_into): New method.
3060 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3061 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3062 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3064 2018-06-04 Richard Biener <rguenther@suse.de>
3066 PR tree-optimization/85955
3067 * builtins.c (fold_builtin_sincos): Convert pointers to
3068 destination to appropriate type before dereferencing.
3070 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3072 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3074 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3076 * expr.c (expand_expr_real_1): Force the operand into memory if
3077 its TYPE_MODE is BLKmode and if there is no integer mode for
3078 the number of bits being extracted.
3080 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3084 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3085 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3087 2018-06-04 Richard Biener <rguenther@suse.de>
3089 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3090 (cleanup_tree_cfg_noloop): ... single caller. Do
3091 start_recording_case_labels later.
3093 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3095 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3096 to _IMMINTRIN_H_INCLUDED.
3097 * config/i386/pconfigintrin.h: Ditto.
3098 * config/i386/waitpkgintrin.h: Ditto.
3099 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3100 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3101 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3102 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3103 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3104 waitpkgintrin.h and cldemoteintrin.h.
3106 2018-06-04 Richard Biener <rguenther@suse.de>
3108 PR tree-optimization/86038
3109 * tracer.c (find_best_successor): Check probability for
3110 being initialized, bail out if not.
3112 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3115 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3116 of bits to ignore when comparing architectures.
3118 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3120 PR tree-optimization/69615
3121 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3122 maximum or minimum of the type, try to merge it also as if
3123 range1 is + [-, x - 1] or + [x + 1, -].
3126 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3128 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3130 PR tree-optimization/86034
3131 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3132 the unsigned bitfield type in a bit insertion sequence if it does not
3133 have a larger precision than the bitfield size.
3134 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3136 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3138 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3140 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3142 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3143 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3144 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3145 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3147 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3149 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3150 Disable -fdelete-null-pointer-checks for ELF toolchain.
3152 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3153 Kito Cheng <kito.cheng@gmail.com>
3155 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3156 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3157 (nds32*-*-*): Add glibc and uclibc conditions.
3158 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3159 (TARGET_EXCEPT_UNWIND_INFO): Define.
3160 * config/nds32/elf.h: New file.
3161 * config/nds32/linux.h: New file.
3162 * config/nds32/nds32-elf.opt: New file.
3163 * config/nds32/nds32-linux.opt: New file.
3164 * config/nds32/nds32-fp-as-gp.c
3165 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3166 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3168 (nds32_asm_file_end): Ditto.
3169 (nds32_print_operand): Ditto.
3170 (nds32_insert_attributes): Ditto.
3171 (nds32_init_libfuncs): New function.
3172 (TARGET_HAVE_TLS): Define.
3173 (TARGET_INIT_LIBFUNCS): Define.
3174 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3176 (TARGET_ELF): Apply different mcmodel setting.
3177 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3178 been migrated into elf.h and linux.h files.
3179 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3180 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3181 (mcmodel): The content has been migrated into nds32-elf.opt and
3182 nds32-linux.opt files.
3183 * config/nds32/t-elf: New file.
3184 * config/nds32/t-linux: New file.
3186 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3187 Shiva Chen <shiva0217@gmail.com>
3189 * config/nds32/constants.md (unspec_volatile_element): Add
3190 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3191 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3193 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3194 (make_pass_nds32_fp_as_gp): Declare.
3195 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3197 (nds32_asm_function_end_prologue): Remove unused asm output.
3198 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3199 (nds32_asm_file_start): Output necessary fp_as_gp information.
3200 (nds32_option_override): Adjust register usage.
3201 (nds32_expand_prologue): Consider fp_as_gp situation.
3202 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3203 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3206 (simple_return): Ditto.
3207 (omit_fp_begin): Output special directive for fp_as_gp.
3208 (omit_fp_end): Output special directive for fp_as_gp.
3209 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3210 mforbid-fp-as-gp): New options.
3212 2018-06-01 Mark Wielaard <mark@klomp.org>
3214 * dwarf2out.c (dwarf2out_finish): Remove generation of
3215 DW_AT_loclists_base.
3217 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3219 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3220 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3221 (struct merged_store_group): Add bit_insertion field.
3222 (dump_char_array): Use standard hexadecimal format.
3223 (merged_store_group::merged_store_group): Set bit_insertion to false.
3224 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3225 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3226 also print the mask in the dump file.
3227 (pass_store_merging::gate): Minor tweak.
3228 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3229 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3230 stores with INTEGER_CST stores.
3231 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3232 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3233 and use it throughout. Generate bit insertion sequences if need be.
3234 (pass_store_merging::process_store): Remove redundant condition.
3235 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3237 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3239 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3240 the 128-bit floating point types. Fix function comment.
3242 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3244 * config/aarch64/aarch64-simd.md
3245 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3247 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3250 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3252 PR tree-optimization/85989
3253 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3255 (backprop::intersect_uses): Check it when deciding whether this
3256 is a backedge reference.
3257 (backprop::process_block): Add each phi to m_visited_phis
3258 after visiting it, then clear it at the end.
3260 2018-06-01 Richard Biener <rguenther@suse.de>
3262 * tree-vectorizer.h (vect_dr_stmt): New function.
3263 (vect_get_load_cost): Adjust.
3264 (vect_get_store_cost): Likewise.
3265 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3266 Use vect_dr_stmt instead of DR_SMTT.
3267 (vect_record_base_alignments): Likewise.
3268 (vect_calculate_target_alignment): Likewise.
3269 (vect_compute_data_ref_alignment): Likewise and make static.
3270 (vect_update_misalignment_for_peel): Likewise.
3271 (vect_verify_datarefs_alignment): Likewise.
3272 (vector_alignment_reachable_p): Likewise.
3273 (vect_get_data_access_cost): Likewise. Pass down
3274 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3275 (vect_get_peeling_costs_all_drs): Likewise.
3276 (vect_peeling_hash_get_lowest_cost): Likewise.
3277 (vect_enhance_data_refs_alignment): Likewise.
3278 (vect_find_same_alignment_drs): Likewise.
3279 (vect_analyze_data_refs_alignment): Likewise.
3280 (vect_analyze_group_access_1): Likewise.
3281 (vect_analyze_group_access): Likewise.
3282 (vect_analyze_data_ref_access): Likewise.
3283 (vect_analyze_data_ref_accesses): Likewise.
3284 (vect_vfa_segment_size): Likewise.
3285 (vect_small_gap_p): Likewise.
3286 (vectorizable_with_step_bound_p): Likewise.
3287 (vect_prune_runtime_alias_test_list): Likewise.
3288 (vect_analyze_data_refs): Likewise.
3289 (vect_supportable_dr_alignment): Likewise.
3290 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3291 (vect_gen_prolog_loop_niters): Likewise.
3292 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3293 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3295 (vect_recog_mask_conversion_pattern): Likewise.
3296 (vect_try_gather_scatter_pattern): Likewise.
3297 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3298 to vect_get_store_cost.
3299 (vect_get_store_cost): Get stmt_info instead of DR.
3300 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3301 (vect_get_load_cost): Get stmt_info instead of DR.
3303 2018-06-01 Richard Biener <rguenther@suse.de>
3306 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3307 vars wrapped in MEM_REFs.
3309 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3311 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3312 Fix subreg tests so that we only return a choice between
3313 GENERAL_REGS and FP_REGS if the original classes included both.
3315 2018-06-01 Richard Biener <rguenther@suse.de>
3318 * tree-ssa-structalias.c (get_function_part_constraint):
3319 Handle NULL fi->decl.
3320 (find_func_aliases_for_call): Properly handle indirect
3321 fi from direct call.
3322 (find_func_clobbers): Likewise.
3323 (ipa_pta_execute): Likewise.
3324 (create_variable_info_for): For functions that are ifunc_resolver
3325 resolve to a varinfo that contains the result of the resolver call.
3326 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3329 2018-05-31 Michael Collison <michael.collison@arm.com>
3331 * config/aarch64/aarch64.md:
3332 (*fix_to_zero_extenddfdi2): New pattern.
3333 * gcc.target/aarch64/fix_extend1.c: New testcase.
3335 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3339 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3340 and BUILT_IN_STRNCMP_EQ.
3341 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3342 BUILT_IN_STRNCMP_EQ.
3343 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3344 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3345 (gimple_fold_builtin): Likewise.
3346 * tree-ssa-strlen.c (compute_string_length): New function.
3347 (determine_min_obsize): New function.
3348 (handle_builtin_string_cmp): New function to handle calls to
3349 string compare functions.
3350 (strlen_optimize_stmt): Add handling to builtin string compare
3352 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3353 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3354 * tree.c (build_common_builtin_nodes): Add new defines of
3355 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3357 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3360 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3361 functions with naked attribute.
3363 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3365 * config/i386/sse.md (avx_vec_concat<mode>):
3366 Substitute concat_tg_mode mode attribute with xtg_mode.
3367 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3368 (concat_tg_mode): Remove mode attribute.
3370 2018-05-31 Martin Sebor <msebor@redhat.com>
3373 * calls.c (alloc_max_size): Correct a logic error/typo.
3374 Treat excessive arguments as infinite. Warn for invalid arguments.
3375 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3377 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3380 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3381 and movx for Haswell.
3383 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3384 Cesar Philippidis <cesar@codesourcery.com>
3387 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3388 when emitting error on private/firstprivate reductions.
3389 * omp-low.c (lower_omp_target): Avoid reference-type processing
3390 on pointers for firstprivate clause.
3392 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3394 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3397 * config/aarch64/aarch64-simd.md
3398 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3399 (aarch64_ld1_x3_<mode>): Likewise
3400 (aarch64_st1x2<VALLDIF:mode>): Likewise
3401 (aarch64_st1_x2_<mode>): Likewise
3402 (aarch64_st1x3<VALLDIF:mode>): Likewise
3403 (aarch64_st1_x3_<mode>): Likewise
3404 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3405 (vld1_s8_x3): Likewise.
3406 (vld1_u16_x3): Likewise.
3407 (vld1_s16_x3): Likewise.
3408 (vld1_u32_x3): Likewise.
3409 (vld1_s32_x3): Likewise.
3410 (vld1_u64_x3): Likewise.
3411 (vld1_s64_x3): Likewise.
3412 (vld1_f16_x3): Likewise.
3413 (vld1_f32_x3): Likewise.
3414 (vld1_f64_x3): Likewise.
3415 (vld1_p8_x3): Likewise.
3416 (vld1_p16_x3): Likewise.
3417 (vld1_p64_x3): Likewise.
3418 (vld1q_u8_x3): Likewise.
3419 (vld1q_s8_x3): Likewise.
3420 (vld1q_u16_x3): Likewise.
3421 (vld1q_s16_x3): Likewise.
3422 (vld1q_u32_x3): Likewise.
3423 (vld1q_s32_x3): Likewise.
3424 (vld1q_u64_x3): Likewise.
3425 (vld1q_s64_x3): Likewise.
3426 (vld1q_f16_x3): Likewise.
3427 (vld1q_f32_x3): Likewise.
3428 (vld1q_f64_x3): Likewise.
3429 (vld1q_p8_x3): Likewise.
3430 (vld1q_p16_x3): Likewise.
3431 (vld1q_p64_x3): Likewise.
3432 (vst1_s64_x2): Likewise.
3433 (vst1_u64_x2): Likewise.
3434 (vst1_f64_x2): Likewise.
3435 (vst1_s8_x2): Likewise.
3436 (vst1_p8_x2): Likewise.
3437 (vst1_s16_x2): Likewise.
3438 (vst1_p16_x2): Likewise.
3439 (vst1_s32_x2): Likewise.
3440 (vst1_u8_x2): Likewise.
3441 (vst1_u16_x2): Likewise.
3442 (vst1_u32_x2): Likewise.
3443 (vst1_f16_x2): Likewise.
3444 (vst1_f32_x2): Likewise.
3445 (vst1_p64_x2): Likewise.
3446 (vst1q_s8_x2): Likewise.
3447 (vst1q_p8_x2): Likewise.
3448 (vst1q_s16_x2): Likewise.
3449 (vst1q_p16_x2): Likewise.
3450 (vst1q_s32_x2): Likewise.
3451 (vst1q_s64_x2): Likewise.
3452 (vst1q_u8_x2): Likewise.
3453 (vst1q_u16_x2): Likewise.
3454 (vst1q_u32_x2): Likewise.
3455 (vst1q_u64_x2): Likewise.
3456 (vst1q_f16_x2): Likewise.
3457 (vst1q_f32_x2): Likewise.
3458 (vst1q_f64_x2): Likewise.
3459 (vst1q_p64_x2): Likewise.
3460 (vst1_s64_x3): Likewise.
3461 (vst1_u64_x3): Likewise.
3462 (vst1_f64_x3): Likewise.
3463 (vst1_s8_x3): Likewise.
3464 (vst1_p8_x3): Likewise.
3465 (vst1_s16_x3): Likewise.
3466 (vst1_p16_x3): Likewise.
3467 (vst1_s32_x3): Likewise.
3468 (vst1_u8_x3): Likewise.
3469 (vst1_u16_x3): Likewise.
3470 (vst1_u32_x3): Likewise.
3471 (vst1_f16_x3): Likewise.
3472 (vst1_f32_x3): Likewise.
3473 (vst1_p64_x3): Likewise.
3474 (vst1q_s8_x3): Likewise.
3475 (vst1q_p8_x3): Likewise.
3476 (vst1q_s16_x3): Likewise.
3477 (vst1q_p16_x3): Likewise.
3478 (vst1q_s32_x3): Likewise.
3479 (vst1q_s64_x3): Likewise.
3480 (vst1q_u8_x3): Likewise.
3481 (vst1q_u16_x3): Likewise.
3482 (vst1q_u32_x3): Likewise.
3483 (vst1q_u64_x3): Likewise.
3484 (vst1q_f16_x3): Likewise.
3485 (vst1q_f32_x3): Likewise.
3486 (vst1q_f64_x3): Likewise.
3487 (vst1q_p64_x3): Likewise.
3489 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3491 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3492 user_label_prefix to name.
3494 * tree-core.h: Update comment about the format of NAME string
3495 passed to handler in attribute_spec.
3497 * config/msp430/msp430.md: Remove erroneous subreg expression from
3498 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3499 zero_extend{q,h}isi2.
3501 2018-05-30 Borislav Petkov <bp@suse.de>
3503 * doc/extend.texi: Document some architecture specific
3504 constraints and sort entries.
3506 2018-05-30 Martin Sebor <msebor@redhat.com>
3509 * builtins.c (expand_builtin_stpcpy_1): New function.
3510 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3511 only if the former succeeds.
3513 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3515 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3518 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3520 * doc/invoke.texi (-flinker-output): Document
3522 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3524 * passes.c (ipa_write_summaries): Only modify statements if body
3526 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3527 incrementally linking.
3528 (ipa_passes): Likewise.
3529 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3530 pass down resolution info.
3531 * common.opt (flag_incremental_link): Update info.
3532 * gcc.c (plugin specs): Turn flinker-output=* to
3533 -plugin-opt=-linker-output-known
3534 * toplev.c (compile_file): Also cut compilation when doing incremental
3536 * flag-types. (enum lto_partition_model): Add
3537 LTO_LINKER_OUTPUT_NOLTOREL.
3538 (invoke.texi): Add -flinker-output docs.
3539 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3540 link same way as WPA; do not stream in dead initializers.
3542 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3543 darf2out_register_external_decl): Support incremental link.
3545 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3547 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3549 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3551 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3552 it down to simple_object_copy_lto_debug_sections.
3553 (run_gcc): Determine incremental LTO link time and configure
3554 lto1 into non-wpa mode, disable renaming of debug sections.
3556 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3558 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3559 descriptions of various incorrectly documented functions.
3561 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3564 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3565 address check not strict.
3567 2018-05-30 Richard Biener <rguenther@suse.de>
3569 PR tree-optimization/85964
3570 * tracer.c (better_p): Drop initialized count check, we only
3571 call the function with initialized counts now.
3572 (find_best_successor): Do find a best edge if one
3573 has uninitialized count.
3574 (find_best_predecessor): Likewise. Do BB frequency check only
3575 if count is initialized.
3577 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3579 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3580 (aarch64_ldrstr_offset_compare): New.
3581 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3582 load/store orderings.
3583 (aarch64_gen_adjusted_ldpstp): Likewise.
3585 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3587 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3588 Check for subset of GENERAL_REGS and FP_REGS.
3589 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3592 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3594 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3595 and wi::to_poly_offset. Add the current offset and then check
3596 whether the sum fits, rather than using an unchecked addition of
3597 a checked term. Check for a shwi rather than a uhwi.
3598 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3599 (store_constructor): Use poly_int_tree_p.
3600 (expand_expr_real_1): Likewise.
3601 * function.c (assign_temp): Likewise.
3602 * fold-const.c (const_binop): Use poly_int_tree_p and
3604 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3606 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3607 to_poly_offset for MEM offsets.
3608 * ipa-icf.c (sem_variable::equals): Likewise.
3609 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3610 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3611 wi::to_poly_offset for BIT_FIELD_REF offsets.
3612 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3614 * var-tracking.c (emit_note_insn_var_location): Use
3615 tree_to_poly_uint64.
3617 2018-05-29 Jim Wilson <jimw@sifive.com>
3619 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3621 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3624 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3625 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3627 (sse4_1_round<mode>2): Use nonimmediate_operand
3628 for operand 1 predicate.
3630 2018-05-29 Martin Sebor <msebor@redhat.com>
3631 Richard Biener <rguenther@suse.de>
3634 * calls.c (get_size_range): Call determine_value_range instead
3635 of get_value_range..
3636 * tree-vrp.h (determine_value_range): Declared new function.
3637 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3639 2018-05-29 Richard Biener <rguenther@suse.de>
3641 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3642 sure to use non-pattern stmts for get_earlier_stmt arguments.
3643 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3644 called on pattern stmts.
3645 (get_later_stmt): Likewise.
3647 2018-05-29 Martin Liska <mliska@suse.cz>
3649 PR gcov-profile/85759
3650 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3653 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3655 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3657 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3661 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3662 VEC_PACK_FLOAT_EXPR): New tree codes.
3663 * tree-pretty-print.c (op_code_prio): Handle
3664 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3665 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3666 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3667 * tree-inline.c (estimate_operator_cost): Likewise.
3668 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3669 * fold-const.c (const_binop): Likewise.
3670 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3671 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3672 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3673 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3674 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3675 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3676 * expr.c (expand_expr_real_2): Likewise.
3677 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3678 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3679 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3681 * optabs.c (expand_widen_pattern_expr): For
3682 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3683 sign from result type rather than operand's type.
3684 (expand_binop_directly): For vec_packu_float_optab and
3685 vec_packs_float_optab allow result type to be different from operand's
3687 * optabs-tree.c (optab_for_tree_code): Handle
3688 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3689 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3690 * tree-vect-generic.c (expand_vector_operations_1): Handle
3691 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3692 VEC_PACK_FLOAT_EXPR.
3693 * tree-vect-stmts.c (supportable_widening_operation): Handle
3695 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3696 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3697 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3698 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3699 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3701 (vec_pack<floatprefix>_float_<mode>): New expander.
3702 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3704 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3705 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3706 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3707 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3708 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3710 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3711 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3712 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3713 VEC_PACK_FLOAT_EXPR): Document.
3715 2018-05-29 Richard Biener <rguenther@suse.de>
3717 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3719 (stmt_vec_info_vec): Make pointer.
3720 (init_stmt_vec_info_vec): Remove.
3721 (free_stmt_vec_info_vec): Likewise.
3722 (set_stmt_vec_info_vec): New function.
3723 (free_stmt_vec_infos): Likewise.
3724 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3725 (set_vinfo_for_stmt): Likewise.
3726 (get_earlier_stmt): Likewise.
3727 (get_later_stmt): Likewise.
3728 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3729 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3730 (vec_info::~vec_info): Free stmt_vec_infos.
3731 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3732 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3733 (pass_slp_vectorize::execute): Likewise.
3734 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3735 (free_stmt_vec_info_vec): Likewise.
3736 (set_stmt_vec_info_vec): New function.
3737 (free_stmt_vec_infos): Likewise.
3738 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3739 the global stmt_vec_info_vec.
3740 * tree-parloops.c (gather_scalar_reductions): Use
3741 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3744 2018-05-29 Richard Biener <rguenther@suse.de>
3746 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3748 2018-05-29 Martin Liska <mliska@suse.cz>
3749 David Malcolm <dmalcolm@redhat.com>
3751 * vec.c (test_reverse): New.
3752 (vec_c_tests): Add new test.
3753 * vec.h (vl_ptr>::reverse): New function.
3755 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3757 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3759 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3762 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3764 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3766 2018-05-28 Richard Biener <rguenther@suse.de>
3768 PR tree-optimization/85933
3769 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3770 look at stmts marked as vectorizable.
3772 2018-05-28 Richard Biener <rguenther@suse.de>
3774 PR tree-optimization/85934
3775 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3776 vector boolean check before scalar optimization.
3778 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3780 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3783 2018-05-28 Mark Wielaard <mark@klomp.org>
3785 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3786 if it is an expression containing a minus sign.
3788 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3790 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3792 2018-05-27 Paul Koning <ni1d@arrl.net>
3794 * config/pdp11/pdp11.md (truncsihi2): Remove.
3796 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3797 Chung-Ju Wu <jasonwucj@gmail.com>
3799 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3801 (unaligned_store_dw): Ditto.
3802 * config/nds32/nds32-memory-manipulation.c
3803 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3804 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3805 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3806 (emit_setmem_word_loop): Rename to ...
3807 (emit_setmem_doubleword_loop): ... this.
3808 (nds32_gen_dup_4_byte_to_word_value): New function.
3809 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3810 (nds32_expand_setmem_loop): Refine implementation.
3811 (nds32_expand_setmem_loop_v3m): Ditto.
3812 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3815 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3817 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3819 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3821 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3822 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3823 machine->attr_no_prologue_p.
3824 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3825 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3826 (nds32_expand_epilogue): Consider attr_naked_p.
3827 (nds32_expand_epilogue_v3pop): Likewise.
3828 (nds32_can_use_return_insn): Likewise.
3829 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3830 attr_no_prologue_p fields.
3831 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3833 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3836 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3838 * config/i386/sse.md
3839 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3841 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3843 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3845 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3847 (*<floatsuffix>floatv2div2sf2): Rename to ...
3848 (*float<floatunssuffix>v2div2sf2): ... this.
3849 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3850 (float<floatunssuffix>v2div2sf2_mask): ... this.
3851 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3852 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3853 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3855 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3857 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3859 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3861 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3863 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3865 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3866 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3867 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3868 gen_ufix_truncv8dfv8si2.
3869 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3870 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3871 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3872 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3873 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3874 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3875 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3876 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3878 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3882 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3884 2018-05-25 Jim Wilson <jimw@sifive.com>
3886 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3887 * config/riscv/riscv.c (struct machine_function): Add
3888 interrupt_handler_p and attribute_checked_p fields.
3889 (riscv_attribute_table): Add interrupt.
3890 (riscv_interrupt_type_p): New.
3891 (riscv_save_reg_p): Save extra regs for interrupt handler.
3892 (riscv_use_save_libcall): Return false for interrupt handler.
3893 (riscv_first_stack_step): Add forward declaration.
3894 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3895 for interrupt handler with large frame. Use it for saved reg list.
3896 (riscv_expand_prologue): Move flag_stack_usage_info support to
3897 eliminate duplication.
3898 (riscv_expand_epilogue): Generate mret for interrupt handler.
3899 (riscv_epilogue_uses): New.
3900 (riscv_can_use_return_insn): Return false for interrupt handler.
3901 (riscv_function_ok_for_sibcall): Likewise.
3902 (riscv_set_current_function): Add interrupt handler support.
3903 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3904 * config/riscv/riscv.md (UNSPECV_MRET): New.
3906 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3908 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3910 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3912 PR tree-optimization/85712
3913 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3914 this candidate has already been replaced in-situ by a copy.
3916 2018-05-25 Jason Merrill <jason@redhat.com>
3918 PR c++/80485 - inline function non-zero address.
3919 * symtab.c (nonzero_address): Check DECL_COMDAT.
3921 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3924 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3925 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3926 extension of SImode operation. Use const123_operand predicate.
3927 (*saddsi_1): Remove.
3928 (*saddl_se_1): Ditto.
3930 (*ssubl_se_1): Ditto.
3931 * config/alpha/predicates.md (const123_operand): New predicate.
3932 * config/alpha/constraints.md (P): Use IN_RANGE.
3934 2018-05-25 Richard Biener <rguenther@suse.de>
3936 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3938 (ref_maybe_used_by_stmt_p): Likewise.
3939 (stmt_may_clobber_ref_p): Likewise.
3940 (stmt_may_clobber_ref_p_1): Likewise.
3941 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3943 (ref_maybe_used_by_stmt_p): Likewise.
3944 (stmt_may_clobber_ref_p): Likewise.
3945 (stmt_may_clobber_ref_p_1): Likewise.
3946 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3947 the alias oracle to disambiguate DRs with stmts DR analysis
3949 (vect_analyze_data_refs): Do not give up on not analyzable
3950 DRs for BB vectorization. Remove code truncating the dataref
3953 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3956 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3957 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3958 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3960 2018-05-25 Richard Biener <rguenther@suse.de>
3962 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3963 function, combining stmt data ref gathering and fatal analysis
3965 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3966 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3967 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3968 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3969 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3971 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3973 PR tree-optimization/85720
3974 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3975 SCC if all partitions are builtins.
3976 (version_loop_by_alias_check): New parameter. Generate cancelable
3977 runtime alias check if all partitions are builtins.
3978 (distribute_loop): Update call to above function.
3980 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3982 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3983 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3984 (parm_default_def_partition_arg): Ditto.
3985 (set_parm_default_def_partition): Ditto.
3986 (get_parm_default_def_partitions): Ditto and make it static.
3987 (get_undefined_value_partitions): Ditto and make it static.
3988 (remove_ssa_form): Refactor call to init_var_map here.
3989 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3990 computation for loop region.
3991 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3992 (register_default_def): Delete.
3993 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3994 (parm_default_def_partition_arg): Ditto.
3995 (set_parm_default_def_partition): Ditto.
3996 (get_parm_default_def_partitions): Ditto and make it static.
3997 (get_undefined_value_partitions): Ditto and make it static.
3998 (coalesce_with_default, coalesce_with_default): Update comment.
3999 (create_coalesce_list_for_region): New func factored out from
4000 create_outofssa_var_map.
4001 (populate_coalesce_list_for_outofssa): New func factored out from
4002 create_outofssa_var_map and coalesce_ssa_name.
4003 (create_outofssa_var_map): Delete.
4004 (coalesce_ssa_name): Refactor to support live range computation.
4005 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4006 (get_parm_default_def_partitions): Delete.
4007 (get_undefined_value_partitions): Ditto.
4008 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4009 computation for loop region.
4010 (new_tree_live_info, loe_visit_block): Ditto.
4011 (live_worklist, set_var_live_on_entry): Ditto.
4012 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4013 * tree-ssa-live.h (struct _var_map): New fields.
4014 (init_var_map): Change decl.
4015 (region_contains_p): New.
4017 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4019 * tree-ssa-live.h (live_merge_and_clear): Delete.
4021 2018-05-25 Richard Biener <rguenther@suse.de>
4024 * tree-dump.c (dequeue_and_dump): Remove access to removed
4025 operand 2 of a SWITCH_EXPR.
4027 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4029 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4030 multiplication and division.
4031 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4032 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4033 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4034 (cond_udiv_optab, cond_umod_optab): New optabs.
4035 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4036 (IFN_COND_RDIV): New internal functions.
4037 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4038 TRUNC_MOD_EXPR and RDIV_EXPR.
4039 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4040 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4042 (SVE_INT_BINARY): Include mult.
4043 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4044 (optab, sve_int_op): Handle mult.
4045 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4047 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4048 for SVE_INT_BINARY_SD.
4050 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4052 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4053 (optab, sve_int_op): Handle div and udiv.
4054 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4055 for SVE_INT_BINARY_SD.
4056 (*<optab><mode>3): New insn for the same.
4058 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4060 * tree-vect-patterns.c: Include predict.h.
4061 (vect_recog_divmod_pattern): Restrict check for division support
4062 to when optimizing for size.
4064 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4066 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4067 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4068 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4069 (gimple_match_op::set_op): Likewise.
4070 (gimple_resimplify4): Declare.
4071 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4072 (expr::gen_transform): Likewise.
4073 (decision_tree::gen): Generate a simplification routine for 4 operands.
4074 * gimple-match-head.c (gimple_simplify): Add an overload for
4075 4 operands. In the top-level function, handle up to 4 call
4076 arguments and call gimple_resimplify4.
4077 (gimple_resimplify4): New function.
4078 (build_call_internal): Pass a fourth operand.
4079 (maybe_push_to_seq): Likewise.
4080 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4081 Fold VEC_COND_EXPRs of an operation and a default value into
4082 an IFN_COND_* function if possible.
4083 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4085 (SVE_COND_FP_BINARY): Include them.
4086 (optab, sve_fp_op): Handle them.
4087 (SVE_INT_BINARY_REV): New code iterator.
4088 (SVE_COND_FP_BINARY_REV): New int iterator.
4089 (commutative): New int attribute.
4090 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4092 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4094 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4095 (*cond_<optab><mode>): New patterns for reversed operands.
4097 2018-05-25 Richard Biener <rguenther@suse.de>
4099 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4100 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4101 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4102 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4103 * tree-vect-data-refs.c (everywhere): Adjust users.
4104 * tree-vect-loop.c (everywhere): Likewise.
4105 * tree-vect-slp.c (everywhere): Likewise.
4106 * tree-vect-stmts.c (everywhere): Likewise.
4107 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4109 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4111 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4113 (gcc_cv_as_section_exclude): ... this.
4114 Try Solaris as #exclude syntax.
4115 * configure: Regenerate.
4116 * config.in: Regenerate.
4117 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4119 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4120 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4122 * varasm.c (default_elf_asm_named_section): Don't check if
4123 HAVE_GAS_SECTION_EXCLUDE is defined.
4125 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4127 * doc/md.texi: Update the documentation of the cond_* optabs
4128 to mention the new final operand. Fix GET_MODE_NUNITS call.
4129 Describe the scalar case too.
4130 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4131 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4133 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4134 (get_conditional_internal_fn): Update comment.
4135 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4136 accumulator value as a final argument to conditional functions.
4137 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4138 a define_expand and add an "else" operand. Assert for now that
4139 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4140 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4141 (*cond_<optab><mode>): New patterns.
4142 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4143 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4144 (UNSPEC_COND_EOR): Delete.
4145 (optab): Remove associated mappings.
4146 (SVE_INT_BINARY): New code iterator.
4147 (sve_int_op): Remove int attribute and add "minus" to the code
4149 (SVE_COND_INT_OP): Delete.
4150 (SVE_COND_FP_OP): Rename to...
4151 (SVE_COND_FP_BINARY): ...this.
4153 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4155 * optabs.c (can_reuse_operands_p): New function.
4156 (maybe_legitimize_operands): Try to reuse the results for
4159 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4161 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4162 Add {q} suffix to insn mnemonic.
4164 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4166 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4167 (msp430_warn_func_return): New.
4169 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4171 * fold-const.c (tree_nonzero_bits): New function.
4172 * fold-const.h (tree_nonzero_bits): Likewise.
4173 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4174 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4176 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4180 * varasm.c (assemble_alias): Check ifunc_resolver only on
4183 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4186 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4187 when memory input operand is handled.
4189 2018-05-24 Luis Machado <luis.machado@linaro.org>
4191 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4193 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4195 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4197 * match.pd: Delay FMA folds until after vectorization.
4199 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4202 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4203 address check not strict.
4205 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4207 * gimple-match.h (gimple_match_op): New class.
4208 (mprts_hook): Replace parameters with a gimple_match_op *.
4209 (maybe_build_generic_op): Likewise.
4210 (gimple_simplified_result_is_gimple_val): Replace parameters with
4211 a const gimple_match_op *.
4212 (gimple_simplify): Replace code_helper * and tree * parameters with
4213 a gimple_match_op * parameter.
4214 (gimple_resimplify1): Replace code_helper *, tree and tree *
4215 parameters with a gimple_match_op * parameter.
4216 (gimple_resimplify2): Likewise.
4217 (gimple_resimplify3): Likewise.
4218 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4219 parameters with a gimple_match_op * parameter.
4220 * gimple-match-head.c (gimple_simplify): Change prototypes of
4221 auto-generated functions to take a gimple_match_op * instead of
4222 separate code_helper * and tree * parameters. Make the same
4223 change in the top-level overload and update calls to the
4224 gimple_resimplify routines. Update calls to the auto-generated
4225 functions and to maybe_push_res_to_seq in the publicly-facing
4226 operation-specific gimple_simplify overloads.
4227 (gimple_match_op::MAX_NUM_OPS): Define.
4228 (gimple_resimplify1): Replace rcode and ops with a single res_op
4229 parameter. Update call to gimple_simplify.
4230 (gimple_resimplify2): Likewise.
4231 (gimple_resimplify3): Likewise.
4232 (mprts_hook): Replace parameters with a gimple_match_op *.
4233 (maybe_build_generic_op): Likewise.
4234 (build_call_internal): Replace type, nargs and ops with
4235 a gimple_match_op *.
4236 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4237 with a single gimple_match_op *. Update calls to mprts_hook,
4238 build_call_internal and gimple_simplified_result_is_gimple_val.
4239 Factor out code that is common to the tree_code and combined_fn cases.
4240 * genmatch.c (expr::gen_transform): Replace tem_code and
4241 tem_ops with a gimple_match_op called tem_op. Update calls
4242 to the gimple_resimplify functions and maybe_push_res_to_seq.
4243 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4244 res_ops. Update call to the gimple_resimplify functions.
4245 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4246 (decision_tree::gen): Make the functions take a gimple_match_op *
4247 called res_op instead of separate res_code and res_ops parameters.
4248 Update call accordingly.
4249 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4250 and ops with a single res_op parameter. Update calls to
4251 maybe_build_generic_op and maybe_push_res_to_seq.
4252 (fold_stmt_1): Update calls to gimple_simplify and
4253 replace_stmt_with_simplification.
4254 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4255 and gimple_simplified_result_is_gimple_val.
4256 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4258 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4259 with a gimple_match_op *.
4260 (vn_nary_build_or_lookup): Likewise. Update call to
4261 vn_nary_build_or_lookup_1.
4262 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4263 gimple_match_op *. Update calls to the gimple_resimplify routines
4264 and to gimple_simplified_result_is_gimple_val.
4265 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4266 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4267 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4268 (visit_nary_op): Likewise.
4269 (visit_reference_op_load): Likewise.
4271 2018-05-23 Luis Machado <luis.machado@linaro.org>
4273 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4274 modifier for printing the step amount.
4276 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4279 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4282 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4284 * doc/sourcebuild.texi (Endianness): New subsubsection.
4286 2018-05-23 Luis Machado <luis.machado@linaro.org>
4288 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4289 <prefetch_dynamic_strides>: New const bool field.
4290 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4291 prefetch_dynamic_strides.
4292 (exynosm1_prefetch_tune): Likewise.
4293 (thunderxt88_prefetch_tune): Likewise.
4294 (thunderx_prefetch_tune): Likewise.
4295 (thunderx2t99_prefetch_tune): Likewise.
4296 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4298 (aarch64_override_options_internal): Update to set
4299 PARAM_PREFETCH_DYNAMIC_STRIDES.
4300 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4301 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4302 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4303 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4304 prefetch-dynamic-strides setting.
4306 2018-05-23 Luis Machado <luis.machado@linaro.org>
4308 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4309 <minimum_stride>: New const int field.
4310 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4311 minimum_stride field defaulting to -1.
4312 (exynosm1_prefetch_tune): Likewise.
4313 (thunderxt88_prefetch_tune): Likewise.
4314 (thunderx_prefetch_tune): Likewise.
4315 (thunderx2t99_prefetch_tune): Likewise.
4316 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4317 <default_opt_level>: Set to 3.
4318 (aarch64_override_options_internal): Update to set
4319 PARAM_PREFETCH_MINIMUM_STRIDE.
4320 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4321 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4322 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4323 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4324 stride is constant and is below the minimum stride threshold.
4326 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4328 * config/arm/arm-cpus.in (mode26): Delete.
4329 (armv4): Delete mode26 reference.
4330 * config/arm/arm.c (arm_configure_build_target): Delete use of
4333 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4335 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4337 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4338 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4340 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4341 (floatunsdidf2): Ditto.
4343 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4345 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4346 (fixuns_trunc<mode>si2_avx512f): Ditto.
4347 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4348 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4349 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4351 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4353 PR rtl-optimization/79985
4354 * df-scan.c (df_insn_refs_collect): Remove special case for
4355 global registers and asm statements.
4357 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4359 * extend.texi (Global Register Variables): Rewrite the bullet list.
4360 Note that the register is available for allocation. Note that access
4361 via inline asm must use constraints. Add note about async-signal
4362 handlers. Remove paragraph about automagic register selection.
4364 2018-05-23 Richard Biener <rguenther@suse.de>
4366 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4367 of fixed offset from memset VN.
4369 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4371 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4373 (alloc_cand_and_find_basis): Initialize first_interp field.
4374 (slsr_process_mul): Modify first_interp field.
4375 (slsr_process_add): Likewise.
4376 (slsr_process_cast): Modify first_interp field for each new
4378 (slsr_process_copy): Likewise.
4379 (dump_candidate): Dump first_interp field.
4380 (replace_mult_candidate): Process all interpretations, not just
4382 (replace_rhs_if_not_dup): Likewise.
4383 (replace_one_candidate): Likewise.
4385 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4387 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4389 (aarch64_needs_frame_chain): New function.
4390 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4392 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4395 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4396 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4397 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4398 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4399 as true for strict-align.
4400 (aarch64_can_inline_p): Perform checks even when callee has no
4401 attributes to check for strict alignment.
4402 * doc/extend.texi (AArch64 Function Attributes): Document
4404 * doc/invoke.texi: (AArch64 Options): Likewise.
4406 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4408 PR tree-optimization/85853
4409 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4410 the handling of the root of the node to...
4411 (vect_slp_analyze_node_operations_1): ...this new function,
4412 and run the whole thing with the child nodes' def types
4413 set according to their SLP node's def type.
4415 2018-05-23 Richard Biener <rguenther@suse.de>
4418 * tree-data-ref.c (create_runtime_alias_checks): Defer
4419 and ignore overflow warnings.
4421 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4423 PR tree-optimization/85822
4424 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4427 2018-05-23 Richard Biener <rguenther@suse.de>
4429 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4430 memset constants via native_interpret_expr.
4432 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4435 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4437 (cgraph_node::create_alias): Likewise.
4438 (cgraph_node::get_availability): Check ifunc_resolver instead
4439 of looking up ifunc attribute.
4440 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4441 * varasm.c (do_assemble_alias): Likewise.
4442 (assemble_alias): Likewise.
4443 (default_binds_local_p_3): Likewise.
4444 * cgraph.h (cgraph_node): Add ifunc_resolver.
4445 (cgraph_node::only_called_directly_or_aliased_p): Return false
4447 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4449 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4450 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4451 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4452 instead of looking up ifunc attribute.
4454 2018-05-22 Luis Machado <luis.machado@linaro.org>
4456 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4458 2018-05-22 Martin Sebor <msebor@redhat.com>
4461 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4462 only when expasion succeeds.
4463 (expand_builtin_strcmp): Same.
4464 (expand_builtin_strncmp): Same.
4466 2018-05-22 Martin Sebor <msebor@redhat.com>
4468 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4470 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4471 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4473 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4474 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4476 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4477 Add check for legitimate_address.
4478 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4479 (aarch64_swap_ldrstr_operands): New.
4480 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4483 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4484 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4486 * config/aarch64/aarch64.md: New patterns to generate stp
4488 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4489 single words and double words.
4490 (load_pair_sw, load_pair_dw): Likewise.
4491 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4493 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4495 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4496 for different mode ldpstp and add peephole for merged zero stores.
4498 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4500 (aarch64_gen_store_pair): Rename calls to match new patterns.
4501 (aarch64_gen_load_pair): Rename calls to match new patterns.
4502 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4503 (load_pair<DREG:mode><DREG2:mode>): ... This.
4504 (store_pair<mode>): Rename to...
4505 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4506 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4508 (V_INT_EQUIV): Handle SImode.
4509 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4512 2018-05-22 Martin Sebor <msebor@redhat.com>
4515 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4516 or ajust the presumed bound on an operation to avoid unnecessary
4519 2018-05-22 Martin Sebor <msebor@redhat.com>
4521 PR tree-optimization/85826
4522 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4523 assuming that a DECL necesarily has a constant size.
4525 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4528 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4530 2018-05-22 Richard Biener <rguenther@suse.de>
4532 PR tree-optimization/85834
4533 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4534 non-constant and non-zero memset arguments.
4536 2018-05-22 Martin Liska <mliska@suse.cz>
4539 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4541 2018-05-22 Richard Biener <rguenther@suse.de>
4543 PR tree-optimization/85863
4544 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4545 comparisons when vectype is specified.
4546 (vectorizable_condition): Do not specify vectype for
4547 vect_is_simple_cond when SLP vectorizing.
4549 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4552 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4553 define __ibm128 as long double.
4554 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4555 as a distinct type when IEEE 128-bit support is enabled.
4556 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4557 128-bit types to use the correct functions.
4558 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4559 convert between 128-bit floating point types that have different
4560 modes but the same representation, instead of using gen_lowpart to
4562 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4564 (IFKF_reg): New attributes to give the register constraints for
4566 (extend<mode>tf2_internal): New insns to mark an explicit
4567 conversion between 128-bit floating point types that have a
4568 different mode but share the same representation.
4570 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4572 PR tree-optimization/85814
4573 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4574 a null return from get_strinfo when unsharing the next
4575 strinfo in the chain.
4577 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4580 * varasm.c (weak_finish): Clean up weak_decls.
4582 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4584 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4585 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4586 UNSPEC_UADALP values.
4587 * config/aarch64/iterators.md (ABAL): New int iterator.
4590 (sur): Add mappings for the above.
4591 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4593 (aarch64_<sur>abal<mode>_4): Likewise.
4594 (aarch64_<sur>adalp<mode>_3): Likewise.
4595 (<sur>sadv16qi): New define_expand.
4597 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4599 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4600 (*movdf_internal): Ditto.
4601 (*rcpsf2_sse): Ditto.
4602 (*rsqrtsf2_sse): Ditto.
4603 (*sqrt<mode>2_sse): Ditto.
4605 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4607 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4609 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4610 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4611 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4612 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4614 * config/aarch64/arm_neon.h: Likewise.
4615 * config/aarch64/iterators.md (VQ_I): New.
4617 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4619 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4621 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4624 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4626 * config/nds32/constraints.md (S): New constraint.
4627 * config/nds32/nds32.md (call_internal): Use constraint S.
4628 (call_value_internal): Likewise.
4629 (sibcall_internal): Likewise.
4630 (sibcall_value_internal): Likewise.
4632 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4633 Chung-Ju Wu <jasonwucj@gmail.com>
4635 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4638 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4639 Chung-Ju Wu <jasonwucj@gmail.com>
4641 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4642 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4643 (nds32_rtx_costs_impl): Simplify.
4644 (nds32_address_cost_impl): Simplify.
4645 (nds32_init_rtx_costs): New function.
4646 (nds32_rtx_costs_speed_prefer): Likewise.
4647 (nds32_rtx_costs_size_prefer): Likewise.
4648 (nds32_address_cost_speed_prefer): Likewise.
4649 (nds32_address_cost_speed_fwprop): Likewise.
4650 (nds32_address_cost_size_prefer): Likewise.
4651 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4652 * config/nds32/nds32.c (nds32_option_override): Use
4653 nds32_init_rtx_costs function.
4655 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4657 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4658 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4659 (TARGET_PIPELINE_N8): Likewise.
4660 (TARGET_PIPELINE_N10): Likewise.
4661 (TARGET_PIPELINE_N13): Likewise.
4662 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4664 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4666 * config/nds32/nds32-fpu.md: Update copyright year.
4668 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4670 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4672 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4674 * config/nds32/nds32.c
4675 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4676 * config/nds32/nds32.opt (minline-asm-r15): New option.
4678 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4680 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4682 * config/nds32/nds32.md (abssi2): New pattern.
4684 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4686 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4687 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4688 Merge insn pattern from sse_cvtsi2ss<round_name> and
4689 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4690 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4691 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4692 using SWI48 mode iterator.
4693 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4694 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4695 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4696 pattern from sse_cvttss2si<round_saeonly_name>
4697 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4698 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4699 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4700 using SWI48 mode iterator.
4701 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4702 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4703 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4704 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4705 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4706 using SWI48 mode iterator.
4707 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4708 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4709 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4710 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4711 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4712 SWI48 mode iterator.
4713 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4714 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4715 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4716 pattern from sse_cvttsd2si<round_saeonly_name>
4717 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4719 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4721 * config/nds32/nds32-md-auxiliary.c
4722 (nds32_valid_smw_lwm_base_p): Refine.
4723 (nds32_output_smw_single_word): Refine.
4724 (nds32_output_smw_double_word): New.
4725 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4727 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4729 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4730 (nds32_output_stack_pop): Refine.
4731 (nds32_expand_unaligned_load): Refine.
4732 (nds32_expand_unaligned_store): Refine.
4734 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4735 Chung-Ju Wu <jasonwucj@gmail.com>
4737 * config/nds32/constants.md: Add TP_REGNUM constant.
4738 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4739 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4741 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4742 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4743 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4744 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4745 and PIC code generation.
4746 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4748 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4750 * config/nds32/nds32.md: Support TLS and PIC.
4751 * config/nds32/nds32.c: Support TLS and PIC.
4752 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4753 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4756 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4758 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4759 mode with E_ prefix.
4761 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4762 Chung-Ju Wu <jasonwucj@gmail.com>
4764 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4765 * config/nds32/nds32-md-auxiliary.c
4766 (symbolic_reference_mentioned_p): New.
4767 (nds32_legitimize_ict_address): New.
4768 (nds32_expand_ict_move): New.
4769 (nds32_indirect_call_referenced_p): New.
4770 (nds32_symbol_binds_local_p): Delete.
4771 (nds32_long_call_p): Modify.
4772 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4773 * config/nds32/nds32-protos.h
4774 (symbolic_reference_mentioned_p): Declare.
4775 (nds32_legitimize_ict_address): Declare.
4776 (nds32_expand_ict_move): Declare.
4777 (nds32_indirect_call_referenced_p): Declare.
4778 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4779 (nds32_relax_group): Use nds32_ict_const_p as condition.
4780 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4781 (nds32_asm_file_start): Output ict_model directive in asm code.
4782 (nds32_legitimate_address_p): Consider indirect call.
4783 (nds32_print_operand): Consider indirect call.
4784 (nds32_print_operand_address): Consider indirect call.
4785 (nds32_insert_attributes): Handle "indirect_call" attribute.
4786 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4787 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4788 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4789 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4790 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4791 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4792 (TARGET_ICT_MODEL_SMALL): Define.
4793 (TARGET_ICT_MODEL_LARGE): Define.
4794 * config/nds32/nds32.md (movsi): Consider ict model.
4795 (call, call_value): Consider ict model.
4796 (sibcall, sibcall_value): Consider ict model.
4797 * config/nds32/nds32.opt (mict-model): New option.
4798 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4801 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4802 Monk Chiang <sh.chiang04@gmail.com>
4803 Jim Wilson <jimw@sifive.com>
4805 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4806 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4807 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4808 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4809 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4810 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4811 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4812 compute save_libcall_adjustment properly.
4813 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4814 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4815 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4816 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4817 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4818 (ABI_SPEC): Handle mabi=ilp32e.
4819 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4820 (RVE): Add RVE mask.
4821 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4822 <-march>: Add rv32e as an example.
4824 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4827 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4828 (intra_create_variable_infos): Handle C++ constructors.
4830 2018-05-18 Martin Liska <mliska@suse.cz>
4832 * passes.def: Remove a redundant pass.
4834 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4837 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4839 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4841 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4843 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4845 (armv2, armv2a, armv3, armv3m): Delete architectures.
4846 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4847 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4848 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4850 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4851 (*mulsidi3adddi): Likewise.
4852 (mulsidi3): Likewise.
4853 (*mulsidi3_nov6): Likewise.
4854 (umulsidi3): Likewise.
4855 (umulsidi3_nov6): Likewise.
4856 (umaddsidi4): Likewise.
4857 (*umulsidi3adddi): Likewise.
4858 (smulsi3_highpart): Likewise.
4859 (*smulsi3_highpart_nov6): Likewise.
4860 (umulsi3_highpart): Likewise.
4861 (*umulsi3_highpart_nov6): Likewise.
4862 * config/arm/arm.h (arm_arch3m): Delete.
4863 * config/arm/arm.c (arm_arch3m): Delete.
4864 (arm_option_override_internal): Update armv3-related comment.
4865 (arm_configure_build_target): Delete use of isa_bit_mode32.
4866 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4867 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4868 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4870 (mulusa3): Likewise.
4871 * config/arm/arm-protos.h (arm_arch3m): Delete.
4872 * config/arm/arm-tables.opt: Regenerate.
4873 * config/arm/arm-tune.md: Likewise.
4874 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4875 deleted architectures.
4877 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4879 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4880 (armv5t, armv5te): New features.
4881 (ARMv5, ARMv5e): Delete fgroups.
4882 (ARMv5t, ARMv5te): Adjust for above changes.
4884 (armv5, armv5e): Delete arches.
4885 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4887 (*call_reg_arm): Likewise.
4888 (*call_value_reg_armv5): Likewise.
4889 (*call_value_reg_arm): Likewise.
4890 (*call_symbol): Likewise.
4891 (*call_value_symbol): Likewise.
4892 (*sibcall_insn): Likewise.
4893 (*sibcall_value_insn): Likewise.
4895 (prefetch): Likewise.
4896 (define_split and define_peephole2 dependent on arm_arch5):
4898 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4900 (TARGET_ARM_QBIT): Likewise.
4901 (TARGET_DSP_MULTIPLY): Likewise.
4902 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4903 (arm_arch5, arm_arch5e): Delete.
4904 (arm_arch5t, arm_arch5te): Declare.
4905 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4906 (arm_arch5t): Declare.
4907 (arm_option_reconfigure_globals): Update for the above.
4908 (arm_options_perform_arch_sanity_checks): Update comment, replace
4909 use of arm_arch5 with arm_arch5t.
4910 (use_return_insn): Likewise.
4911 (arm_emit_call_insn): Likewise.
4912 (output_return_instruction): Likewise.
4913 (arm_final_prescan_insn): Likewise.
4914 (arm_coproc_builtin_available): Likewise.
4915 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4916 arm_arch5e with arm_arch5t and arm_arch5te.
4917 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4918 (arm_arch5t, arm_arch5te): Declare.
4919 * config/arm/arm-tables.opt: Regenerate.
4920 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4921 * config/arm/t-multilib: Likewise.
4922 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4923 instead of arm_arch5.
4924 (*call_reg_thumb1): Likewise.
4925 (*call_value_reg_thumb1_v5): Likewise.
4926 (*call_value_reg_thumb1): Likewise.
4927 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4929 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4931 2018-05-18 Martin Liska <mliska@suse.cz>
4933 PR gcov-profile/84846
4934 * doc/gcov.texi: Document -t option of gcov tool.
4936 2018-05-18 Martin Liska <mliska@suse.cz>
4938 PR gcov-profile/84846
4939 * gcov.c (print_usage): Add new -t option.
4940 (process_args): Handle the option.
4941 (generate_results): Use stdout as output when requested by
4944 2018-05-18 Martin Liska <mliska@suse.cz>
4946 PR gcov-profile/84846
4947 * coverage.c (coverage_init): Write PWD to .gcno file.
4948 * doc/gcov.texi: Document how working directory is printed.
4949 * gcov-dump.c (dump_gcov_file): Print PWD.
4950 * gcov.c (output_intermediate_file): Likewise.
4951 (read_graph_file): Read PWD string.
4952 (output_lines): Print PWD.
4954 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4957 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4958 for retval and return false if all args to phi are zero.
4960 2018-05-18 Richard Biener <rguenther@suse.de>
4962 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4964 (evrp_dom_walker::before_dom_children): Call it.
4966 2018-05-18 Richard Biener <rguenther@suse.de>
4968 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4969 results when processing array refs with variable index.
4971 2018-05-18 Toon Moene <toon@moene.org>
4973 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4974 directly after that of -floop-interchange. Indicate that both
4975 options are enabled by default when specifying -O3.
4977 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4979 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4980 iterator. Delete separate integer-mode vec_set<mode> expander.
4981 (aarch64_simd_vec_setv2di): Delete.
4982 (vec_setv2di): Delete.
4983 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4984 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4985 the "w, r" alternative.
4987 2018-05-18 Martin Liska <mliska@suse.cz>
4989 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4990 * tree-pass.h (make_pass_lower_switch_O0): New function.
4991 * tree-switch-conversion.c (node_has_low_bound): Remove.
4992 (node_has_high_bound): Likewise.
4993 (node_is_bounded): Likewise.
4994 (class pass_lower_switch): Make it a template type and create
4996 (pass_lower_switch::execute): Add template argument.
4997 (make_pass_lower_switch): New function.
4998 (make_pass_lower_switch_O0): New function.
4999 (do_jump_if_equal): Remove.
5000 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5001 all the hard work to tree optimization passes.
5003 2018-05-18 Martin Liska <mliska@suse.cz>
5005 * dbgcnt.c (limit_low): Renamed from limit.
5006 (limit_high): New variable.
5007 (dbg_cnt_is_enabled): Check for upper limit.
5008 (dbg_cnt): Adjust dumping.
5009 (dbg_cnt_set_limit_by_index): Add new argument for high
5011 (dbg_cnt_set_limit_by_name): Likewise.
5012 (dbg_cnt_process_single_pair): Parse new format.
5013 (dbg_cnt_process_opt): Use strtok.
5014 (dbg_cnt_list_all_counters): Remove 'value' and add
5016 * doc/invoke.texi: Document changes.
5018 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5020 * doc/sourcebuild.texi (scalar_all_fma): Document.
5021 * tree.def (FMA_EXPR): Delete.
5022 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5023 * internal-fn.c (ternary_direct): New macro.
5024 (expand_ternary_optab_fn): Likewise.
5025 (direct_ternary_optab_supported_p): Likewise.
5026 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5027 * builtins.c (fold_builtin_fma): Delete.
5028 (fold_builtin_3): Don't call it.
5029 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5030 * expr.c (expand_expr_real_2): Likewise.
5031 * fold-const.c (operand_equal_p): Likewise.
5032 (fold_ternary_loc): Likewise.
5033 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5034 * gimple.c (DEFTREECODE): Likewise.
5035 * gimplify.c (gimplify_expr): Likewise.
5036 * optabs-tree.c (optab_for_tree_code): Likewise.
5037 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5038 * tree-eh.c (operation_could_trap_p): Likewise.
5039 (stmt_could_throw_1_p): Likewise.
5040 * tree-inline.c (estimate_operator_cost): Likewise.
5041 * tree-pretty-print.c (dump_generic_node): Likewise.
5042 (op_code_prio): Likewise.
5043 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5044 * tree-ssa-operands.c (get_expr_operands): Likewise.
5045 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5046 * fold-const-call.h (fold_fma): Delete.
5047 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5048 CFN_FNMA and CFN_FNMS.
5050 * genmatch.c (combined_fn): New enum.
5051 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5052 (commutative_op): New function.
5053 (commutate): Use it. Handle more than 2 operands.
5054 (dt_operand::gen_gimple_expr): Use commutative_op.
5055 (parser::parse_expr): Allow :c to be used with non-binary
5056 operators if the commutative operand is known.
5057 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5058 CFN_FMS, CFN_FNMA and CFN_FNMS.
5059 (backprop::process_assign_use): Remove FMA_EXPR handling.
5060 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5061 (gen_hsa_fma): New function.
5062 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5063 IFN_FNMA and IFN_FNMS.
5064 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5065 * gimple-fold.h (follow_all_ssa_edges): Declare.
5066 * gimple-fold.c (follow_all_ssa_edges): New function.
5067 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5068 gimple_build interface and use follow_all_ssa_edges to fold the result.
5069 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5070 instead of checking for optabs directly.
5071 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5072 rather than FMA_EXPRs.
5073 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5074 call to IFN_FMA instead of an FMA_EXPR.
5076 2018-05-17 Jim Wilson <jimw@sifive.com>
5078 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5079 sign extended promoted subreg, and the range does not have the sign bit
5080 set, then do a sign extend.
5082 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5083 test, check for sign extended subreg and/or constant operands, and
5084 do a sign extend in that case.
5086 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5088 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5089 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5091 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5092 Change logics_shift_reg to logics_shift_imm.
5093 (thunderx2t99_fp_loadpair_basic): Delete.
5094 (thunderx2t99_fp_storepair_basic): Delete.
5095 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5096 (thunderx2t99_asimd_polynomial): Delete.
5097 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5098 and neon_fp_mul_d_scalar_q.
5099 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5100 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5101 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5102 (thunderx2t99_asimd_lut): Add missing tbl types.
5103 (thunderx2t99_asimd_ext): Delete.
5104 (thunderx2t99_asimd_load1_1_mult): Delete.
5105 (thunderx2t99_asimd_load1_2_mult): Delete.
5106 (thunderx2t99_asimd_load1_ldp): New.
5107 (thunderx2t99_asimd_load1): New.
5108 (thunderx2t99_asimd_load2): Add missing *load2* types.
5109 (thunderx2t99_asimd_load3): New.
5110 (thunderx2t99_asimd_load4): New.
5111 (thunderx2t99_asimd_store1_1_mult): Delete.
5112 (thunderx2t99_asimd_store1_2_mult): Delete.
5113 (thunderx2t99_asimd_store2_mult): Delete.
5114 (thunderx2t99_asimd_store2_onelane): Delete.
5115 (thunderx2t99_asimd_store_stp): New.
5116 (thunderx2t99_asimd_store1): New.
5117 (thunderx2t99_asimd_store2): New.
5118 (thunderx2t99_asimd_store3): New.
5119 (thunderx2t99_asimd_store4): New.
5121 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5123 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5124 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5126 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5127 Segher Boessenkool <segher@kernel.crashing.org>
5130 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5133 2018-05-17 Richard Biener <rguenther@suse.de>
5135 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5136 for pruning loop and prune defs feeding only already visited PHIs.
5138 2018-05-17 Richard Biener <rguenther@suse.de>
5140 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5142 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5143 Richard Biener <rguenther@suse.de>
5145 PR tree-optimization/85793
5146 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5147 for VMAT_ELEMENTWISE.
5149 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5151 * internal-fn.h (lookup_internal_fn): Declare
5152 * internal-fn.c (lookup_internal_fn): New function.
5153 * gimple.c (gimple_build_call_from_tree): Handle calls to
5155 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5156 internal function names.
5157 * tree-pretty-print.c (dump_generic_node): Likewise.
5158 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5160 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5162 * gimple-fold.h (gimple_build): Make the function forms take
5163 combined_fn rather than built_in_function.
5164 (gimple_simplify): Likewise.
5165 * gimple-match-head.c (gimple_simplify): Likewise.
5166 * gimple-fold.c (gimple_build): Likewise.
5167 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5168 rather than gimple_build_call_internal.
5169 (get_initial_defs_for_reduction): Likewise.
5170 (vect_create_epilog_for_reduction): Likewise.
5171 (vectorizable_live_operation): Likewise.
5173 2018-05-17 Martin Liska <mliska@suse.cz>
5175 * gimple-ssa-sprintf.c (format_directive): Do not use
5176 space in between 'G_' and '('.
5178 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5181 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5182 even if the mask is not all ones.
5185 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5187 (ix86_gimple_fold_builtin): Likewise.
5190 * config/i386/i386.c: Include tree-vector-builder.h.
5191 (ix86_vector_shift_count): New function.
5192 (ix86_fold_builtin): Fold shift builtins by scalar count.
5193 (ix86_gimple_fold_builtin): Likewise.
5195 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5196 _mm512_setzero): New intrinsics.
5198 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5199 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5201 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5202 code generation for cases where splatting a value is not useful.
5203 * simplify-rtx.c (simplify_ternary_operation): Simplify
5204 vec_merge across a vec_duplicate and a paradoxical subreg forming
5205 a vector mode to a vec_concat.
5207 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5209 * config.gcc: Support "goldmont-plus".
5210 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5212 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5213 PROCESSOR_GOLDMONT_PLUS.
5214 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5215 (processor_target_table): Add "goldmont-plus".
5216 (PTA_GOLDMONT_PLUS): Define.
5217 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5218 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5219 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5220 (fold_builtin_cpu): Add "goldmont-plus".
5221 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5222 (ix86_option_override_internal): Add "goldmont-plus".
5223 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5224 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5225 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5226 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5228 2018-05-17 Richard Biener <rguenther@suse.de>
5230 PR tree-optimization/85757
5231 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5232 remove defs that only feed that PHI from further processing.
5234 2018-05-16 Jim Wilson <jimw@sifive.com>
5236 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5238 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5240 2018-05-16 Mark Wielaard <mark@klomp.org>
5242 * dwarf2out.c (count_index_strings): New function.
5243 (output_indirect_strings): Call count_index_strings and generate
5244 header for dwarf_version >= 5.
5246 2018-05-16 Mark Wielaard <mark@klomp.org>
5248 * dwarf2out.c (dwarf_FORM): New function.
5249 (set_indirect_string): Use dwarf_FORM.
5250 (reset_indirect_string): Likewise.
5251 (size_of_die): Likewise.
5252 (value_format): Likewise.
5253 (output_die): Likewise.
5254 (add_skeleton_AT_string): Likewise.
5255 (output_macinfo_op): Likewise.
5256 (index_string): Likewise.
5257 (output_index_string_offset): Likewise.
5258 (output_index_string): Likewise.
5259 (count_index_strings): Likewise.
5261 2018-05-16 Carl Love <cel@us.ibm.com>
5263 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5264 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5266 2018-05-16 Martin Jambor <mjambor@suse.cz>
5268 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5269 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5271 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5273 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5274 (fnma<mode>4): Likewise.
5275 (fms<mode>4): Likewise.
5276 (fnms<mode>4): Likewise.
5277 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5278 (aarch64_fnma<mode>4): Likewise.
5279 (aarch64_fms<mode>4): Likewise.
5280 (aarch64_fnms<mode>4): Likewise.
5281 (aarch64_fnmadd<mode>4): Likewise.
5283 2018-05-16 Jason Merrill <jason@redhat.com>
5285 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5287 2018-05-16 Richard Biener <rguenther@suse.de>
5289 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5290 (dump_stmt_cost): Declare.
5291 (add_stmt_cost): Dump cost we add.
5292 (add_stmt_costs): New function.
5293 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5295 (vect_analyze_stmt): Adjust prototype.
5296 (vectorizable_condition): Likewise.
5297 (vectorizable_live_operation): Likewise.
5298 (vectorizable_reduction): Likewise.
5299 (vectorizable_induction): Likewise.
5300 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5301 cost vector to pass to vectorizable_ and record afterwards.
5302 (vect_model_reduction_cost): Take cost vector argument and adjust.
5303 (vect_model_induction_cost): Likewise.
5304 (vectorizable_reduction): Likewise.
5305 (vectorizable_induction): Likewise.
5306 (vectorizable_live_operation): Likewise.
5307 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5308 SLP_TREE_NUMBER_OF_VEC_STMTS.
5309 (vect_analyze_slp_cost_1): Remove.
5310 (vect_analyze_slp_cost): Likewise.
5311 (vect_slp_analyze_node_operations): Take visited args and
5312 a target cost vector. Avoid processing already visited stmt sets.
5313 (vect_slp_analyze_operations): Use a local cost vector to gather
5314 costs and register those of non-discarded instances.
5315 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5316 (vect_schedule_slp_instance): Remove copying of
5317 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5319 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5320 adding cost. Record cost entry location.
5321 (vect_prologue_cost_for_slp_op): Function to compute cost of
5322 a constant or invariant generated for SLP vect in the prologue,
5323 split out from vect_analyze_slp_cost_1.
5324 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5325 (vect_model_promotion_demotion_cost): Likewise.
5326 (vect_model_store_cost): Likewise, make static.
5327 (vect_model_load_cost): Likewise.
5328 (vectorizable_bswap): Add cost vector arg and adjust.
5329 (vectorizable_call): Likewise.
5330 (vectorizable_simd_clone_call): Likewise.
5331 (vectorizable_conversion): Likewise.
5332 (vectorizable_assignment): Likewise.
5333 (vectorizable_shift): Likewise.
5334 (vectorizable_operation): Likewise.
5335 (vectorizable_store): Likewise.
5336 (vectorizable_load): Likewise.
5337 (vectorizable_condition): Likewise.
5338 (vectorizable_comparison): Likewise.
5339 (can_vectorize_live_stmts): Likewise.
5340 (vect_analyze_stmt): Likewise.
5341 (vect_transform_stmt): Adjust calls to vectorizable_*.
5342 * tree-vectorizer.c: Include gimple-pretty-print.h.
5343 (dump_stmt_cost): New function.
5345 2018-05-16 Richard Biener <rguenther@suse.de>
5347 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5348 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5349 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5350 (check_name): New callback.
5351 (dse_classify_store): Track cycles via a visited bitmap of PHI
5352 defs and simplify handling of in-loop and across loop dead stores
5353 and properly fail for loop-variant refs. Handle byte-tracking with
5354 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5357 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5359 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5360 (vect_get_mask_type_for_stmt): Likewise.
5361 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5363 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5364 to determine the statement's vector type and the vector type that
5365 should be used for calculating nunits. Deal with cases in which
5366 the type has to be deferred.
5367 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5368 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5369 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5370 (vect_determine_vf_for_stmt): New functions, split out from...
5371 (vect_determine_vectorization_factor): ...here.
5372 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5373 (vect_get_mask_type_for_stmt): New functions, split out from
5374 vect_determine_vectorization_factor.
5376 2018-05-16 Richard Biener <rguenther@suse.de>
5378 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5379 verify the [VEC_]COND_EXPR embedded comparison.
5381 2018-05-15 Martin Sebor <msebor@redhat.com>
5383 PR tree-optimization/85753
5384 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5385 RECORD_TYPE in addition to ARRAY_TYPE.
5387 2018-05-15 Martin Sebor <msebor@redhat.com>
5390 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5392 2018-05-15 Richard Biener <rguenther@suse.de>
5394 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5395 add by_clobber_p one. Change algorithm to collect all defs
5396 representing uses we need to walk and try reducing them to
5397 a single one before failing.
5398 (dse_dom_walker::dse_optimize_stmt): Adjust.
5400 2018-05-13 Mark Wielaard <mark@klomp.org>
5402 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5403 (size_of_loc_descr): Likewise.
5404 (output_loc_operands): Likewise.
5405 (output_loc_operands_raw): Likewise.
5406 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5407 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5408 (hash_loc_operands): Likewise.
5409 (compare_loc_operands): Likewise.
5411 2018-05-14 Mark Wielaard <mark@klomp.org>
5413 * dwarf2out.c (count_index_addrs): New function.
5414 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5416 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5418 PR tree-optimization/83648
5419 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5420 return value as malloc candidate.
5422 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5425 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5426 param as true in call to suggest_attribute.
5428 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5430 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5431 -mreadonly-in-sdata.
5433 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5435 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5437 (aarch64_crypto_aesd_fused): Likewise.
5439 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5441 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5442 (movsi_aarch64): Likewise.
5443 (load_pairsi): Likewise.
5444 (load_pairdi): Likewise.
5445 (store_pairsi): Likewise.
5446 (store_pairdi): Likewise.
5447 (load_pairsf): Likewise.
5448 (load_pairdf): Likewise.
5449 (store_pairsf): Likewise.
5450 (store_pairdf): Likewise.
5451 (zero_extend): Likewise.
5452 (trunc): Swap alternatives.
5453 (fcvt_target): Add '?' to prefer w over r.
5455 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5458 * config/i386/i386.md: Disallow non-commutative arithmetics in
5459 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5460 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5461 in the peephole2 before it.
5463 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5465 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5466 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5467 (ix86_handle_option): Handle -mcldemote.
5468 * config.gcc: New header.
5469 * config/i386/cldemoteintrin.h: New file.
5470 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5471 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5473 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5474 OPTION_MASK_ISA_CLDEMOTE.
5475 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5476 (ix86_valid_target_attribute_inner_p): Ditto.
5477 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5478 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5479 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5480 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5481 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5483 * config/i386/i386.opt: Add -mcldemote.
5484 * config/i386/x86intrin.h: New header.
5485 * doc/invoke.texi: Add -mcldemote.
5487 2018-05-14 Richard Biener <rguenther@suse.de>
5489 * doc/match-and-simplify.texi: Adjust :s documentation.
5491 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5493 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5494 intended memcpy size.
5495 (REORDER_45): Likewise.
5497 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5499 * sort.cc: New file.
5500 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5501 * vec.c (qsort_chk): Use gcc_qsort.
5502 * Makefile.in (OBJS-libcommon): Add sort.o.
5503 (build/sort.o): New target. Use it...
5504 (BUILD_RTL): ... here, and...
5505 (build/gencfn-macros): ... here, and...
5506 (build/genmatch): ... here.
5508 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5509 Chung-Ju Wu <jasonwucj@gmail.com>
5511 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5512 * config/nds32/nds32-graywolf.md: New file.
5513 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5514 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5516 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5517 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5518 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5519 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5520 * config/nds32/pipelines.md: Include n15 settings.
5522 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5523 Chung-Ju Wu <jasonwucj@gmail.com>
5525 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5526 * config/nds32/nds32-n13.md: New file.
5527 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5528 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5530 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5531 * config/nds32/nds32.md (pipeline_model): Add n13.
5532 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5533 * config/nds32/pipelines.md: Include n13 settings.
5535 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5536 Chung-Ju Wu <jasonwucj@gmail.com>
5538 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5539 * config/nds32/nds32-n10.md: New file.
5540 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5541 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5543 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5544 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5545 * config/nds32/nds32.md (pipeline_model): Add n10.
5546 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5547 * config/nds32/pipelines.md: Include n10 settings.
5549 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5550 Kito Cheng <kito.cheng@gmail.com>
5551 Chung-Ju Wu <jasonwucj@gmail.com>
5553 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5554 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5555 Add enum values for DSP extension instructions.
5556 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5558 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5559 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5561 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5562 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5563 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5564 * config/nds32/nds32-intrinsic.md: Likewise.
5565 * config/nds32/nds32_intrinsic.h: Likewise.
5566 * config/nds32/nds32-md-auxiliary.c: Likewise.
5567 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5568 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5569 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5570 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5571 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5572 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5573 TYPE_DMAC in switch statement.
5574 * config/nds32/nds32.c: New checking and implementation for DSP
5575 extension instructions.
5576 * config/nds32/nds32.h: Likewise.
5577 * config/nds32/nds32.md: Likewise.
5578 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5579 * config/nds32/predicates.md: Implement new predicates for DSP
5582 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5584 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5585 Reformat alternatives and attributes so it is easier to identify
5586 which constraints/attributes go with which instruction.
5587 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5588 (mov<mode>_softfloat32, FMOVE64): Likewise.
5589 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5590 (mov<mode>_softfloat64, FMOVE64): Likewise.
5592 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5594 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5596 (Basic PowerPC Built-in Functions): The new name of the
5597 subsection previously known as "PowerPC Built-in Functions".
5598 (Basic PowerPC Built-in Functions Available on all Configurations):
5600 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5601 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5602 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5603 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5605 2018-05-11 Martin Jambor <mjambor@suse.cz>
5608 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5611 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5614 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5616 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5618 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5619 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5620 (ix86_handle_option): Handle -mwaitpkg.
5621 * config.gcc: New header.
5622 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5623 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5624 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5626 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5627 OPTION_MASK_ISA_WAITPKG.
5628 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5629 (ix86_option_override_internal): Add PTA_WAITPKG.
5630 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5631 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5632 IX86_BUILTIN_TPAUSE.
5633 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5634 __builtin_ia32_umwait and __builtin_ia32_tpause.
5635 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5636 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5637 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5638 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5639 UNSPECV_TPAUSE): New.
5640 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5641 * config/i386/i386.opt: Add -mwaitpkg.
5642 * config/i386/waitpkgintrin.h: New file.
5643 * config/i386/x86intrin.h: New header.
5644 * doc/invoke.texi: Add -mwaitpkg.
5646 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5649 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5651 (cortex-m0): Use armv6s-m isa.
5652 (cortex-m0plus): Likewise.
5653 (cortex-m1): Likewise.
5654 (cortex-m0.small-multiply): Likewise.
5655 (cortex-m0plus.small-multiply): Likewise.
5656 (cortex-m1.small-multiply): Likewise.
5658 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5659 Jakub Jelinek <jakub@redhat.com>
5661 PR tree-optimization/85692
5662 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5663 source permute as well.
5665 2018-05-11 Martin Liska <mliska@suse.cz>
5668 * doc/extend.texi: Document LLVM style format for no_sanitize
5671 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5673 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5674 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5675 (mode_supports_vsx_dform_quad): Likewise.
5676 (mode_supports_vmx_dform): Move these functions to be next to the
5677 other mode_supports functions.
5678 (mode_supports_dq_form): Likewise.
5679 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5680 mode_supports_dq_form.
5681 (reg_offset_addressing_ok_p): Likewise.
5682 (offsettable_ok_by_alignment): Likewise.
5683 (rs6000_legitimate_offset_address_p): Likewise.
5684 (legitimate_lo_sum_address_p): Likewise.
5685 (rs6000_legitimize_address): Likewise.
5686 (rs6000_legitimize_reload_address): Likewise.
5687 (rs6000_secondary_reload_inner): Likewise.
5688 (rs6000_preferred_reload_class): Likewise.
5689 (rs6000_output_move_128bit): Likewise.
5691 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5693 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5694 Generate SImode target register for null target.
5695 <case IX86_BUILTIN_XGETBV>: Ditto.
5696 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5697 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5699 2018-05-10 Carl Love <cel@us.ibm.com>
5701 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5702 dcbtt and dcbtstt if operands[2] is 0.
5704 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5707 * config/i386/sse.md (usadv64qi): New expander.
5709 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5711 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5712 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5713 -maltivec=be support.
5714 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5715 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5716 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5717 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5718 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5719 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5720 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5721 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5722 altivec_vsumsws): Adjust.
5723 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5724 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5725 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5727 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5728 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5729 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5730 (altivec_lve<VI_char>x): Delete expand.
5731 (*altivec_lve<VI_char>x_internal): Rename to...
5732 (altivec_lve<VI_char>x): ... this.
5733 (altivec_lvxl_<mode>): Delete expand.
5734 (*altivec_lvxl_<mode>_internal): Rename to ...
5735 (altivec_lvxl_<mode>): ... this.
5736 (altivec_stvxl_<mode>): Delete expand.
5737 (*altivec_stvxl_<mode>_internal): Rename to ...
5738 (altivec_stvxl_<mode>): ... this.
5739 (altivec_stve<VI_char>x): Delete expand.
5740 (*altivec_stve<VI_char>x_internal): Rename to ...
5741 (altivec_stve<VI_char>x): ... this.
5742 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5743 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5744 reduc_plus_scal_<mode>): Adjust.
5745 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5747 (rs6000_cpu_cpp_builtins): Adjust.
5748 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5749 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5750 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5751 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5752 -maltivec=be support.
5753 (rs6000_split_vec_extract_var): Adjust.
5754 (rs6000_split_v4si_init): Adjust.
5755 (swap_selector_for_mode): Delete.
5756 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5757 altivec_expand_stvex_be): Delete.
5758 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5759 -maltivec=be support.
5760 (rs6000_gimple_fold_builtin): Ditto.
5761 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5763 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5764 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5765 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5766 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5767 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5768 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5769 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5770 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5771 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5772 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5773 anonymous split): Adjust.
5774 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5775 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5777 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5779 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5780 when --with-gxx-include-dir is also specified.
5781 * configure: Regenerate.
5783 2018-05-09 Jim Wilson <jimw@sifive.com>
5786 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5787 * config/riscv/t-withmultilib: New.
5788 * config/riscv/withmultilib.h: New.
5789 * doc/install.texi: Document RISC-V --with-multilib-list support.
5791 2018-05-09 Richard Biener <rguenther@suse.de>
5793 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5795 (vect_bb_vectorization_profitable_p): Adjust. Compute
5796 actual scalar cost using the cost vector and the add_stmt_cost
5799 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5801 PR rtl-optimization/85645
5802 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5803 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5805 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5807 PR rtl-optimization/85645
5808 * shrink-wrap.c (spread_components): Return a boolean saying if
5809 anything was changed.
5810 (try_shrink_wrapping_separate): Iterate spread_components until
5811 nothing changes anymore.
5813 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5815 PR rtl-optimization/85645
5816 * regrename.c (build_def_use): Also kill the chains that include the
5817 destination of a REG_CFA_REGISTER note.
5819 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5821 PR rtl-optimization/85645
5822 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5823 insn that has a REG_CFA_REGISTER note.
5825 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5827 * cfgexpand.c (expand_clobber): New function.
5828 (expand_gimple_stmt_1): Use it.
5829 * tree-vect-stmts.c (vect_clobber_variable): New function,
5831 (vectorizable_simd_clone_call): ...here.
5832 (vectorizable_store): Emit a clobber either side of an
5833 IFN_STORE_LANES sequence.
5834 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5836 2018-05-09 Tom de Vries <tom@codesourcery.com>
5839 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5840 (define_insn "trap_if_false"): Add exit after trap.
5842 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5844 PR rtl-optimization/85638
5845 * bb-reorder.c: Include common/common-target.h.
5846 (create_forwarder_block): New function extracted from...
5847 (fix_up_crossing_landing_pad): ...here. Rename into...
5848 (dw2_fix_up_crossing_landing_pad): ...this.
5849 (sjlj_fix_up_crossing_landing_pad): New function.
5850 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5851 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5852 from both partitions and exit the loop after one iteration.
5854 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5857 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5859 (Basic PowerPC Built-in Functions): The new name of the
5860 subsection previously known as "PowerPC Built-in Functions".
5861 (Basic PowerPC Built-in Functions Available on all Configurations):
5863 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5865 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5867 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5869 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5872 2018-05-08 Jim Wilson <jimw@sifive.com>
5874 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5875 (LD_EMUL_SUFFIX): New.
5876 (LINK_SPEC): Use it.
5878 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5880 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5882 (Basic PowerPC Built-in Functions): The new name of the
5883 subsection previously known as "PowerPC Built-in Functions".
5884 (Basic PowerPC Built-in Functions Available on all Configurations):
5886 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5888 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5890 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5892 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5895 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5898 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5899 after cmpelim optimization.
5901 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5903 * config.gcc: Support "goldmont".
5904 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5905 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5907 * config/i386/i386.c (m_GOLDMONT): Define.
5908 (processor_target_table): Add "goldmont".
5909 (PTA_GOLDMONT): Define.
5910 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5911 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5912 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5913 (fold_builtin_cpu): Add "goldmont".
5914 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5915 (ix86_option_override_internal): Add "goldmont".
5916 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5917 (processor_type): Add PROCESSOR_GOLDMONT.
5918 * config/i386/i386.md: Add CPU "glm".
5919 * config/i386/glm.md: New file.
5920 * config/i386/x86-tune.def: Add m_GOLDMONT.
5921 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5923 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5926 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5928 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5929 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5930 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5933 * config/i386/i386.c (ix86_fold_builtin): Handle
5934 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5937 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5938 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5940 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5943 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5944 (check_arch): Likewise.
5945 (check_fpu): Return the result rather than printing it.
5946 (end arch): Fix operator precedence.
5947 (end cpu): Likewise.
5948 (END): Print the result from check_fpu.
5950 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5951 Alan Hayward <alan.hayward@arm.com>
5952 David Sherwood <david.sherwood@arm.com>
5954 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5955 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5956 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5957 (*fcmuo<mode>_and): New patterns.
5959 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5961 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5962 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5963 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5964 (cmp_op, sve_imm_con): New code attributes.
5965 (SVE_COND_INT_CMP, imm_con): Delete.
5966 (cmp_op): Remove above unspecs from int attribute.
5967 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5969 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5970 comparison-specific unspecs.
5971 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5972 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5973 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5974 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5975 (*vec_fcm<cmp_op><mode>): Rename to...
5976 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5977 (*vec_fcmuo<mode>): Rename to...
5978 (*fcmuo<mode>): ...this and adjust likewise.
5979 (*pred_fcm<cmp_op><mode>): New pattern.
5980 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5981 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5983 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5985 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5986 (aarch64_emit_sve_predicated_cond): New function.
5987 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5988 (aarch64_emit_unspec_cond_or): Replace with...
5989 (aarch64_emit_sve_or_conds): ...this new function. Use
5990 aarch64_emit_sve_ptrue_op for the individual comparisons and
5991 aarch64_emit_binop to OR them together.
5992 (aarch64_emit_inverted_unspec_cond): Replace with...
5993 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5994 aarch64_emit_sve_ptrue_op for the comparison and
5995 aarch64_emit_unop to invert the result.
5996 (aarch64_expand_sve_vec_cmp_float): Update after the above
5997 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5999 2018-05-07 Nathan Sidwell <nathan@acm.org>
6001 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6002 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6003 (Backwards Compatibility): Likewise.
6005 2018-05-07 Luis Machado <luis.machado@linaro.org>
6009 2018-05-07 Luis Machado <luis.machado@linaro.org>
6011 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6012 <prefetch_dynamic_strides>: New const bool field.
6013 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6014 prefetch_dynamic_strides.
6015 (exynosm1_prefetch_tune): Likewise.
6016 (thunderxt88_prefetch_tune): Likewise.
6017 (thunderx_prefetch_tune): Likewise.
6018 (thunderx2t99_prefetch_tune): Likewise.
6019 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6021 (aarch64_override_options_internal): Update to set
6022 PARAM_PREFETCH_DYNAMIC_STRIDES.
6023 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6024 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6025 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6026 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6027 prefetch-dynamic-strides setting.
6029 2018-05-07 Luis Machado <luis.machado@linaro.org>
6031 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6032 <minimum_stride>: New const int field.
6033 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6034 minimum_stride field.
6035 (exynosm1_prefetch_tune): Likewise.
6036 (thunderxt88_prefetch_tune): Likewise.
6037 (thunderx_prefetch_tune): Likewise.
6038 (thunderx2t99_prefetch_tune): Likewise.
6039 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6040 (aarch64_override_options_internal): Update to set
6041 PARAM_PREFETCH_MINIMUM_STRIDE.
6042 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6043 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6044 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6045 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6046 stride is constant and is below the minimum stride threshold.
6048 2018-05-07 Luis Machado <luis.machado@linaro.org>
6050 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6053 2018-05-07 Luis Machado <luis.machado@linaro.org>
6055 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6056 <prefetch_dynamic_strides>: New const bool field.
6057 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6058 prefetch_dynamic_strides.
6059 (exynosm1_prefetch_tune): Likewise.
6060 (thunderxt88_prefetch_tune): Likewise.
6061 (thunderx_prefetch_tune): Likewise.
6062 (thunderx2t99_prefetch_tune): Likewise.
6063 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6065 (aarch64_override_options_internal): Update to set
6066 PARAM_PREFETCH_DYNAMIC_STRIDES.
6067 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6068 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6069 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6070 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6071 prefetch-dynamic-strides setting.
6073 2018-05-07 Luis Machado <luis.machado@linaro.org>
6075 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6076 <minimum_stride>: New const int field.
6077 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6078 minimum_stride field.
6079 (exynosm1_prefetch_tune): Likewise.
6080 (thunderxt88_prefetch_tune): Likewise.
6081 (thunderx_prefetch_tune): Likewise.
6082 (thunderx2t99_prefetch_tune): Likewise.
6083 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6084 (aarch64_override_options_internal): Update to set
6085 PARAM_PREFETCH_MINIMUM_STRIDE.
6086 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6087 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6088 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6090 stride is constant and is below the minimum stride threshold.
6092 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6095 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6096 the type is addressable. Don't force op into register if it has
6099 2018-05-05 Roland McGrath <mcgrathr@google.com>
6102 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6103 any section for which we don't know a specific type it should have,
6104 regardless of name. Previously this was done only for the exact
6105 names ".init_array", ".fini_array", and ".preinit_array".
6106 (default_elf_asm_named_section): Add comment about
6107 relationship with default_section_type_flags and SECTION_NOTYPE.
6108 (get_section): Don't consider it a type conflict if one side has
6109 SECTION_NOTYPE and the other doesn't, as long as neither has the
6110 SECTION_BSS et al used in the default_section_type_flags logic.
6112 2018-05-05 Tom de Vries <tom@codesourcery.com>
6115 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6116 (workaround_barsyncs): New function.
6117 (nvptx_reorg): Use workaround_barsyncs.
6118 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6119 (define_expand "nvptx_membar_cta"): New define_expand.
6120 (define_insn "*nvptx_membar_cta"): New insn.
6122 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6124 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6125 To improve optimization opportunities.
6126 * builtin-types.def: The new needed builtin types for the above.
6128 2018-05-04 Richard Biener <rguenther@suse.de>
6130 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6131 * gimple-ssa-store-merging.c
6132 (imm_store_chain_info::output_merged_store): Remove redundant create,
6133 release split_store vector contents on failure.
6134 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6135 scalar stmt vector on cache hit.
6137 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6139 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6141 * config.gcc (powerpc-xilinx-eabi*): Remove.
6142 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6144 (fusion_addis_mem_combo_load): Ditto.
6145 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6147 (rs6000_cpu_cpp_builtins): Ditto.
6148 * config/rs6000/rs6000-linux.c
6149 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6150 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6151 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6153 (rs6000_setup_reg_addr_masks): Ditto.
6154 (rs6000_init_hard_regno_mode_ok): Ditto.
6155 (rs6000_option_override_internal): Ditto.
6156 (legitimate_lo_sum_address_p): Ditto.
6157 (rs6000_legitimize_address): Ditto.
6158 (rs6000_legitimize_reload_address): Ditto.
6159 (rs6000_legitimate_address_p): Ditto.
6160 (abi_v4_pass_in_fpr): Ditto.
6161 (setup_incoming_varargs): Ditto.
6162 (rs6000_gimplify_va_arg): Ditto.
6163 (rs6000_split_multireg_move): Ditto.
6164 (rs6000_savres_strategy): Ditto.
6165 (rs6000_emit_prologue_components): Ditto.
6166 (rs6000_emit_epilogue_components): Ditto.
6167 (rs6000_emit_prologue): Ditto.
6168 (rs6000_emit_epilogue): Ditto.
6169 (rs6000_elf_file_end): Ditto.
6170 (rs6000_function_value): Ditto.
6171 (rs6000_libcall_value): Ditto.
6172 * config/rs6000/rs6000.h: Ditto.
6173 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6174 (TARGET_MINMAX): ... this. New.
6175 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6176 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6177 (*movsi_internal1_single): Delete.
6178 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6179 mfpu=, mxilinx-fpu): Delete.
6180 * config/rs6000/singlefp.h: Delete.
6181 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6182 * config/rs6000/t-rs6000: Ditto.
6183 * config/rs6000/t-xilinx: Delete.
6184 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6185 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6186 (VStype_simple): Delete.
6187 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6188 * config/rs6000/xfpu.h: Delete.
6189 * config/rs6000/xfpu.md: Delete.
6190 * config/rs6000/xilinx.h: Delete.
6191 * config/rs6000/xilinx.opt: Delete.
6192 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6193 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6195 2018-05-04 Tom de Vries <tom@codesourcery.com>
6198 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6201 2018-05-04 Richard Biener <rguenther@suse.de>
6204 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6205 (expand_complex_div_wide): Likewise.
6206 (expand_complex_operations_1): Likewise.
6207 (expand_complex_libcall): Preserve EH info of the original stmt.
6208 (tree_lower_complex): Handle removed blocks.
6209 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6210 on complex multiplication and division libcall builtins.
6212 2018-05-04 Richard Biener <rguenther@suse.de>
6215 * fold-const.c (negate_expr_p): Restrict negation of operand
6216 zero of a division to when we know that can happen without
6218 (fold_negate_expr_1): Likewise.
6220 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6223 * real.h (real_nextafter): Declare.
6224 * real.c (real_nextafter): New function.
6225 * fold-const-call.c (fold_const_nextafter): New function.
6226 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6227 CASE_CFN_NEXTTOWARD.
6228 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6229 even when arg1_mode is different from arg0_mode.
6231 2018-05-03 Nathan Sidwell <nathan@acm.org>
6233 * doc/extend.texi (Deprecated Features): Remove
6235 (Backwards Compatibility): Likewise.
6236 * doc/invoke.texi (C++ Language Options): Likewise.
6237 (C++ Dialect Options): Likewise.
6239 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6242 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6243 _mm512_mask_mullox_epi64): New intrinsics.
6245 2018-05-03 Tom de Vries <tom@codesourcery.com>
6248 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6249 dump files): Add offload-tree.
6251 2018-05-03 Richard Biener <rguenther@suse.de>
6253 PR tree-optimization/85615
6254 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6255 to loops not nested in BBs loop father to avoid creating multi-entry
6258 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6260 PR tree-optimization/70291
6261 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6262 arguments. Change return type to tree. Emit libcall as a new
6263 statement rather than replacing existing one when inplace_p is true.
6264 (expand_complex_multiplication_components): New function.
6265 (expand_complex_multiplication): Expand floating-point complex
6266 multiplication using the above.
6267 (expand_complex_division): Rename inner_type parameter to type.
6268 Update expand_complex_libcall call-site.
6269 (expand_complex_operations_1): Update expand_complex_multiplication
6270 and expand_complex_division call-sites.
6272 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6275 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6276 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6277 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6278 the highest significant bit of the shift count mask is clear. In
6279 check whether and[sq]i3 is needed verify that all significant bits
6280 of the shift count other than the highest are set.
6282 2018-05-02 Tom de Vries <tom@codesourcery.com>
6285 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6286 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6287 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6288 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6289 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6290 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6291 __builtin_goacc_parlevel_size.
6293 2018-05-02 Richard Biener <rguenther@suse.de>
6295 PR tree-optimization/85597
6296 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6297 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6300 2018-05-02 Tom de Vries <tom@codesourcery.com>
6303 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6304 dump files): Add ltrans-tree.
6306 2018-05-02 Tom de Vries <tom@codesourcery.com>
6309 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6310 dump files): Add wpa-ipa.
6312 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6314 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6315 powerpc*-*-linux*paired* target.
6316 * config/rs6000/750cl.h: Delete.
6317 * config/rs6000/paired.h: Delete.
6318 * config/rs6000/paired.md: Delete.
6319 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6321 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6322 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6323 comment. Remove paired float support.
6324 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6325 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6327 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6328 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6330 * config/rs6000/rs6000.c: Remove paired float support.
6331 (paired_expand_vector_init, paired_expand_vector_move,
6332 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6333 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6334 paired_expand_builtin, paired_expand_predicate_builtin,
6335 paired_init_builtins): Delete.
6336 * config/rs6000/rs6000.h: Remove paired float support.
6337 * config/rs6000/rs6000.md: Remove paired float support.
6338 (move_from_CR_ov_bit): Delete.
6339 * config/rs6000/rs6000.opt (mpaired): Delete.
6340 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6341 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6343 2018-05-02 Richard Biener <rguenther@suse.de>
6346 * gimplify.c (gimplify_save_expr): When in SSA form allow
6347 SAVE_EXPRs to compute to SSA vars.
6349 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6352 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6353 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6354 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6355 clobber operands[2], instead use a new pseudo. Formatting fixes.
6357 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6359 PR tree-optimization/85586
6360 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6361 exit early for statements in the same group if the accesses are
6364 2018-05-02 Tom de Vries <tom@codesourcery.com>
6367 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6370 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6372 PR tree-optimization/85143
6373 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6375 2018-05-01 Tom de Vries <tom@codesourcery.com>
6378 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6379 not found" error message.
6381 2018-05-01 Tom de Vries <tom@codesourcery.com>
6384 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6385 * vec.c (test_ordered_remove_if): New function.
6386 (vec_c_tests): Call test_ordered_remove_if.
6387 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6388 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6389 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6390 VEC_ORDERED_REMOVE_IF.
6392 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6394 PR tree-optimization/82665
6395 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6396 pointer subtraction where arguments come from a memchr call.
6398 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6400 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6401 --push-state --as-needed and --pop-state instead of --as-needed and
6402 --no-as-needed if ld supports it.
6403 * configure: Regenerated.
6406 * doc/install.texi2html: Replace _002d with - and _002a with * in
6407 generated html files using sed.
6409 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6412 * gcc-rich-location.c (blank_line_before_p): New function.
6413 (use_new_line): New function.
6414 (gcc_rich_location::add_fixit_insert_formatted): New function.
6415 * gcc-rich-location.h
6416 (gcc_rich_location::add_fixit_insert_formatted): New function.
6418 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6420 * selftest.c (assert_streq): Rename "expected" and "actual" to
6421 "val1" and "val2". Extend NULL-handling to cover both inputs
6422 symmetrically, while still requiring both to be non-NULL for a pass.
6423 * selftest.h (assert_streq): Rename "expected" and "actual" to
6425 (ASSERT_EQ): Likewise.
6426 (ASSERT_EQ_AT): Likewise.
6427 (ASSERT_KNOWN_EQ): Likewise.
6428 (ASSERT_KNOWN_EQ_AT): Likewise.
6429 (ASSERT_NE): Likewise.
6430 (ASSERT_MAYBE_NE): Likewise.
6431 (ASSERT_MAYBE_NE_AT): Likewise.
6432 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6433 the assertion to pass.
6434 (ASSERT_STREQ_AT): Likewise.
6436 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6438 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6439 interaction with -pie.
6441 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6443 * selftest.h: Fix alphabetization of per-source-file selftest
6446 2018-04-30 Jason Merrill <jason@redhat.com>
6448 PR c++/61982 - dead stores to destroyed objects.
6449 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6452 2018-04-30 Jason Merrill <jason@redhat.com>
6454 * tree.c (build_clobber): New.
6455 * tree.h: Declare it.
6456 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6458 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6460 * diagnostic-show-locus.c (layout::layout): Update for
6461 location_get_source_line returning a char_span.
6462 (struct char_span): Move to input.h.
6463 (struct correction): Update for fields in char_span becoming
6465 (struct source_line): Update for location_get_source_line
6466 returning a char_span.
6467 (layout::print_line): Likewise.
6468 * edit-context.c (edited_file::print_content): Likewise.
6469 (edited_file::print_diff_hunk): Likewise.
6470 (edited_file::print_run_of_changed_lines): Likewise.
6471 (edited_file::get_num_lines): Likewise.
6472 (edited_line::edited_line): Likewise.
6473 * final.c (asm_show_source): Likewise.
6474 * input.c (location_get_source_line): Convert return type
6475 from const char * to char_span, losing the final "line_len"
6477 (dump_location_info): Update for the above.
6478 (get_substring_ranges_for_loc): Likewise. Use a char_span
6479 when handling the literal within the line.
6480 (test_reading_source_line): Update for location_get_source_line
6481 returning a char_span.
6482 * input.h (class char_span): Move here from
6483 diagnostic-show-locus.c, converting from a struct to a class.
6484 Make data members private.
6485 (char_span::operator bool): New.
6486 (char_span::length): New.
6487 (char_span::get_buffer): New.
6488 (char_span::operator[]): New.
6489 (char_span::subspan): Make const.
6490 (char_span::xstrdup): New.
6491 (location_get_source_line): Convert return type from const char *
6492 to char_span, losing the final "line_size" param.
6494 2018-04-30 Jan Hubicka <jh@suse.cz>
6496 * lto-wrapper.c (ltrans_priorities): New static var.
6497 (cmp_priority): New.
6498 (run_gcc): Read priorities and if doing parallel build order
6499 the Makefile by them.
6501 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6503 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6505 2018-04-30 Richard Biener <rguenther@suse.de>
6507 * tree-cfg.c (verify_address): Remove base argument, add
6508 flag whether to check TREE_ADDRESSABLE and do that.
6509 (verify_expr): Remove.
6510 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6511 (verify_gimple_assign_single): Likewise.
6512 (verify_gimple_switch): Likewise.
6513 (verify_expr_location_1): Dereference tp once. Add (disabled)
6514 piece from verify_expr.
6515 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6517 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6519 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6521 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6523 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6524 (small_data_pattern): Likewise.
6525 (arc_rewrite_small_data): Likewise.
6526 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6527 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6528 (get_symbol_alignment): New function.
6529 (legitimate_small_data_address_p): Likewise.
6530 (legitimate_scaled_address): Update, call
6531 legitimate_small_data_address_p.
6532 (output_sdata): New static variable.
6533 (arc_print_operand): Update how we handle small data operands.
6534 (arc_print_operand_address): Likewise.
6535 (arc_legitimate_address_p): Update, use
6536 legitimate_small_data_address_p.
6537 (arc_rewrite_small_data_p): Remove.
6538 (arc_rewrite_small_data_1): Likewise.
6539 (arc_rewrite_small_data): Likewise.
6540 (small_data_pattern): Likewise.
6541 (compact_sda_memory_operand): Update to use
6542 legitimate_small_data_address_p and get_symbol_alignment.
6543 (prepare_move_operands): Don't rewite sdata pattern.
6544 (prepare_extend_operands): Remove.
6545 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6547 (zero_extendqisi2): Likewise.
6548 (zero_extendhisi2): Likewise.
6549 (extendqihi2): Likewise.
6550 (extendqisi2): Likewise.
6551 (extendhisi2): Likewise.
6555 * config/arc/constraints.md (Usd): Change it to memory constraint.
6557 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6559 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6560 as source of std instructions.
6561 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6562 allow 6-bit constants as source for store instructions.
6563 (movdi_insn): Update instruction pattern to allow 6-bit constants
6564 as source for store instructions.
6566 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6568 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6570 2018-04-30 Nathan Sidwell <nathan@acm.org>
6571 Sandra Loosemore <sandra@codesourcery.com>
6573 * dumpfile.c (dump_open): Allow '-' for stdout.
6574 * doc/invoke.texi (Developer Options): Document dump filename
6575 determination early. Document stdin/stdout selection.
6577 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6579 Microblaze Target: PIC data text relative
6581 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6582 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6584 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6585 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6586 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6587 New addressing mode for data-text relative position indepenedent code.
6588 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6589 'ADDRESS_SYMBOLIC_TXT_REL'.
6590 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6591 (microblaze_legitimate_pic_operand): Exclude function calls from
6592 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6593 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6595 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6596 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6597 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6598 for 'address + offset'.
6599 (microblaze_expand_prologue): Add new function prologue call for
6601 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6602 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6603 table in case of TARGET_PIC_DATA_TEXT_REL.
6604 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6605 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6606 Add new macros 'UNSPEC_TEXT',
6607 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6608 + exclude function calls from 'UNSPEC_PLT' in case of data text
6610 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6611 new target hook for generating address diff vector tables in case of
6613 * doc/tm.texi : Regenerate.
6614 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6615 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6616 of addr diff vector generation.
6617 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6618 target hook definition.
6619 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6620 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6621 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6622 option for data text relative.
6624 2018-04-30 Richard Biener <rguenther@suse.de>
6626 * tree-chrec.h (evolution_function_is_constant_p): Remove
6628 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6630 2018-04-30 Richard Biener <rguenther@suse.de>
6633 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6635 2018-04-30 Richard Biener <rguenther@suse.de>
6637 PR tree-optimization/28364
6638 PR tree-optimization/85275
6639 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6640 copying first exit test.
6642 2018-04-28 Mark Wielaard <mark@klomp.org>
6644 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6646 (dwarf_AT): Handle DW_AT_addr_base.
6647 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6649 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6652 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6653 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6654 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6655 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6657 2018-04-28 Richard Biener <rguenther@suse.de>
6659 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6660 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6661 to reflect use. Only add interesting stmts.
6663 2018-04-27 Martin Jambor <mjambor@suse.cz>
6666 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6667 the jump function allows for passing through aggregate values.
6669 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6671 * input.h (in_system_header_at): Convert from macro to inline
6673 (from_macro_expansion_at): Likewise.
6674 (from_macro_definition_at): Likewise.
6676 2018-04-27 Jeff Law <law@redhat.com>
6678 * config.gcc: Mark tile* targets as deprecated/obsolete.
6680 2018-04-27 Richard Biener <rguenther@suse.de>
6682 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6685 2018-04-27 Richard Biener <rguenther@suse.de>
6687 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6689 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6691 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6692 with Yd constraint. Set "preferred_for_speed" attribute from
6693 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6695 (*movdi_internal): Ditto.
6696 (movti_interunit splitters): Remove
6697 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6698 (movdi_interunit splitters): Ditto.
6699 * config/i386/constraints.md (Ye): Remove.
6700 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6702 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6705 * config/aarch64/constraints.md (Usg): Limit to 31.
6708 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6710 PR tree-optimization/85529
6711 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6712 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6713 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6714 zero extension or masking of the MSB bit.
6715 (optimize_range_tests): Add FIRST_BB argument, pass it through
6716 to optimize_range_tests_var_bound.
6717 (maybe_optimize_range_tests, reassociate_bb): Adjust
6718 optimize_range_tests callers.
6720 2018-04-26 Richard Biener <rguenther@suse.de>
6721 Jakub Jelinek <jakub@redhat.com>
6723 * cgraph.h (symbol_table): Just declare debug method here.
6724 * symtab.c (symbol_table::debug): Define.
6726 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6728 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6730 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6732 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6733 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6734 (*movdi_internal): Substitute Yi and Yj constraint with x
6735 and Ym and Yn constraint with y constraint. Update "isa"
6736 attribute and set "preferred_for_speed" attribute from
6737 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6738 (*movsi_internal): Ditto.
6739 (*movdf_internal): Ditto.
6740 (*movsf_internal): Ditto.
6741 (*zero_extendsidi2): Ditto.
6742 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6743 (sse2_loadld): Ditto.
6744 (*vec_extract<ssevecmodelower>_0): Ditto.
6745 (*vec_extractv4si_0_zext_sse4): Ditto.
6746 (vec_concatv2di): Ditto.
6747 (*vec_dup<mode>): Ditto.
6748 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6749 * config/i386/constraints.md (Yi): Remove.
6754 2018-04-26 Nathan Sidwell <nathan@acm.org>
6756 * dumpfile.c (dump_open): New.
6757 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6758 (dump_finish): Detect stdio/stderr by value not name.
6760 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6762 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6764 2018-04-26 Tom de Vries <tom@codesourcery.com>
6767 * config/nvptx/nvptx.c (verify_neutering_jumps)
6768 (verify_neutering_labels): New function
6769 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6771 2018-04-26 Tom de Vries <tom@codesourcery.com>
6774 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6775 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6778 2018-04-26 Richard Biener <rguenther@suse.de>
6779 Tom de Vries <tom@codesourcery.com>
6782 * lto-streamer-out.c (output_function): Fixup loops if required to match
6783 discovery done in the reader.
6785 2018-04-26 Richard Biener <rguenther@suse.de>
6787 PR tree-optimization/85116
6788 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6789 have a loop exit from the single latch predecessor. Remove
6790 case of header with just condition.
6791 (ch_base::copy_headers): Exclude infinite loops from any
6793 (pass_ch::execute): Record exits.
6795 2018-04-26 Richard Biener <rguenther@suse.de>
6797 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6798 prologue cost vector and pass it to vect_get_load_cost.
6799 (vect_get_peeling_costs_all_drs): Likewise.
6800 (vect_peeling_hash_get_lowest_cost): Likewise.
6801 (vect_enhance_data_refs_alignment): Likewise.
6803 2018-04-26 Richard Biener <rguenther@suse.de>
6806 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6807 checking of integer<->pointer conversions.
6808 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6809 sign-/zero-extending pointer types.
6810 (expand_omp_for_static_chunk): Likewise.
6812 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6813 Jean Lee <xiaoyur347@gmail.com>
6815 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6816 (TARGET_ASAN_SHADOW_OFFSET): Define.
6817 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6818 true for -fsanitize=address.
6820 2018-04-25 Mark Wielaard <mark@klomp.org>
6822 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6825 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6827 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6828 than "alu", remove explicit "memory" and "imm_disp" attributes.
6829 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6832 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6833 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6834 gen_lowpart_no_emit.
6836 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6839 * config/i386/i386.c (ix86_expand_builtin): Change memory
6840 operand to XI, extend p0 to Pmode.
6841 * config/i386/i386.md: Change unspec volatile and operand
6842 1 mode to XI, change operand 0 mode to P.
6844 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6846 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6847 GET_MODE_MASK before any checking.
6848 (nds32_can_use_bset_p): Likewise.
6849 (nds32_can_use_btgl_p): Likewise.
6851 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6853 * config/nds32/nds32-doubleword.md: New define_split pattern for
6854 illegal register number.
6856 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6858 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6860 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6862 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6864 2018-04-25 Richard Biener <rguenther@suse.de>
6866 * lto-streamer.h (LTO_major_version): Bump to 8.
6868 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6870 * BASE-VER: Set to 9.0.0.
6872 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6874 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6875 in __abskf2 and __powikf2.
6877 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6880 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6881 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6882 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6883 Use the above on operand 2. Reindent.
6884 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6886 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6889 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6891 * config/i386/i386.opt (mcet): Removed.
6892 * doc/install.texi: Remove -mcet documentation.
6893 * doc/invoke.texi: Likewise.
6895 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6898 * doc/install.texi: Remove -mcet from bootstrap-cet.
6900 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6903 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6904 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6908 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6909 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6910 containing a CONST_VECTOR.
6912 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6914 * doc/install.texi: Update newlib dependency for nvptx.
6916 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6919 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6920 instead of INTVAL when shifting x left.
6922 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6924 PR tree-optimization/85478
6925 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6926 vect_grouped_store_supported for single element vectors.
6928 2018-04-24 Richard Biener <rguenther@suse.de>
6931 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6932 load cost increase to the case of non-constant step.
6934 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6937 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6938 destination if any_malformed_asm.
6940 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6943 * expr.c (store_field): In the bitfield case, if the value comes from
6944 a function call and is returned in registers by means of a PARALLEL,
6945 do not change the mode of the temporary unless BLKmode and VOIDmode.
6947 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6949 PR rtl-optimization/85423
6950 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6951 dependencies to debug insns when the previous insn is non-debug.
6953 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6955 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6956 enums into a single definition.
6957 (fls): Fix predicates and printing.
6960 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6962 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6963 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6964 and short u6 immediate.
6965 (check_if_valid_sleep_operand): Remove.
6966 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6968 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6970 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6971 flag_always_save_lp condition.
6972 * config/nds32/nds32.opt (malways-save-lp): New option.
6974 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6976 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6977 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6978 * config/nds32/nds32.h
6979 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6980 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6982 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6984 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6985 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6987 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6988 Chung-Ju Wu <jasonwucj@gmail.com>
6990 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6992 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6993 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6995 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6997 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6999 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7001 * config/nds32/nds32-protos.h (nds32_data_alignment,
7002 nds32_local_alignment): Declare.
7003 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7004 nds32_local_alignment): New functions.
7005 (TARGET_CONSTANT_ALIGNMENT): Define.
7006 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7008 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7010 * config/nds32/nds32.c
7011 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7012 (TARGET_MODES_TIEABLE_P): Likewise.
7014 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7016 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7019 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7020 Chung-Ju Wu <jasonwucj@gmail.com>
7022 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7023 for unaligned access.
7024 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7026 * config/nds32/nds32-intrinsic.md: Likewise.
7027 * config/nds32/nds32_intrinsic.h: Likewise.
7028 * config/nds32/nds32.h (nds32_builtins): Likewise.
7029 * config/nds32/nds32.opt (munaligned-access): New option.
7030 * config/nds32/nds32.c (nds32_asm_file_start): Display
7031 flag_unaligned_access status.
7033 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7035 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7036 -mno-relax is present.
7037 * config/riscv/linux.h (LINK_SPEC): Ditto.
7039 2018-04-20 Martin Sebor <msebor@redhat.com>
7042 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7044 (gimple_fold_builtin_stxcpy_chk): Same.
7045 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7047 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7050 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7051 __powikf2 when long double is IEEE 128-bit.
7053 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7055 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7056 step to make sure stack always aligned.
7058 2018-04-20 Carl Love <cel@us.ibm.com>
7061 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7062 size check for arg0.
7064 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7065 Tom de Vries <tom@codesourcery.com>
7068 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7069 Emit insns for calls too.
7070 (nvptx_find_par): Always look for worker-level predecessor insn.
7071 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7073 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7074 (nvptx_process_pars): Propagate frames for calls.
7076 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7079 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7081 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7082 (ix86_handle_option): Don't handle OPT_mibt.
7083 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7085 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7087 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7088 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7089 (ix86_target_macros): Define __CET__ with flag_cf_protection
7090 for -fcf-protection.
7091 * config/i386/i386.c (isa2_opts): Remove -mibt.
7092 * config/i386/i386.h (TARGET_IBT): Removed.
7093 (TARGET_IBT_P): Likewise.
7094 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7095 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7096 * config/i386/i386.opt (mcet): Update help message.
7099 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7100 -mcet as an alias for -mshstk.
7102 2018-04-20 Richard Biener <rguenther@suse.de>
7105 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7106 complexity by forcing a single use of the multiply operand.
7108 2018-04-20 Martin Jambor <mjambor@suse.cz>
7111 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7112 recursion dependency to only apply to non-clones.
7114 2018-04-20 Martin Jambor <mjambor@suse.cz>
7117 * ipa-cp.c (create_specialized_node): Check that clones of
7118 self-recursive edges exist during IPA-CP.
7120 2018-04-19 Toon Moene <toon@moene.org>
7122 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7125 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7127 PR tree-optimization/85467
7128 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7129 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7130 VECTOR_CST element to type.
7132 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7135 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7136 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7137 (builtin_longjmp): Likewise.
7138 (save_stack_nonlocal): New pattern.
7139 (restore_stack_nonlocal): Likewise.
7141 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7144 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7145 Replace ASM_OUTPUT_LABEL with fprintf.
7147 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7150 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7151 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7152 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7153 define __IBT__ and __SHSTK__ for -fcf-protection.
7154 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7156 (ix86_trampoline_init): Likewise.
7157 (x86_output_mi_thunk): Likewise.
7158 (ix86_notrack_prefixed_insn_p): Likewise.
7159 (ix86_option_override_internal): Don't disallow -fcf-protection.
7160 * config/i386/i386.md (rdssp<mode>): Also enable for
7162 (incssp<mode>): Likewise.
7163 (nop_endbr): Likewise.
7164 * config/i386/i386.opt (mcet): Change help message to built-in
7168 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7169 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7170 enable CET built-in functions.
7172 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7174 * common/config/i386/i386-common.c
7175 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7176 OPTION_MASK_ISA_MOVDIRI_UNSET,
7177 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7178 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7179 * config.gcc (movdirintrin.h): New header.
7180 * config/i386/cpuid.h (bit_MOVDIRI,
7181 bit_MOVDIR64B): New bits.
7182 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7184 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7185 (VOID, PVOID, PCVOID)): New function types.
7186 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7187 __builtin_ia32_directstoreu_u64,
7188 __builtin_ia32_movdir64b): New builtins.
7189 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7190 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7192 (ix86_valid_target_attribute_inner_p): Ditto.
7193 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7194 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7195 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7196 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7197 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7198 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7199 (movdiri<mode>, movdir64b_<mode>): New.
7200 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7201 * config/i386/immintrin.h: Include movdirintrin.h.
7202 * config/i386/movdirintrin.h: New file.
7203 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7205 2018-04-19 Richard Biener <rguenther@suse.de>
7208 * cfg.c (clear_bb_flags): When loop state says we have
7209 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7211 2018-04-19 Richard Biener <rguenther@suse.de>
7213 PR tree-optimization/84737
7214 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7215 copying restrict info.
7216 (vect_setup_realignment): Use it.
7217 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7218 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7219 the first DR to all generated stores.
7220 (vectorizable_load): Likewise for loads.
7222 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7224 PR tree-optimization/85446
7225 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7226 the integral and pointer types to have the same precision.
7228 * doc/install.texi: Document --disable-cet being the default and
7231 2018-04-18 Martin Liska <mliska@suse.cz>
7233 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7236 2018-04-18 Martin Liska <mliska@suse.cz>
7239 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7242 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7243 arguments if they are comparable.
7245 2018-04-18 Martin Liska <mliska@suse.cz>
7248 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7251 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7254 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7257 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7258 ENDBR after calling __morestack.
7260 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7263 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7264 by using gcc_base_ver to generate a gcc_driver_version, and use
7265 it when generating GCC_DRIVER_NAME.
7266 * configure: Regenerate.
7268 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7271 * config.gcc: Obsolete powerpc*-*-*spe*.
7273 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7276 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7277 (stabstr_D): Change type of unum from unsigned int to
7278 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7281 2018-04-17 Jim Wilson <jimw@sifive.com>
7284 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7285 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7286 Set arg_pointer_offset after using pretend_args_size.
7288 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7290 PR rtl-optimization/85431
7291 * dse.c (record_store): Ignore zero width stores.
7294 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7295 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7296 __builtin_stack_restore rather than after it.
7297 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7298 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7299 argument instead of virtual_dynamic_stack_rtx.
7301 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7303 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7305 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7306 Add note to error message to explain internal mapping of overloaded
7307 built-in function name to non-overloaded built-in function name.
7308 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7311 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7314 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7315 where the inputs overlap with the output.
7317 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7319 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7320 (=v, v) alternative and explicit "memory" attribute.
7321 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7322 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7324 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7325 "sselog1" type instead of "sselog".
7326 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7327 "sselog". Remove explicit "memory" attribute.
7328 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7329 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7331 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7332 "sselog1" type instead of "sselog". Remove explicit "memory"
7334 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7335 use "sselog1" type instead of "sselog". Remove explicit "memory"
7337 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7338 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7340 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7341 "sselog1" type instead of "sselog". Remove explicit "memory"
7343 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7344 use "sselog1" type instead of "sselog". Remove explicit "memory"
7348 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7351 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7354 2018-04-17 Martin Jambor <mjambor@suse.cz>
7357 * ipa-cp.c (create_specialized_node): Call
7358 expand_all_artificial_thunks if necessary.
7360 2018-04-17 Martin Liska <mliska@suse.cz>
7363 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7364 in message, remote space in between '_G' and '('.
7366 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7369 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7370 avx512f_vmcmp<mode>3<round_saeonly_name>,
7371 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7372 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7373 avx512f_rndscale<mode><round_saeonly_name>,
7374 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7375 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7376 Use %<iptr>2 instead of %2 for -masm=intel.
7377 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7378 avx512f_vcvttss2usi<round_saeonly_name>,
7379 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7381 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7382 avx512f_vcvttsd2usi<round_saeonly_name>,
7383 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7384 Use %q1 instead of %1 for -masm=intel.
7385 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7386 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7387 of %3 for -masm=intel.
7388 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7390 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7392 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7394 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7395 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7397 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7399 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7400 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7401 %g1 and one with %0 and %1.
7402 (avx512er_vmrcp28<mode><round_saeonly_name>,
7403 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7405 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7406 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7407 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7408 of %0 and %{%4%} for -masm=intel.
7409 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7410 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7411 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7412 order of %0 and %{%5%}%{z%} for -masm=intel.
7414 2018-04-17 Jan Hubicka <jh@suse.cz>
7417 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7419 2018-04-17 Martin Liska <mliska@suse.cz>
7422 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7423 for target_clone error message. Make default implementation
7424 clone to be a local declaration.
7425 (separate_attrs): Add new argument and check for an empty
7427 (expand_target_clones): Handle it.
7428 (ipa_target_clone): Make redirection just for target_clones
7431 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7432 Tom de Vries <tom@codesourcery.com>
7435 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7436 tiled basic blocks without omp continue statements.
7438 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7441 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7442 vec_extract expression as having side effects to make sure it gets
7445 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7448 * config/i386/i386.c (get_builtin_code_for_version): Check
7451 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7454 * gcc/config.gcc: Support "skylake".
7455 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7457 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7458 (processor_target_table): Add "skylake".
7459 (ix86_option_override_internal): Add "skylake".
7460 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7461 PROCESSOR_CANNONLAKE.
7462 (get_builtin_code_for_version): Fix priority for
7463 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7464 PROCESSOR_SKYLAKE-AVX512.
7465 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7466 (processor_type): Add PROCESSOR_SKYLAKE.
7468 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7469 Jason Merrill <jason@redhat.com>
7472 * convert.c (convert_to_integer_1): Use direct recursion for
7473 enumeral types and types with a precision less than the number
7474 of bits in their mode.
7476 2018-04-16 Julia Koval <julia.koval@intel.com>
7479 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7480 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7482 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7485 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7486 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7487 and -mno-direct-move.
7489 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7492 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7493 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7494 Ensure negative shifts result in {0}.
7496 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7498 PR rtl-optimization/79916
7499 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7500 regs (if any) to define how to gnerate SD moves when LRA is in
7503 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7505 PR rtl-optimization/85393
7506 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7507 * except.c (expand_dw2_landing_pad_for_region): Make static.
7508 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7509 a label and unconditional jump to old_bb, rather than
7510 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7513 PR rtl-optimization/85376
7514 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7515 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7516 instead of a specific value.
7518 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7519 Bin Cheng <bin.cheng@arm.com>
7521 PR tree-optimization/82965
7522 PR tree-optimization/83991
7523 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7524 by_profile_only parameter.
7525 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7526 information if the loop was predicted to iterate too many times.
7527 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7529 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7532 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7535 2018-04-13 Martin Liska <mliska@suse.cz>
7536 Jakub Jelinek <jakub@redhat.com>
7539 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7540 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7541 * builtins.c (expand_builtin_memory_copy_args): Use
7542 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7543 handle dest_addr == pc_rtx.
7545 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7548 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7549 asked to not generate direct moves.
7550 (fix_trunc<mode>si2_stfiwx): Similar.
7551 (fix_trunc<mode>si2_internal): Similar.
7553 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7556 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7557 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7558 lookup if dest in some wider mode is known to be const0_rtx and
7559 if so, record permanent equivalence for it to be ZERO_EXTEND of
7560 the narrower mode destination.
7562 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7564 * lto-streamer-out.c (output_function): Revert 259346.
7565 * omp-expand.c (expand_oacc_for): Likewise.
7567 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7569 PR rtl-optimization/85354
7570 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7571 * sel-sched.c (sel_global_init): ... here.
7573 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7576 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7577 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7578 mode for PE-COFF targets.
7579 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7580 (i386_pe_asm_lto_end): Likewise.
7581 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7582 (TARGET_ASM_LTO_END): Likewise.
7583 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7584 (i386_pe_asm_lto_start): New function.
7585 (i386_pe_asm_lto_end): Likewise.
7587 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7588 Richard Biener <rguenther@suse.de>
7591 * lto-streamer-out.c (output_function): Fix CFG loop state before
7593 * omp-expand.c (expand_oacc_for): Handle calls to internal
7594 functions like regular functions.
7596 2018-04-12 Richard Biener <rguenther@suse.de>
7599 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7600 for the early LTO debug to properly generate references to it
7601 during DIE emission. Do not re-use that for the skeleton for
7603 (dwarf2out_early_finish): Likewise.
7605 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7608 * config/i386/sse.md
7609 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7610 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7611 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7612 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7613 and output is a reg, avoid creating invalid lowpart subreg, but
7614 instead split into a 512-bit move. Don't split if not AVX512VL,
7615 input is xmm16+ reg and output is a mem.
7616 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7617 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7618 xmm16+ reg and output is a mem.
7620 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7622 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7623 also for flag_dwarf2_cfi_asm.
7625 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7627 PR rtl-optimization/85342
7628 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7629 a bool scalar var inside of the loop instead. Don't try to update
7630 recog_data.operand after failed apply_change_group.
7632 2018-04-12 Tom de Vries <tom@codesourcery.com>
7635 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7636 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7637 array with flexible array member as array without given dimension.
7638 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7639 argument for undefined param to true.
7641 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7644 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7645 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7646 from PowerPC section.
7647 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7648 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7649 help text that is too long.
7650 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7651 help text that is too long.
7652 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7653 help text that is too long.
7655 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7657 * config/alpha/alpha.md (stack_probe_internal): Rename
7658 from "probe_stack". Update all callers.
7660 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7662 PR rtl-optimization/84566
7663 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7664 sched_macro_fuse_insns.
7666 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7669 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7670 (compute_block_dependences): ... from here.
7672 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7674 PR tree-optimization/85331
7675 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7676 from int to HOST_WIDE_INT.
7678 2018-04-11 Martin Jambor <mjambor@suse.cz>
7681 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7682 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7683 not the same as the source val.
7684 (cgraph_edge_brings_value_p): New parameter.
7685 (gather_edges_for_value): Pass destination value to
7686 cgraph_edge_brings_value_p.
7687 (perhaps_add_new_callers): Likewise.
7688 (get_info_about_necessary_edges): Likewise and exclude values brought
7689 only by self-recursive edges.
7690 (create_specialized_node): Redirect only clones of self-calling edges.
7691 (+self_recursive_pass_through_p): New function.
7692 (find_more_scalar_values_for_callers_subset): Use it.
7693 (find_aggregate_values_for_callers_subset): Likewise.
7694 (known_aggs_to_agg_replacement_list): Removed.
7695 (decide_whether_version_node): Re-calculate known constants for all
7696 remaining context clones.
7698 2018-04-11 Richard Biener <rguenther@suse.de>
7701 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7702 from early DWARF output.
7703 (dwarf2out_early_finish): Output line info unconditionally into
7704 early DWARF and add reference to it.
7706 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7709 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7710 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7711 other than V2DFmode using iptr mode attribute.
7712 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7714 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7716 PR rtl-optimization/84659
7717 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7719 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7722 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7724 (output_loc_list): Pass address of a dummy size variable even in the
7725 locview handling loop.
7726 (index_location_lists): Add comment on why skip_loc_list_entry can't
7729 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7732 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7735 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7738 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7739 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7740 and -mstring-compare-inline-limit.
7742 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7745 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7746 for stack clash protection in a register whenever we need it to be in
7749 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7751 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7752 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7754 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7757 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7759 (mlong-double-): Ditto.
7760 * config/rs6000/sysv4.opt (msdata=): Ditto.
7761 (mtls-size=): Ditto.
7763 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7765 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7766 erroneous entries for
7767 "vector int vec_ldl (int, long int *)", and
7768 "vector unsigned int vec_ldl (int, unsigned long int *)".
7769 Add comments and entries for
7770 "vector bool char vec_ldl (int, bool char *)",
7771 "vector bool short vec_ldl (int, bool short *)",
7772 "vector bool int vec_ldl (int, bool int *)",
7773 "vector bool long long vec_ldl (int, bool long long *)",
7774 "vector pixel vec_ldl (int, pixel *)",
7775 "vector long long vec_ldl (int, long long *)",
7776 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7777 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7778 type tree bool_long_long_type_node and correct definition of
7779 bool_V2DI_type_node to make reference to this new type tree.
7780 (rs6000_mangle_type): Replace erroneous reference to
7781 bool_long_type_node with bool_long_long_type_node.
7782 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7783 comments to emphasize sign distinctions for char and int types and
7784 replace RS6000_BTI_bool_long constant with
7785 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7786 use of RS6000_BTI_pixel.
7787 (bool_long_type_node): Remove this macro definition.
7788 (bool_long_long_type_node): New macro definition
7790 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7792 PR rtl-optimization/85300
7793 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7794 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7795 simplify_unary_operation fails.
7797 2018-04-10 Martin Liska <mliska@suse.cz>
7799 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7800 cgraph_edge and ipa_ref.
7802 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7806 * config/i386/sse.md
7807 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7808 computation of the VEC_MERGE selector from mask.
7809 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7810 Fix decoding of the VEC_MERGE selector into mask.
7812 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7814 PR tree-optimization/85286
7815 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7817 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7819 * final.c (final_1): Set insn_last_address as well as
7820 insn_current_address.
7822 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7825 * explow.c (emit_stack_probe): Call validize_mem on memory location
7826 before passing it to gen_probe_stack. Create address operand and
7827 legitimize it for the probe_stack_address case.
7829 2018-04-09 Jan Hubicka <jh@suse.cz>
7832 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7833 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7834 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7836 (free_lang_data): Rebuild type inheritance graph.
7838 2018-04-09 Martin Sebor <msebor@redhat.com>
7840 * invoke.texi (-finline-small-functions): Mention other optimization
7842 (-findirect-inlining, -fpartial-inlining): Same.
7843 (-finline-functions-called-once): Same.
7844 (-freorder-blocks-and-partition): Same.
7846 2018-04-09 Jan Hubicka <jh@suse.cz>
7849 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7850 jumps; choose last target that matches the criteria (i.e.
7851 no partition changes for non-crossing jumps).
7852 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7853 support for redirecting crossing jumps to non-crossing.
7855 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7857 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7858 also for naked functions.
7860 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7862 * config/arc/arc.md (add_shift): New pattern.
7863 (add_shift2): Likewise.
7864 (sub_shift): Likewise.
7865 (sub_shift_cmp0_noout): Likewise.
7866 (compare_si_ashiftsi): Likewise.
7867 (xbfu_cmp0_noout): New combine pattern.
7868 (xbfu_cmp0"): Likewise.
7869 (movsi_set_cc_insn): Place the predicable variant first.
7870 (commutative_binary_cmp0_noout): Remove clobber.
7871 (commutative_binary_cmp0): New pattern.
7872 (noncommutative_binary_cmp0): Likewise.
7873 (noncommutative_binary_cmp0_noout): Likewise.
7874 (noncommutative_binary_comparison_result_used): Removed.
7875 (rsub_cmp0): New pattern.
7876 (rsub_cmp0_noout): Likewise.
7877 (extzvsi): Changed, keep only meaningful variants.
7878 (SQH, SEZ): New iterators.
7879 (SQH_postfix): New mode attribute.
7880 (SEZ_prefix): New code attribute.
7881 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7882 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7883 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7885 (noncommutative_operator): Check the availability of barrel
7888 2018-04-09 Richard Biener <rguenther@suse.de>
7890 PR tree-optimization/85284
7891 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7892 Only use the niter constraining form of simple_iv when the exit
7895 2018-04-09 Tom de Vries <tom@codesourcery.com>
7898 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7899 (define_expand "*memory_barrier"): New define_expand.
7900 (define_insn "memory_barrier"): New insn.
7902 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7904 PR rtl-optimization/80463
7905 PR rtl-optimization/83972
7906 PR rtl-optimization/83480
7908 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7909 correct producer for the insn.
7910 (tidy_control_flow): Fixup seqnos in case of debug insns.
7912 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7914 PR rtl-optimization/83913
7916 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7917 different sched-times when merging exprs.
7919 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7921 PR rtl-optimization/83962
7923 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7924 tidy_fallthru_edge and tidy_control_flow.
7926 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7928 PR rtl-optimization/83530
7930 * sel-sched.c (force_next_insn): New global variable.
7931 (remove_insn_for_debug): When force_next_insn is true, also leave only
7932 next insn in the ready list.
7933 (sel_sched_region): When the region wasn't scheduled, make another pass
7934 over it with force_next_insn set to 1.
7936 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7938 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7940 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7941 for interrupt control.
7942 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7943 functions for interrupt control.
7944 * config/nds32/nds32-intrinsic.md: Likewise.
7945 * config/nds32/nds32_intrinsic.h: Likewise.
7946 * config/nds32/nds32.h (nds32_builtins): Likewise.
7948 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7950 * config/nds32/nds32.c (nds32_init_machine_status,
7951 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7952 strict_aligned_p field.
7953 (nds32_expand_to_rtl_hook): New function.
7954 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7955 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7957 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7958 Chung-Ju Wu <jasonwucj@gmail.com>
7960 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7961 * config/nds32/nds32-n7.md: New file.
7962 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7963 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7965 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7966 * config/nds32/nds32.md (pipeline_model): Add n7.
7967 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7968 * config/nds32/pipelines.md: Include n7 settings.
7970 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7971 Chung-Ju Wu <jasonwucj@gmail.com>
7973 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7974 * config/nds32/nds32-e8.md: New file.
7975 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7976 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7978 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7979 * config/nds32/nds32.md (pipeline_model): Add e8.
7980 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7981 * config/nds32/pipelines.md: Include e8 settings.
7983 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7984 Chung-Ju Wu <jasonwucj@gmail.com>
7986 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7987 * config/nds32/nds32-n8.md: New file.
7988 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7989 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7991 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7992 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7993 * config/nds32/nds32.md (pipeline_model): Add n8.
7994 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7995 * config/nds32/pipelines.md: Include n8 settings.
7997 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7998 Chung-Ju Wu <jasonwucj@gmail.com>
8000 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8001 * config/nds32/nds32-n9-2r1w.md: New file.
8002 * config/nds32/nds32-n9-3r2w.md: New file.
8003 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8004 nds32_register_ports): New or modify for cpu n9.
8005 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8007 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8008 * config/nds32/nds32-utils.c: New file.
8009 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8010 TARGET_MUL_SLOW): Define.
8011 * config/nds32/nds32.md (pipeline_model): New attribute.
8012 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8013 New options that support cpu n9.
8014 * config/nds32/pipelines.md: Include n9 settings.
8015 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8017 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8019 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8020 information if necessary.
8021 (output_cond_branch_compare_zero): Likewise.
8022 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8023 (nds32_target_alignment): Refine for alignment.
8024 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8025 (FUNCTION_BOUNDARY): Modify.
8026 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8028 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8030 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8032 * config/nds32/constants.md (unspec_volatile_element): Add values for
8033 TLB operation and data prefetch.
8034 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8035 functions for TLB operation and data prefetch.
8036 * config/nds32/nds32-intrinsic.md: Likewise.
8037 * config/nds32/nds32_intrinsic.h: Likewise.
8038 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8039 (nds32_print_operand): Likewise.
8040 * config/nds32/nds32.h (nds32_builtins): Likewise.
8042 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8043 Andrew Pinski <pinsika@gcc.gnu.org>
8046 * match.pd: Use constant_boolean_node of correct type instead of
8047 boolean_true_node or boolean_false_node for simplifying
8048 pointer comparisons to zero.
8050 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8052 PR tree-optimization/80021
8053 * tree.c (verify_type_variant): Make error call in verify_variant_match
8054 translatable and remove final full stop.
8056 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8058 * config/nds32/constants.md (unspec_volatile_element): Add
8059 UNSPEC_VOLATILE_EH_RETURN.
8060 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8061 nds32_output_stack_pop): Support dwarf exception handling process.
8062 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8063 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8064 exception handling process.
8065 (nds32_compute_stack_frame): Likewise.
8066 (nds32_return_addr_rtx): Likewise.
8067 (nds32_initial_elimination_offset): Likewise.
8068 (nds32_expand_prologue): Likewise.
8069 (nds32_expand_epilogue): Likewise.
8070 (nds32_dynamic_chain_address): New function.
8071 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8073 (DYNAMIC_CHAIN_ADDRESS): Define.
8074 (EH_RETURN_DATA_REGNO): Define.
8075 (EH_RETURN_STACKADJ_RTX): Define.
8076 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8077 patterns for dwarf exception handling.
8079 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8081 * config/nds32/nds32.h: Clean up obsolete macros.
8083 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8085 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8086 Add enum values for particular instructions.
8087 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8088 particular intrinsic functions.
8089 * config/nds32/nds32-intrinsic.md: Likewise.
8090 * config/nds32/nds32_intrinsic.h: Likewise.
8091 * config/nds32/nds32.h (nds32_builtins): Likewise.
8092 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8093 (btst, ave): New patterns for particular instructions.
8095 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8097 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8098 Add enum values for atomic load/store and memory sync.
8099 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8101 * config/nds32/nds32-intrinsic.md: Likewise.
8102 * config/nds32/nds32_intrinsic.h: Likewise.
8103 * config/nds32/nds32.h (nds32_builtins): Likewise.
8105 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8107 PR tree-optimization/85257
8108 * fold-const.c (native_encode_vector): If not all elts could fit
8109 and off is -1, return 0 rather than offset.
8110 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8111 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8112 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8113 adjust buffer in native_interpret_expr call.
8115 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8117 * config/nds32/constants.md (unspec_volatile_element): Add cache
8118 control enum values.
8119 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8120 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8121 * config/nds32/nds32.c (nds32_cctl_names): New.
8122 (nds32_print_operand): Handle cache control register names.
8123 * config/nds32/nds32.h (nds32_builtins): New enum values.
8124 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8126 * config/nds32/nds32.md (type): Add mmu.
8127 * config/nds32/pipelines.md (simple_insn): Add mmu.
8129 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8131 * config/nds32/nds32.md (type): Remove call.
8132 * config/nds32/pipelines.md (simple_insn): Likewise.
8134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8136 * config/nds32/constants.md (unspec_volatile_element): Add
8137 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8138 UNSPEC_VOLATILE_FMFCFG.
8139 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8140 description for fmfcfg and fmfcsr.
8141 (bdesc_1arg): Add fmtcsr.
8142 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8143 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8144 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8145 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8146 unspec_fmfcfg): New patterns.
8147 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8148 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8149 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8150 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8151 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8152 __nds32__fmfcfg): Define.
8154 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8156 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8157 intrinsic register names.
8158 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8159 intrinsic register enum values and macros.
8161 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8163 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8164 for load/store addressing form.
8165 (nds32_print_operand_address): Likewise.
8167 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8170 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8171 based on LABEL_REF. Remove useless assertion.
8172 (pic_address_needs_scratch): Fix formatting.
8173 (sparc_legitimize_pic_address): Minor tweaks.
8174 (sparc_delegitimize_address): Adjust assertion accordingly.
8175 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8176 into symbolic_operand.
8177 (movsi_high_pic_label_ref): Likewise.
8178 (movsi_lo_sum_pic_label_ref): Likewise.
8179 (movdi_pic_label_ref): Likewise.
8180 (movdi_high_pic_label_ref): Likewise.
8181 (movdi_lo_sum_pic_label_ref): Likewise.
8183 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8185 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8186 custom LIB_SPEC setup.
8188 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8189 Kito Cheng <kito.cheng@gmail.com>
8191 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8192 * config/riscv/freebsd.h: New.
8194 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8196 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8197 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8200 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8201 Kito Cheng <kito.cheng@gmail.com>
8203 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8204 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8205 * config/nds32/nds32-protos.h (nds32_output_call,
8206 nds32_output_return): Declare.
8207 * config/nds32/nds32.md: Refine all the call and return patterns.
8209 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8212 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8213 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8215 PR rtl-optimization/84872
8216 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8217 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8220 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8222 * expr.c (copy_blkmode_to_reg): Revert 254862.
8223 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8225 2018-04-06 Richard Biener <rguenther@suse.de>
8228 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8229 after seeing a component reference with an adjacent field. Treat
8230 refs to arrays at struct end of external decls similar to
8231 refs to unconstrained commons.
8233 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8236 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8237 look through SAVE_EXPRs with non-side-effects argument. Adjust
8239 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8242 2018-04-06 Richard Biener <rguenther@suse.de>
8245 * alias.c (find_base_term): New wrapper around find_base_term
8246 unwinding CSELIB_VAL_PTR changes.
8247 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8250 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8252 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8254 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8255 constant definitions.
8256 ("nop"): lr 0,0 -> nopr r0
8257 ("nop_lr0", "nop_lr1"): New insn definitions.
8259 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8261 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8262 NDS32_V3PUSH_AVAILABLE_P macro.
8264 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8265 Chung-Ju Wu <jasonwucj@gmail.com>
8267 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8268 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8269 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8270 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8271 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8272 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8273 * config/nds32/constraints.md: New constraints and checking for hard
8274 float configuration.
8275 * config/nds32/iterators.md: New mode iterator and attribute for hard
8276 float configuration.
8277 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8279 * config/nds32/nds32-fpu.md: New file.
8280 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8281 deal with hard float code generation.
8282 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8284 (abi_type, float_reg_number): New enum type.
8285 * config/nds32/nds32-predicates.c: New predicates for hard float.
8286 * config/nds32/nds32-protos.h: Declare functions for hard float.
8287 * config/nds32/nds32.c: Implementation for hard float configuration.
8288 * config/nds32/nds32.h: Definitions for hard float configuration.
8289 * config/nds32/nds32.md: Include hard float machine description and
8290 modify patterns for hard float configuration.
8291 * config/nds32/nds32.opt: New options for hard float configuration.
8292 * config/nds32/predicates.md: New predicates for hard float
8295 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8297 * common/config/nds32/nds32-common.c
8298 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8300 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8303 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8304 CONSTRUCTOR_ELT (ctor, ...)->value.
8306 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8309 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8311 2018-04-05 Tom de Vries <tom@codesourcery.com>
8314 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8317 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8318 Kito Cheng <kito.cheng@gmail.com>
8320 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8321 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8322 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8323 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8325 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8326 Kito Cheng <kito.cheng@gmail.com>
8328 * config/nds32/constraints.md (Ufe): New memory constraint.
8329 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8330 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8331 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8333 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8334 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8336 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8338 * config/nds32/nds32.md: Use optimize_size in the condition for
8339 alu-shift instructions.
8341 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8343 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8345 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8347 * config/nds32/nds32.md (negsi2): Refine pattern.
8349 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8350 Chung-Ju Wu <jasonwucj@gmail.com>
8352 * config/nds32/iterators.md (shift_rotate): New code iterator.
8353 (shift): New code attribute.
8354 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8355 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8356 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8357 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8358 bit-wise operations.
8359 (andsi3, *andsi3): Ditto.
8360 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8361 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8362 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8363 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8364 nds32_ior_operand, nds32_xor_operand): New predicates.
8366 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8368 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8369 (addsi3, subsi3): ... this.
8371 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8373 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8375 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8377 * config/nds32/nds32.md: Adjust indention.
8379 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8381 * config/nds32/nds32.md (feature): New attribute.
8383 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8385 * config/nds32/nds32.md (subtype): New attribute.
8387 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8390 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8391 expansion to perform a bitwise AND of the argument followed by a
8392 boolean negation of the result.
8394 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8396 PR rtl-optimization/84878
8397 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8398 the basic block. Assert the use reference is not artificial and that
8399 it has an associated insn.
8401 2018-04-04 Michael Matz <matz@suse.de>
8403 * builtins.c (compute_objsize): Pass correct operand
8404 to array_at_struct_end_p.
8406 2018-04-04 Richard Biener <rguenther@suse.de>
8409 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8410 from contexts for DINFO_LEVEL_TERSE and below.
8412 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8414 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8415 resiter_operand condition.
8416 * config/nds32/nds32.md (*move<mode>): Ditto.
8418 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8419 Monk Chiang <sh.chiang04@gmail.com>
8421 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8423 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8425 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8427 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8428 Kito Cheng <kito.cheng@gmail.com>
8430 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8431 nds32_cond_code_str, output_cond_branch,
8432 output_cond_branch_compare_zero, nds32_expand_cbranch,
8433 nds32_expand_cstore, nds32_expand_movcc,
8434 nds32_output_cbranchsi4_equality_zero,
8435 nds32_output_cbranchsi4_equality_reg,
8436 nds32_output_cbranchsi4_equality_reg_or_const_int,
8437 nds32_output_cbranchsi4_greater_less_zero: New functions.
8438 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8439 nds32_expand_cstore, nds32_expand_movcc,
8440 nds32_output_cbranchsi4_equality_zero,
8441 nds32_output_cbranchsi4_equality_reg,
8442 nds32_output_cbranchsi4_equality_reg_or_const_int,
8443 nds32_output_cbranchsi4_greater_less_zero): Declare.
8444 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8445 nds32_rimm11s_operand): New predicates.
8446 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8447 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8450 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8452 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8453 * config/nds32/nds32.md: Ditto.
8454 * config/nds32/pipelines.md: Ditto.
8456 2018-04-04 Richard Biener <rguenther@suse.de>
8458 PR tree-optimization/85168
8459 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8460 propagating abnormals.
8462 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8464 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8466 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8467 Kito Cheng <kito.cheng@gmail.com>
8469 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8470 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8471 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8472 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8473 * config/nds32/nds32.md (sibcall_internal): New.
8474 (sibcall_register): Remove.
8475 (sibcall_immediate): Remove.
8476 (sibcall_value_internal): New.
8477 (sibcall_value_register): Remove.
8478 (sibcall_value_immediate): Remove.
8479 * config/nds32/predicates.md (nds32_general_register_operand): New.
8480 (nds32_call_address_operand): New.
8482 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8484 PR rtl-optimization/85167
8485 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8486 bb_defs if *split_p, instead preinitialize it to NULL.
8488 PR tree-optimization/85156
8489 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8490 evaluating the argument multiple times.
8492 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8494 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8496 (_mm_cvtpd_ps): Likewise.
8497 (_mm_cvttpd_epi32): Likewise.
8498 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8499 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8500 vector, pixel, and bool following altivec.h include.
8502 2018-04-03 Martin Sebor <msebor@redhat.com>
8504 * doc/extend.texi (Common Function Attributes): Clarify.
8505 (const attribute): Likewise.
8506 (pure attribute): Likewise.
8508 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8511 * config/i386/i386.c (ix86_expand_vector_set): Use
8512 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8514 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8516 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8517 instructions when changing rounding bits to preserve precision bits
8518 in the x87 control word.
8520 2018-04-03 Martin Liska <mliska@suse.cz>
8522 PR tree-optimization/82491
8523 * rtl.h (strip_offset_and_add): Replace += suboffset with
8524 poly_uint64 () + suboffset.
8526 2018-03-29 Martin Liska <mliska@suse.cz>
8527 Martin Jambor <mjambor@suse.cz>
8530 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8531 param_type is not an integral or pointer type.
8533 2018-04-03 Richard Biener <rguenther@suse.de>
8535 * sese.h (recompute_all_dominators): Remove.
8537 2018-04-02 Martin Sebor <msebor@redhat.com>
8539 * doc/invoke.texi (-Wrestrict): Fix typos.
8541 2018-04-02 Jim Wilson <jimw@sifive.com>
8543 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8544 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8545 (<optab>di3, <optab>si3_extend): Likewise.
8546 (<optab>si3_mask, <optab>si3_mask_1): New.
8547 (<optab>di3_mask, <optab>di3_mask_1): New.
8548 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8549 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8550 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8552 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8554 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8557 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8559 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8560 (nds32_canonicalize_comparison): New function.
8562 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8563 Kito Cheng <kito.cheng@gmail.com>
8564 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8566 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8567 * config/nds32/constants.md (unspec_volatile_element): Add
8568 UNSPEC_VOLATILE_RELAX_GROUP.
8569 * config/nds32/nds32-relax-opt.c: New file.
8570 * config/nds32/nds32-predicates.c
8571 (nds32_symbol_load_store_p): New function.
8572 * config/nds32/nds32-protos.h
8573 (nds32_symbol_load_store_p): Declare function.
8574 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8575 * config/nds32/nds32.c
8576 (nds32_register_pass): New function to register pass.
8577 (nds32_register_passes): New function to register passes.
8578 * config/nds32/nds32.md (relax_group): New pattern.
8579 * config/nds32/nds32.opt (mrelax-hint): New option.
8580 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8582 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8584 * config/nds32/t-nds32: Modify files dependency.
8586 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8588 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8589 (PROFILE_HOOK): Define its implementation.
8591 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8593 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8594 type and 32-bit size.
8596 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8599 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8600 (V_128_256): New mode iterator.
8601 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8602 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8603 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8605 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8606 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8608 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8611 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8612 NaN inputs correctly.
8614 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8617 * config/rs6000/vsx.md (??r): New mode attribute.
8618 (*vsx_mov<mode>_64bit): Use it.
8619 (*vsx_mov<mode>_32bit): Likewise.
8621 2018-03-30 Martin Sebor <msebor@redhat.com>
8623 PR tree-optimization/84818
8624 * builtins.c (check_access): Use warning_n.
8626 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8629 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8631 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8634 2018-03-30 Julia Koval <julia.koval@intel.com>
8637 * x86-tune.def (movx, partial_reg_dependency): Enable for
8640 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8643 * lra-constraints.c (process_alt_operands): Move setting
8644 this_alternative_matches below.
8646 2018-03-29 Martin Liska <mliska@suse.cz>
8649 * doc/invoke.texi: Document how LTO works with debug info.
8650 Describe auto-load support of binutils. Mention 'x86-64'
8651 as valid option value of -march option.
8653 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8655 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8658 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8659 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8660 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8663 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8666 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8667 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8668 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8669 for RS6000_BTM_POWERPC64.
8670 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8671 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8672 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8677 2018-03-28 Carl Love <cel@us.ibm.com>
8680 2017-09-27 Carl Love <cel@us.ibm.com>
8682 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8683 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8684 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8687 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8689 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8690 instead of __vector bool.
8691 (_mm_max_pu8): Likewise.
8692 (_mm_min_pi16): Likewise.
8694 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8697 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8698 (DIVWEUO): Likewise.
8700 (DIVDEUO): Likewise.
8701 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8702 DIVWEUO and DIVDEUO.
8703 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8704 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8705 (div_extend): Likewise.
8706 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8708 (__builtin_divweuo): Likewise.
8709 (__builtin_divdeo): Likewise.
8710 (__builtin_divdeuo): Likewise.
8712 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8715 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8716 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8718 PR tree-optimization/82004
8719 * gimple-match-head.c (optimize_pow_to_exp): New function.
8720 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8721 Don't fold to exp if optimize_pow_to_exp is false.
8723 2018-03-28 Martin Liska <mliska@suse.cz>
8726 * calls.c (initialize_argument_information): Fix trailing space.
8727 * common.opt: Fix typo and provide better explanation for
8728 -fsanitize-coverage option.
8729 * config/i386/i386.opt: Fix typo.
8731 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8732 Martin Liska <mliska@suse.cz>
8735 * gimplify.c (asan_poison_variable): Don't do the check for
8736 gimplify_omp_ctxp here.
8737 (gimplify_decl_expr): Do it here.
8738 (gimplify_target_expr): Likewise.
8740 2018-03-28 Martin Liska <mliska@suse.cz>
8743 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8744 chkp_type_bounds_count if MPX is not enabled.
8746 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8748 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8750 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8753 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8754 function to create the function decl for complex long double
8755 multiply and divide for -mabi=ieeelongdouble.
8756 (init_float128_ieee): Call it.
8758 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8761 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8762 -fcf-protection=branch -mibt.
8763 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8765 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8768 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8769 arm_word_relocations.
8771 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8774 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8775 extern array declarations.
8777 2018-03-27 Richard Biener <rguenther@suse.de>
8780 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8781 explicit single_use checks.
8783 2018-03-27 Richard Biener <rguenther@suse.de>
8785 PR tree-optimization/85082
8786 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8789 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8791 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8792 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8793 Turn on fasynchronous-unwind-tables and funwind-tables.
8795 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8798 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8799 (*bmi_blsr_<mode>_ccz): Ditto.
8801 2018-03-26 Tom de Vries <tom@codesourcery.com>
8803 PR tree-optimization/85063
8804 * omp-general.c (offloading_function_p): New function. Factor out
8806 * omp-offload.c (pass_omp_target_link::gate): ... here.
8807 * omp-general.h (offloading_function_p): Declare.
8808 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8809 with attribute omp declare target for offloading functions.
8811 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8813 PR tree-optimization/84005
8814 * tree-data-ref.h (get_base_for_alignment): Declare.
8815 * tree-data-ref.c (get_base_for_alignment_1): New function.
8816 (get_base_for_alignment): Likewise.
8817 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8818 get_base_for_alignment to find a suitable base object, instead
8819 of always using drb->base_address.
8821 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8824 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8825 known size by default.
8827 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8830 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8831 and non BLKmode operands.
8833 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8836 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8837 Clean up attributes.
8839 2018-03-23 Richard Biener <rguenther@suse.de>
8842 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8843 we are going to emit early debug for LTO.
8845 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8848 * function.c (match_asm_constraints_1): Don't optimize if input
8849 doesn't satisfy general_operand predicate for output's mode.
8852 * alias.c (write_dependence_p): Don't require for x_canonicalized
8853 non-VOIDmode if x has VOIDmode.
8856 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8857 just don't try to optimize it rather than assert it never happens.
8859 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8861 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8862 macro expansions for definition of ST_INTERNAL_<mode> and
8863 LD_INTERNAL_<mode> builtins.
8864 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8866 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8868 (altivec_expand_st_builtin): Likewise.
8869 (altivec_expand_builtin): Remove calls to deleted functions.
8870 (rs6000_address_for_altivec): Delete this function.
8871 * config/rs6000/vector.md: Remove expands for
8872 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8874 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8877 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8878 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8879 re-computing once computed.
8880 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8881 (arm_init_machine_status): Initialize
8882 machine->static_chain_stack_bytes.
8884 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8887 * doc/extend.texi: Add four new prototypes for vec_ld.
8888 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8889 definitions for more logical presentation.
8890 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8891 entries for V1TI variants of __builtin_altivec_ld builtin.
8892 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8893 handling of V1TI variant of LVX icode pattern.
8894 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8895 (rs6000_gimple_fold_builtin): Likewise.
8896 (altivec_init_builtins): Add code to define
8897 __builtin_altivec_lvx_v1ti function.
8899 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8902 * function.c (match_asm_constraints_1): Don't do the optimization
8903 if input isn't a REG, SUBREG, MEM or constant.
8905 2018-03-22 Tom de Vries <tom@codesourcery.com>
8907 PR tree-optimization/84956
8908 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8909 bb_has_abnormal_pred.
8911 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8914 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8915 DECL_INITIAL (decl) to decl at the end.
8916 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8919 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8921 * doc/extend.texi (__builtin_tgmath): Document when complex
8922 integer types are treated as _Complex _Float64.
8924 2018-03-21 Tom de Vries <tom@codesourcery.com>
8926 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8928 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8930 PR tree-optimization/84960
8931 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8932 if it is ENTRY block, move them into single succ of ENTRY in that case.
8934 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8936 PR tree-optimization/84811
8937 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8938 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8939 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8940 (known_subrange_p): Update accordingly. Cast each value involved
8941 in the size comparison, rather than casting the result of the
8944 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8946 PR tree-optimization/84982
8947 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8948 by flipping the least significant bit rather than all bits from
8949 bitpos to bitpos + bitsize - 1.
8951 2018-03-21 Nathan Sidwell <nathan@acm.org>
8953 * doc/extend.texi (Deprecated Features): Remove mention of
8954 long-deleted deprecations.
8956 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8959 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8960 * configure: Regenerate.
8962 2018-03-21 Tom de Vries <tom@codesourcery.com>
8964 PR tree-optimization/83126
8965 * tree-parloops.c (num_phis): New function.
8966 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8968 2018-03-21 Nathan Sidwell <nathan@acm.org>
8970 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8971 mention anon-struct/union members and trailing attributes.
8973 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8975 PR tree-optimization/84969
8976 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8977 builtin memset partitions if they set different rhs values.
8979 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8981 PR rtl-optimization/84989
8982 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8983 VEC_DUPLICATE with scalar result mode.
8985 2018-03-21 Martin Liska <mliska@suse.cz>
8988 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8989 not intended return statement.
8991 2018-03-21 Martin Liska <mliska@suse.cz>
8994 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8995 (chkp_find_bound_slots_1): Limit number of iterations.
8997 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9000 * Minor grammar fixes for x86 options.
9002 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9005 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9006 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9008 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9011 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9012 (altivec_lvx_<mode>_1op): Likewise.
9013 (altivec_stvx_<mode>_2op): Likewise.
9014 (altivec_stvx_<mode>_1op): Likewise.
9015 (altivec_lvx_<VM2:mode>): New define_expand.
9016 (altivec_stvx_<VM2:mode>): Likewise.
9017 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9018 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9019 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9020 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9021 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9022 (rs6000_gen_lvx): Likewise.
9023 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9024 (altivec_expand_stv_builtin): Likewise.
9025 (altivec_expand_builtin): Likewise.
9026 * config/rs6000/vector.md: Likewise.
9028 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9031 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9034 2018-03-20 Richard Biener <rguenther@suse.de>
9037 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9038 sign-conversions as zero, fall back to standard scalar_stmt
9041 2018-03-20 Martin Liska <mliska@suse.cz>
9044 * predict.c (rebuild_frequencies): Handle case when we have
9045 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9047 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9050 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9051 flag_section_anchors.
9052 * varasm.c (use_blocks_for_decl_p): Remove hack for
9053 dw2_force_const_mem.
9056 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9058 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9059 be created, use lowpart_subreg of operands[0] rather than operands[0]
9061 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9062 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9063 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9064 and n constraint instead of aarch64_shift_imm_di and Usd.
9065 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9066 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9068 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9071 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9072 to favor GPR over NEON registers.
9073 (<shift>di3_neon): Likewise.
9075 2018-03-20 Tom de Vries <tom@codesourcery.com>
9078 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9079 (nvptx_process_pars): Emit bar.sync asap and alap.
9081 2018-03-20 Tom de Vries <tom@codesourcery.com>
9084 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9085 seen_label if seen_label is already set.
9087 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9090 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9091 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9092 Use 1U instead of 1. Formatting fixes.
9095 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9096 instead of TREE_TYPE (s1) for the return value.
9098 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9100 PR tree-optimization/84946
9101 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9102 bitsize + bitsize in poly_uint64 rather than poly_int64.
9105 * dwarf2asm.c: Include fold-const.c.
9106 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9107 of decl rather than decl itself.
9109 PR rtl-optimization/84643
9110 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9112 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9115 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9116 calling assemble_variable.
9118 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9121 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9122 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9124 2018-03-19 Jim Wilson <jimw@sifive.com>
9127 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9128 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9129 (riscv_first_stack_step): Likewise.
9130 (riscv_option_override): Use STACK_BOUNDARY instead of
9132 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9134 (BIGGEST_ALIGNMENT): Set to 128.
9135 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9136 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9139 2018-03-19 Richard Biener <rguenther@suse.de>
9141 PR tree-optimization/84933
9142 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9143 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9145 2018-03-19 Richard Biener <rguenther@suse.de>
9147 PR tree-optimization/84859
9148 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9149 (cond_if_else_store_replacement): Perform sinking operation on
9150 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9151 Generalize what a BB with a single eligible store is.
9153 2018-03-19 Richard Biener <rguenther@suse.de>
9155 PR tree-optimization/84929
9156 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9157 chrec_is_positive against non-chrec arg.
9159 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9162 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9164 2018-03-18 Martin Liska <mliska@suse.cz>
9166 PR rtl-optimization/84635
9167 * regrename.c (build_def_use): Use matches_mode only when
9170 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9172 PR tree-optimization/84913
9173 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9174 vectorize chains of COND_EXPRs.
9176 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9178 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9180 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9182 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9184 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9186 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9188 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9189 Kito Cheng <kito.cheng@gmail.com>
9191 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9192 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9193 (nds32_adjust_reg_alloc_order): New function.
9194 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9196 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9198 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9199 nds32_print_operand, nds32_print_operand_address): Use
9200 HOST_WIDE_INT_PRINT_DEC instead.
9202 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9204 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9206 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9209 * config/i386/i386.c (initial_ix86_tune_features,
9210 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9212 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9213 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9214 rather than 1u << ix86_tune. Formatting fix.
9215 (ix86_option_override_internal): Change ix86_arch_mask from
9216 unsigned int to unsigned HOST_WIDE_INT, initialize to
9217 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9218 (ix86_function_specific_restore): Likewise.
9220 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9223 * postreload.c (reload_combine_recognize_pattern): Perform
9224 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9225 truncate_int_for_mode the result for the destination's mode.
9228 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9229 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9231 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9234 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9237 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9238 regno_allocno_class_array and sorted_pseudos.
9239 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9240 insns where regno is used.
9242 2018-03-16 Martin Liska <mliska@suse.cz>
9245 * multiple_target.c (create_dispatcher_calls): Redirect
9246 reference in the symbol table.
9248 2018-03-16 Martin Liska <mliska@suse.cz>
9251 * multiple_target.c (create_dispatcher_calls): Redirect also
9254 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9258 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9259 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9260 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9262 2018-03-16 Julia Koval <julia.koval@intel.com>
9264 * doc/invoke.texi (Skylake Server): Add CLWB.
9265 Cannonlake): Remove CLWB.
9267 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9269 PR tree-optimization/84841
9270 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9272 (FLOAT_ONE_CONST_TYPE): Define.
9273 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9274 (sort_by_operand_rank): Put entries with higher constant_type last
9275 rather than first to match comments.
9277 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9279 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9282 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9285 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9286 check and use address of target always.
9288 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9291 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9292 (indirect_thunk_bnd_needed): Likewise.
9293 (indirect_thunks_used): Likewise.
9294 (indirect_thunks_bnd_used): Likewise.
9295 (indirect_return_needed): New.
9296 (indirect_return_bnd_needed): Likewise.
9297 (output_indirect_thunk_function): Add a bool argument for
9299 (output_indirect_thunk_function): Don't generate alias for
9300 function return thunk.
9301 (ix86_code_end): Call output_indirect_thunk_function to generate
9302 function return thunks.
9303 (ix86_output_function_return): Set indirect_return_bnd_needed
9304 and indirect_return_needed instead of indirect_thunk_bnd_needed
9305 and indirect_thunk_needed.
9307 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9309 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9310 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9311 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9313 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9314 Paul Hua <paul.hua.gm@gmail.com>
9317 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9319 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9321 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9322 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9325 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9328 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9329 instead of GET_MODE_SIZE when comparing Units.
9331 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9334 * varasm.c (hash_section): Return an unchangeble hash value
9335 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9336 Return !aarch64_can_use_per_function_literal_pools_p ().
9338 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9341 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9342 rather than address of cmode as last argument to prepare_cmp_insn.
9344 2018-03-15 Julia Koval <julia.koval@intel.com>
9346 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9347 F_AVX512VNNI, F_AVX512BITALG): New.
9349 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9352 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9353 insn for floating-point loads and stores.
9355 2018-03-14 Carl Love <cel@us.ibm.com>
9357 * config/rs6000/rs6000-c.c: Add macro definitions for
9358 ALTIVEC_BUILTIN_VEC_PERMXOR.
9359 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9360 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9361 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9362 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9364 * config/doc/extend.texi: Add prototypes for vec_permxor.
9366 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9369 * diagnostic-show-locus.c (class layout_point): Convert m_line
9370 from int to linenum_type.
9371 (line_span::comparator): Use linenum "compare" function when
9372 comparing line numbers.
9373 (test_line_span): New function.
9374 (layout_range::contains_point): Convert param "row" from int to
9376 (layout_range::intersects_line_p): Likewise.
9377 (layout::will_show_line_p): Likewise.
9378 (layout::print_source_line): Likewise.
9379 (layout::should_print_annotation_line_p): Likewise.
9380 (layout::print_annotation_line): Likewise.
9381 (layout::print_leading_fixits): Likewise.
9382 (layout::annotation_line_showed_range_p): Likewise.
9383 (struct line_corrections): Likewise for field m_row.
9384 (line_corrections::line_corrections): Likewise for param "row".
9385 (layout::print_trailing_fixits): Likewise.
9386 (layout::get_state_at_point): Likewise.
9387 (layout::get_x_bound_for_row): Likewise.
9388 (layout::print_line): Likewise.
9389 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9390 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9391 * input.c (selftest::test_linenum_comparisons): New function.
9392 (selftest::input_c_tests): Call it.
9393 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9394 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9395 * selftest.h (ASSERT_GT): New macro.
9396 (ASSERT_GT_AT): New macro.
9397 (ASSERT_LT): New macro.
9398 (ASSERT_LT_AT): New macro.
9400 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9402 PR rtl-optimization/84780
9403 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9405 2018-03-14 Martin Liska <mliska@suse.cz>
9407 * tree.c (record_node_allocation_statistics): Use
9408 get_stats_node_kind.
9409 (get_stats_node_kind): New function extracted from
9410 record_node_allocation_statistics.
9411 (free_node): Use get_stats_node_kind.
9413 2018-03-14 Richard Biener <rguenther@suse.de>
9415 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9416 that the value-set of ANTIC_IN doesn't grow.
9419 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9421 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9422 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9424 2018-03-14 Julia Koval <julia.koval@intel.com>
9426 * config.gcc (icelake-client, icelake-server): New.
9428 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9429 (initial_ix86_arch_features): Ditto.
9430 (PTA_SKYLAKE): Add SGX.
9431 (PTA_ICELAKE): Remove.
9432 (PTA_ICELAKE_CLIENT): New.
9433 (PTA_ICELAKE_SERVER): New.
9434 (ix86_option_override_internal): Split up icelake on icelake client and
9436 (get_builtin_code_for_version): Ditto.
9437 (fold_builtin_cpu): Ditto.
9438 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9439 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9440 * config/i386/i386.h (processor_type): Ditto.
9441 * doc/invoke.texi: Ditto.
9443 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9446 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9447 INTEGER_CST offset, add it together with bitpos / 8 and
9448 sign extend based on POINTER_SIZE.
9452 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9455 * config/i386/constraints.md (Yc): New register constraint.
9456 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9457 Use Yc constraint for alternative 2 of operand 0. Remove
9458 preferred_for_speed attribute.
9460 2018-03-14 Richard Biener <rguenther@suse.de>
9462 PR tree-optimization/84830
9463 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9464 with the old one to avoid oscillations.
9466 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9469 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9471 (assign_by_spills): Return a flag of reload assignment failure.
9472 Do not process the reload assignment failures. Do not spill other
9473 reload pseudos if they has the same reg class. Update n if
9475 (lra_assign): Add a return arg. Set up from the result of
9476 assign_by_spills call.
9477 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9478 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9479 usage_insns if it is not NULL.
9480 (spill_hard_reg_in_range): New function.
9481 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9482 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9483 function prototypes.
9484 (lra_assign): Change prototype.
9485 * lra.c (lra): Add code to deal with fails by splitting hard reg
9488 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9490 * config/riscv/riscv.opt (mrelax): New option.
9491 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9492 "norelax" when riscv_mrelax is disabled.
9493 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9495 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9498 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9499 reassociation for int modes.
9501 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9503 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9504 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9506 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9507 * config/aarch64/aarch64-sve.md
9508 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9509 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9510 (*extend<mode><Vwide>2): Rename to...
9511 (aarch64_sve_extend<mode><Vwide>2): ...this.
9512 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9513 renaming the old pattern to...
9514 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9516 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9517 define_expand, renaming the old pattern to...
9518 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9519 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9520 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9521 account when deciding which SVE instruction the optab should use.
9522 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9524 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9526 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9527 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9528 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9529 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9530 (tlsdesc_small_<mode>): Turn a define_expand and use
9531 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9532 (tlsdesc_small_advsimd_<mode>): ...this.
9533 (tlsdesc_small_sve_<mode>): New pattern.
9535 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9537 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9538 (UNSPEC_UMUL_HIGHPART): New constants.
9539 (MUL_HIGHPART): New int iteraor.
9540 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9541 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9543 (*<su>mul<mode>3_highpart): New define_insn.
9545 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9548 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9551 2018-03-13 Martin Liska <mliska@suse.cz>
9554 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9556 (sem_item_optimizer::~sem_item_optimizer): Release it.
9557 (sem_item_optimizer::merge_classes): Register variable aliases.
9558 (sem_item_optimizer::fixup_pt_set): New function.
9559 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9560 * ipa-icf.h: Declare new variables and functions.
9562 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9565 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9566 integer_pow2p@2 and test integer_pow2p in condition.
9567 (A < 0 ? C : 0): Similarly for @1.
9570 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9571 characters starting at p contain '\0' character, don't look beyond
9575 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9576 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9579 * reg-stack.c (change_stack): Change update_end var from int to
9580 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9581 also call set_block_for_insn on the newly added insns and rescan.
9584 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9585 on the last operand.
9588 * tree.c (stabilize_reference_1): Return save_expr (e) for
9589 STATEMENT_LIST even if it doesn't have side-effects.
9591 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9593 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9595 2018-03-12 Renlin Li <renlin.li@arm.com>
9597 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9598 aarch64_output_scalar_simd_mov_immediate.
9600 2018-03-12 Martin Sebor <msebor@redhat.com>
9602 PR tree-optimization/83456
9603 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9604 for perfectly overlapping calls to memcpy.
9605 (gimple_fold_builtin_memory_chk): Same.
9606 (gimple_fold_builtin_strcpy): Handle no-warning.
9607 (gimple_fold_builtin_stxcpy_chk): Same.
9608 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9610 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9612 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9613 parameter. Use it for SFmode.
9614 (rs6000_function_arg_advance_1): Adjust.
9615 (rs6000_function_arg): Adjust.
9616 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9618 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9620 PR rtl-optimization/84169
9621 PR rtl-optimization/84780
9622 * combine.c (can_combine_p): Check for a 2-insn combination whether
9623 the destination register is used between the two insns, too.
9625 2018-03-12 Richard Biener <rguenther@suse.de>
9627 PR tree-optimization/84803
9628 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9629 for refs DR analysis didn't process.
9631 2018-03-12 Richard Biener <rguenther@suse.de>
9633 PR tree-optimization/84777
9634 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9635 force-vectorize loops ignore whether we are optimizing for size.
9637 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9639 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9640 (TARGET_MD_ASM_ADJUST): Define.
9642 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9643 Kito Cheng <kito.cheng@gmail.com>
9644 Chung-Ju Wu <jasonwucj@gmail.com>
9646 * config/nds32/nds32.c (nds32_compute_stack_frame,
9647 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9648 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9649 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9650 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9651 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9652 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9653 * config/nds32/nds32.md (prologue, epilogue): Use macro
9654 NDS32_V3PUSH_AVAILABLE_P to do checking.
9656 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9659 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9660 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9661 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9662 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9663 addition of most attributes on !orig_type_die or the attribute not
9664 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9666 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9667 Chung-Ju Wu <jasonwucj@gmail.com>
9669 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9671 * config/nds32/nds32.opt (mvh): New option.
9673 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9674 Chung-Ju Wu <jasonwucj@gmail.com>
9676 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9678 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9679 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9682 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9683 Chung-Ju Wu <jasonwucj@gmail.com>
9685 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9687 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9688 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9690 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9691 Kito Cheng <kito.cheng@gmail.com>
9692 Chung-Ju Wu <jasonwucj@gmail.com>
9694 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9695 UNSPEC_FFMISM and UNSPEC_FLMISM.
9696 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9697 for ffb, ffmism and flmism.
9698 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9699 (unspec_ffmism): Ditto.
9700 (unspec_flmism): Ditto.
9701 (nds32_expand_builtin_impl): Check if string extension is available.
9702 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9703 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9705 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9708 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9711 * lra-assigns.c (assign_by_spills): Return a flag of reload
9712 assignment failure. Do not process the reload assignment
9713 failures. Do not spill other reload pseudos if they has the same
9715 (lra_assign): Add a return arg. Set up from the result of
9716 assign_by_spills call.
9717 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9718 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9719 usage_insns if it is not NULL.
9720 (spill_hard_reg_in_range): New function.
9721 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9722 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9723 function prototypes.
9724 (lra_assign): Change prototype.
9725 * lra.c (lra): Add code to deal with fails by splitting hard reg
9728 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9731 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9733 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9736 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9737 (dw_val_node): Add val_symbolic_view.
9738 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9739 (symview_upper_bound): New.
9740 (new_line_info_table): Initialize symviews_since_reset.
9741 (dwarf2out_source_line): Count symviews_since_reset and set
9742 symview_upper_bound.
9743 (dw_val_equal_p): Handle symview.
9744 (add_AT_symview): New.
9745 (print_dw_val): Handle symview.
9746 (attr_checksum, attr_checksum_ordered): Likewise.
9747 (same_dw_val_p, size_of_die): Likewise.
9748 (value_format, output_die): Likewise.
9749 (add_high_low_attributes): Use add_AT_symview for entry_view.
9750 (dwarf2out_finish): Reset symview_upper_bound, clear
9753 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9756 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9757 Add strict argument and use it.
9758 (rs6000_split_multireg_move): Update for new strict argument.
9759 (mem_operand_gpr): Disallow all non-offsettable addresses.
9760 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9762 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9765 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9766 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9767 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9770 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9771 decl, use remap_type if we want to use the type.
9773 2018-03-09 Martin Sebor <msebor@redhat.com>
9775 PR tree-optimization/84526
9776 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9778 (builtin_access::generic_overlap): Be prepared to handle non-array
9781 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9783 PR rtl-optimization/84682
9784 * lra-constraints.c (process_address_1): Check is_address flag
9785 for address constraints.
9786 (process_alt_operands): Likewise.
9787 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9788 preprocess_constraints.
9789 * recog.h (preprocess_constraints): Add oploc parameter.
9791 * recog.c (preprocess_constraints): Test address_operand for
9792 CT_ADDRESS constraints.
9794 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9797 * lra-assigns.c (assign_by_spills): Return a flag of reload
9798 assignment failure. Do not process the reload assignment
9799 failures. Do not spill other reload pseudos if they has the same
9801 (lra_assign): Add a return arg. Set up from the result of
9802 assign_by_spills call.
9803 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9804 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9805 usage_insns if it is not NULL.
9806 (spill_hard_reg_in_range): New function.
9807 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9808 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9809 function prototypes.
9810 (lra_assign): Change prototype.
9811 * lra.c (lra): Add code to deal with fails by splitting hard reg
9814 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9817 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9818 Accept complain bool parameter. Only emit errors if it is true.
9819 (arm_parse_cpu_option_name): Likewise.
9820 (arm_target_thumb_only): Adjust callers of the above.
9821 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9822 prototype to take a default true bool parameter.
9823 (arm_parse_arch_option_name): Likewise.
9825 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9826 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9830 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9831 * configure: Regenerate.
9832 * configure.ac ("linker --version-script option"): New.
9833 ("linker soname option"): New.
9835 2018-03-09 Richard Biener <rguenther@suse.de>
9837 PR tree-optimization/84775
9838 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9839 immediate uses of predicate stmts and mark them modified.
9842 PR tree-optimization/84178
9843 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9845 (version_loop_for_if_conversion): Delay update_ssa call.
9846 (tree_if_conversion): Delay update_ssa until after predicate
9849 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9852 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9853 when the function accesses prior frames.
9855 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9858 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9859 gen_llsym, otherwise call maybe_gen_llsym.
9862 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9863 has ',' character inside of it.
9865 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9868 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9869 as clobbering CC_REGNUM.
9871 2018-03-08 Richard Biener <rguenther@suse.de>
9874 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9875 (follow_copies_to_constant): Do not follow SSA names registered
9878 2018-03-08 Richard Biener <rguenther@suse.de>
9880 PR tree-optimization/84178
9881 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9883 (version_loop_for_if_conversion): Delay update_ssa call.
9884 (tree_if_conversion): Delay update_ssa until after predicate
9887 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9889 PR tree-optimization/84178
9890 * tree-if-conv.c (release_bb_predicate): Remove the
9891 the assertion that the stmts have NULL use_ops.
9892 Discard the statements, asserting that they haven't
9893 yet been added to a BB.
9895 2018-03-08 Richard Biener <rguenther@suse.de>
9897 PR tree-optimization/84746
9898 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9899 (phi_translate): Pass in destination ANTIC_OUT set.
9900 (phi_translate_1): Likewise. For a simplified result lookup
9901 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9902 (phi_translate_set): Adjust.
9903 (do_pre_regular_insertion): Likewise.
9904 (do_pre_partial_partial_insertion): Likewise.
9906 2018-03-08 Martin Liska <mliska@suse.cz>
9908 PR gcov-profile/84735
9909 * doc/gcov.texi: Document usage of profile files.
9910 * gcov-io.h: Document changes in the format.
9912 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9916 * dwarf2out.c (struct dw_line_info_table): Update comments for
9918 (FORCE_RESET_NEXT_VIEW): New.
9919 (FORCE_RESETTING_VIEW_P): New.
9920 (RESETTING_VIEW_P): Check for -1 too.
9921 (ZERO_VIEW_P): Likewise.
9922 (new_line_info_table): Force-reset next view.
9923 (dwarf2out_begin_function): Likewise.
9924 (dwarf2out_source_line): Simplify zero_view_p initialization.
9925 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9926 view directly. Omit view when omitting .loc at line 0.
9928 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9930 PR tree-optimization/84740
9931 * tree-switch-conversion.c (process_switch): Call build_constructors
9932 only if info.phi_count is non-zero.
9934 PR tree-optimization/84739
9935 * tree-tailcall.c (find_tail_calls): Check call arguments against
9936 DECL_ARGUMENTS (current_function_decl) rather than
9937 DECL_ARGUMENTS (func) when checking for tail recursion.
9939 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9941 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9942 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9943 Volker Reichelt's entry and add entries for people that perform
9944 GCC fuzzy testing and report numerous bugs.
9946 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9949 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9950 readonly data in sdata, if that is disabled.
9951 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9952 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9953 -mreadonly-in-sdata option.
9955 2018-03-07 Martin Sebor <msebor@redhat.com>
9957 PR tree-optimization/84468
9958 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9959 basic block when looking for nul assignment.
9961 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9964 * except.h (output_function_exception_table): Adjust prototype.
9965 * except.c (output_function_exception_table): Remove FNNAME parameter
9966 and add SECTION parameter. Ouput one part of the table at a time.
9967 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9968 the first part of the exception table and emit unwind directives.
9969 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9970 (i386_pe_seh_cold_init): Likewise.
9971 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9972 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9973 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9974 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9975 * config/i386/winnt.c: Include except.h.
9976 (struct seh_frame_state): Add reg_offset, after_prologue and
9977 in_cold_section fields.
9978 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9979 (i386_pe_seh_cold_init): New function.
9980 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9981 to seh->in_cold_section.
9982 (seh_emit_push): Record the offset of the push.
9983 (seh_emit_save): Record the offet of the save.
9984 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9985 Test seh->after_prologue to disregard the epilogue.
9986 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9987 (i386_pe_end_cold_function): New function.
9989 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9992 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9993 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9996 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9998 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9999 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10001 PR middle-end/84723
10002 * multiple_target.c: Include tree-inline.h and intl.h.
10003 (expand_target_clones): Diagnose and fail if node->definition and
10004 !tree_versionable_function_p (node->decl).
10006 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10008 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10010 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10011 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10012 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10014 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10017 * combine.c (try_combine): Use reg_or_subregno instead of handling
10018 just paradoxical SUBREGs and REGs.
10020 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10022 * config/arc/arc.c (arc_finalize_pic): Remove function.
10023 (arc_must_save_register): We use single base PIC register, remove
10024 checks to save/restore the PIC register.
10025 (arc_expand_prologue): Likewise.
10026 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10028 (arc_verify_short): Likewise.
10029 (arc_attr_type): Likewise.
10030 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10031 (walk_stores): Likewise.
10032 (arc_address_cost): Make it static.
10033 (arc_verify_short): Likewise.
10034 (branch_dest): Likewise.
10035 (arc_attr_type): Likewise.
10036 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10037 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10038 (arc_final_prescan_insn): Remove inserting the nops due to
10039 hardware hazards. It is done in reorg step.
10040 (insn_length_variant_t): Remove.
10041 (insn_length_parameters_t): Likewise.
10042 (arc_insn_length_parameters): Likewise.
10043 (arc_get_insn_variants): Likewise.
10044 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10046 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10048 PR inline-asm/84683
10049 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10052 PR tree-optimization/84687
10053 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10055 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10057 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10059 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10060 Rename to ppc_speculation_barrier.
10061 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10062 __builtin_ppc_speculation_barrier.
10064 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10067 * combine.c (combine_simplify_rtx): Don't try to simplify if
10068 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10071 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10073 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10074 to 32 bytes when compiling for POWER9.
10076 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10079 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10080 regparm >= 3 with no arg reg available also for calls with
10081 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10084 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10086 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10088 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10091 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10093 2018-03-05 Richard Biener <rguenther@suse.de>
10095 PR tree-optimization/84486
10096 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10097 When inserting a __builtin_assume_aligned call set the LHS
10098 SSA name alignment info accordingly.
10100 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10102 PR tree-optimization/84114
10103 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10104 Avoid reassociation of FLOAT_MODE addition.
10106 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10108 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10109 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10110 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10111 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10112 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10113 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10114 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10116 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10117 __builtin_ia32_wbinvd): New builtins.
10118 (SPECIAL_ARGS2): New.
10119 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10120 (SPECIAL_ARGS2): New.
10121 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10122 (ix86_valid_target_attribute_inner_p): Ditto.
10123 (ix86_init_mmx_sse_builtins): Add special_args2.
10124 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10125 TARGET_WBNOINVD_P): New.
10126 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10127 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10128 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10129 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10130 * config/i386/pconfigintrin.h: New file.
10131 * config/i386/wbnoinvdintrin.h: Ditto.
10132 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10134 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10136 2018-03-05 Richard Biener <rguenther@suse.de>
10138 PR tree-optimization/84670
10139 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10141 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10142 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10143 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10144 doesn't grow if all successors (recursively) were visited at least
10147 2018-03-05 Richard Biener <rguenther@suse.de>
10149 PR tree-optimization/84650
10150 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10151 if executed in the loop pipeline.
10153 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10155 * doc/configfiles.texi (Configuration Files): Move info about
10156 conditionalizing $target-protos.h to...
10157 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10158 differs from $target-protos.h.
10160 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10161 Chung-Ju Wu <jasonwucj@gmail.com>
10163 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10164 * config/nds32/nds32-multiple.md (setmemsi): Define.
10165 * config/nds32/nds32-memory-manipulation.c
10166 (nds32_gen_dup_4_byte_to_word_value): New.
10167 (emit_setmem_word_loop): New.
10168 (emit_setmem_byte_loop): New.
10169 (nds32_expand_setmem_loop): New.
10170 (nds32_expand_setmem_loop_v3m): New.
10171 (nds32_expand_setmem_unroll): New.
10172 (nds32_expand_setmem): New.
10174 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10175 Chung-Ju Wu <jasonwucj@gmail.com>
10177 * config/nds32/nds32-memory-manipulation.c
10178 (nds32_emit_load_store): New.
10179 (nds32_emit_post_inc_load_store): New.
10180 (nds32_emit_mem_move): New.
10181 (nds32_emit_mem_move_block): New.
10182 (nds32_expand_movmemsi_loop_unknown_size): New.
10183 (nds32_expand_movmemsi_loop_known_size): New.
10184 (nds32_expand_movmemsi_loop): New.
10185 (nds32_expand_movmemsi_unroll): New.
10186 (nds32_expand_movmemqi): Rename ...
10187 (nds32_expand_movmemsi): ... to this.
10188 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10189 (movmemsi): ... to this.
10190 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10191 (nds32_expand_movmemsi): ... to this.
10193 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10194 Monk Chiang <sh.chiang04@gmail.com>
10195 Chung-Ju Wu <jasonwucj@gmail.com>
10197 * config/nds32/nds32-protos.h
10198 (nds32_expand_load_multiple): New arguments.
10199 (nds32_expand_store_multiple): Ditto.
10200 (nds32_valid_multiple_load_store): Rename ...
10201 (nds32_valid_multiple_load_store_p): ... to this.
10202 * config/nds32/nds32-memory-manipulation.c
10203 (nds32_expand_load_multiple): Refine implementation.
10204 (nds32_expand_store_multiple): Ditto.
10205 * config/nds32/nds32-multiple.md
10206 (load_multiple): Update nds32_expand_load_multiple interface.
10207 (store_multiple): Update nds32_expand_store_multiple interface.
10208 * config/nds32/nds32-predicates.c
10209 (nds32_valid_multiple_load_store): Rename ...
10210 (nds32_valid_multiple_load_store_p): ... to this and refine
10212 * config/nds32/predicates.md
10213 (nds32_load_multiple_and_update_address_operation): New predicate.
10214 (nds32_store_multiple_and_update_address_operation): New predicate.
10216 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10217 Chung-Ju Wu <jasonwucj@gmail.com>
10219 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10220 (combo): New attribute.
10221 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10223 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10225 * config/nds32/nds32.opt: Change -mcmodel= default value.
10227 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10228 Monk Chiang <sh.chiang04@gmail.com>
10229 Chung-Ju Wu <jasonwucj@gmail.com>
10231 * config/nds32/constants.md (unspec_element): New enum.
10232 * config/nds32/constraints.md (Umw): New constraint.
10233 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10234 * config/nds32/nds32-intrinsic.md: Likewise.
10235 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10236 (nds32_valid_smw_lwm_base_p): New.
10237 (nds32_output_smw_single_word): New.
10238 (nds32_output_lmw_single_word): New.
10239 (nds32_expand_unaligned_load): New.
10240 (nds32_expand_unaligned_store): New.
10241 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10242 (nds32_output_smw_single_word): Declare.
10243 (nds32_output_lmw_single_word): Declare.
10244 (nds32_expand_unaligned_load): Declare.
10245 (nds32_expand_unaligned_store): Declare.
10246 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10247 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10248 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10249 NDS32_BUILTIN_UASTORE_DW.
10250 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10253 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10254 Kito Cheng <kito.cheng@gmail.com>
10255 Chung-Ju Wu <jasonwucj@gmail.com>
10257 * config/nds32/nds32-intrinsic.c
10258 (nds32_expand_builtin_null_ftype_reg): Delete.
10259 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10260 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10261 (nds32_read_argument): New.
10262 (nds32_legitimize_target): Ditto.
10263 (nds32_legitimize_argument): Ditto.
10264 (nds32_check_constant_argument): Ditto.
10265 (nds32_expand_unop_builtin): Ditto.
10266 (nds32_expand_unopimm_builtin): Ditto.
10267 (nds32_expand_binop_builtin): Ditto.
10268 (nds32_builtin_decl_impl): Ditto.
10269 (builtin_description): Ditto.
10270 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10271 (nds32_init_builtins_impl): Ditto.
10272 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10273 (nds32_builtin_decl): New.
10274 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10275 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10277 2018-03-02 Jeff Law <law@redhat.com>
10279 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10280 (redundant_insn, fill_simple_delay_slots): Likewise.
10281 (fill_slots_from_thread): Likewise.
10282 * resource.c (mark_referenced_resources): Likewise.
10283 (mark_set_resources, find_dead_or_set_registers): Likewise.
10285 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10287 * substring-locations.h (format_warning_va): Formatting fix for
10288 ATTRIBUTE_GCC_DIAG.
10289 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10291 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10292 * substring-locations.c: Include intl.h.
10293 (format_warning_va): Turned into small wrapper around
10294 format_warning_n_va, renamed to ...
10295 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10296 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10298 (format_warning_at_substring_n): New function.
10299 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10300 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10301 format_warning_at_substring with just a shorter name instead of
10302 const function pointer.
10303 (fmtwarn_n): New function.
10304 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10305 appropriate, get rid of all the fmtstr temporaries, move conditionals
10306 with G_() wrapped string literals directly into fmtwarn arguments,
10307 cast dir.len to (int), formatting fixes.
10309 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10311 * doc/invoke.texi: Remove "Cilk Plus" references.
10313 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10314 Richard Biener <rguenther@suse.de>
10317 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10318 for error or warning attributes if CALL_FROM_THUNK_P is set.
10321 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10324 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10325 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10328 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10329 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10331 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10332 instead of -1U in last predictors element's probability member.
10334 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10337 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10338 arguments if they are comparable.
10340 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10342 PR tree-optimization/84634
10343 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10344 masks and masked_loop_p with a single loop_masks, making sure it's
10345 null for bb vectorization.
10347 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10349 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10350 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10351 loop->force_vectorize to check whether there is no alias.
10353 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10356 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10358 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10360 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10361 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10362 instead of a loop around prev_real_insn.
10363 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10366 PR inline-asm/84625
10367 * config/i386/i386.c (ix86_print_operand): Use conditional
10368 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10371 2018-03-02 Richard Biener <rguenther@suse.de>
10373 PR tree-optimization/84427
10374 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10375 (bitmap_set_subtract_values): Rewrite to handle multiple
10378 (prune_clobbered_mems): Likewise.
10379 (phi_translate): Take edge instead of pred/phiblock.
10380 (phi_translate_1): Likewise.
10381 (phi_translate_set): Likewise. Insert all translated
10382 exprs for a value into the set, keeping possibly multiple
10383 expressions per value.
10384 (compute_antic_aux): Adjust for phi_translate changes.
10385 When intersecting union the expressions and prune those
10386 not in the final value set, keeping possibly multiple
10387 expressions per value. Do not use value-insertion
10388 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10389 all expressions. Add verification that the value-sets
10390 only shrink during iteration.
10391 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10392 (do_pre_regular_insertion): Likewise.
10393 (do_pre_partial_partial_insertion): Likewise.
10395 2018-03-02 Richard Biener <rguenther@suse.de>
10398 * config/darwin.c (saved_debug_info_level): New static global.
10399 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10400 (darwin_asm_lto_end): Restore debug info generation settings.
10402 2018-03-01 Martin Liska <mliska@suse.cz>
10405 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10406 volatile arguments.
10408 2018-03-01 Richard Biener <rguenther@suse.de>
10411 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10412 type annotation with LTO when debug was disabled at compile-time.
10414 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10416 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10418 (mips_final_postscan_insn): Likewise.
10420 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10422 PR rtl-optimization/84528
10423 * alias.c (init_alias_target): Add commentary.
10424 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10425 a unique base value if the frame pointer is not eliminated
10426 to the stack pointer.
10428 2018-03-01 Tom de Vries <tom@codesourcery.com>
10430 PR rtl-optimization/83327
10431 * lra-int.h (hard_regs_spilled_into): Declare.
10432 * lra.c (hard_regs_spilled_into): Define.
10433 (init_reg_info): Init hard_regs_spilled_into.
10434 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10435 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10436 (process_bb_lives): Handle hard_regs_spilled_into.
10437 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10438 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10440 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10442 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10443 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10444 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10445 * config/rs6000/aix72.h: New file.
10447 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10449 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10450 instead of warning_at with conditional singular and plural messages
10454 * stor-layout.c (update_alignment_for_field): For
10455 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10456 && !DECL_PACKED (field), do the alignment update, just use
10457 only desired_align instead of MAX (type_align, desired_align)
10459 (place_field): Don't do known_align < desired_align handling
10460 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10461 is non-NULL, instead do it after rli->prev_field handling and
10462 only if not within a bitfield word. For DECL_PACKED (field)
10463 use type_align of BITS_PER_UNIT.
10465 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10467 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10468 superfluous parentheses and trailing spaces.
10470 2018-02-28 Richard Biener <rguenther@suse.de>
10472 PR tree-optimization/84584
10473 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10474 SCoPs with fake exit edge.
10476 2018-02-28 Martin Liska <mliska@suse.cz>
10479 * timevar.c (timer::print): Fix format to properly print 100%
10482 2018-02-28 Richard Biener <rguenther@suse.de>
10484 PR middle-end/84607
10485 * genmatch.c (capture_info::walk_match): Do not mark
10486 captured expressions without operands as expr_p given
10487 they act more like predicates and should be subject to
10488 "lost tail" side-effect preserving.
10490 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10492 PR rtl-optimization/81611
10493 * auto-inc-dec.c (attempt_change): Move dead note from
10494 mem_insn if it's the next use of regno
10495 (find_address): Take address use of reg holding
10496 non-incremented value. Add parm to limit search to the named
10498 (merge_in_block): Attempt to use a mem insn that is the next
10499 use of the original regno.
10501 2018-02-27 Martin Sebor <msebor@redhat.com>
10504 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10505 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10507 2018-02-27 Martin Sebor <msebor@redhat.com>
10509 PR translation/84207
10510 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10511 n argument to unsigned HOST_WIDE_INT.
10512 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10513 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10514 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10515 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10517 2018-02-27 Richard Biener <rguenther@suse.de>
10519 PR tree-optimization/84512
10520 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10521 Do not use the estimate returned from record_stmt_cost for
10522 the scalar iteration cost but sum properly using add_stmt_cost.
10524 2018-02-27 Richard Biener <rguenther@suse.de>
10526 PR tree-optimization/84466
10527 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10528 Adjust last change to less strictly validate use operands.
10530 2018-02-27 Martin Liska <mliska@suse.cz>
10532 PR gcov-profile/84548
10533 * gcov.c (process_file): Allow partial overlap and consider it
10534 also as group functions.
10535 (output_lines): Properly calculate range of lines for a group.
10537 2018-02-27 Martin Liska <mliska@suse.cz>
10539 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10540 'ggc' suffixes. Change first column width.
10541 (timer::print): Fix formatting of the column.
10543 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10545 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10546 preserve inline entry blocks for the sake of debug inline
10547 entry point markers alone.
10548 (remove_unused_locals): Suggest in comments a better place to
10549 force the preservation of inline entry blocks that are
10550 otherwise unused, but do not preserve them.
10552 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10554 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10556 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10559 * config/i386/constraints.md (Bs): Replace
10560 ix86_indirect_branch_register with
10561 TARGET_INDIRECT_BRANCH_REGISTER.
10563 * config/i386/i386.md (indirect_jump): Likewise.
10564 (tablejump): Likewise.
10565 (*sibcall_memory): Likewise.
10566 (*sibcall_value_memory): Likewise.
10567 Peepholes of indirect call and jump via memory: Likewise.
10568 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10569 (*sibcall_value_GOT_32): Likewise.
10570 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10571 (GOT_memory_operand): Likewise.
10572 (call_insn_operand): Likewise.
10573 (sibcall_insn_operand): Likewise.
10574 (GOT32_symbol_operand): Likewise.
10575 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10577 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10579 PR rtl-optimization/83496
10580 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10581 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10582 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10583 redundant insn, if any.
10584 (relax_delay_slots): Likewise.
10585 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10587 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10589 PR tree-optimization/83965
10590 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10591 that grouped statements are part of a reduction chain. Return
10592 true if the statement is not marked as a reduction itself but
10593 is part of a group.
10594 (vect_recog_dot_prod_pattern): Don't check whether the statement
10595 is part of a group here.
10596 (vect_recog_sad_pattern): Likewise.
10597 (vect_recog_widen_sum_pattern): Likewise.
10599 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10602 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10604 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10607 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10609 (ix86_output_indirect_function_return): New prototype.
10610 (ix86_split_simple_return_pop_internal): Likewise.
10611 * config/i386/i386.c (indirect_return_via_cx): New.
10612 (indirect_return_via_cx_bnd): Likewise.
10613 (indirect_thunk_name): Handle return va CX_REG.
10614 (output_indirect_thunk_function): Create alias for
10615 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10616 (ix86_output_indirect_jmp): Remove the bool argument.
10617 (ix86_output_indirect_function_return): New function.
10618 (ix86_split_simple_return_pop_internal): Likewise.
10619 * config/i386/i386.md (*indirect_jump): Don't pass false
10620 to ix86_output_indirect_jmp.
10621 (*tablejump_1): Likewise.
10622 (simple_return_pop_internal): Change it to define_insn_and_split.
10623 Call ix86_split_simple_return_pop_internal to split it for
10624 -mfunction-return=.
10625 (simple_return_indirect_internal): Call
10626 ix86_output_indirect_function_return instead of
10627 ix86_output_indirect_jmp.
10629 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10632 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10633 memset and value initialization afterwards.
10635 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10637 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10639 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10642 * common/config/aarch64/aarch64-common.c
10643 (aarch_option_optimization_table[]): Switch
10644 off fomit-frame-pointer
10646 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10647 Chung-Ju Wu <jasonwucj@gmail.com>
10649 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10651 (store_multiple): Ditto.
10653 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10655 * config.gcc: Add --with-cpu support for nds32 target.
10656 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10657 * config/nds32/nds32.opt: Add -mcpu= option.
10659 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10661 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10662 isel=yes): Warn for these deprecated options.
10664 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10666 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10667 ISA_2_5_MASKS_EMBEDDED.
10669 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10671 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10672 p->max as pointers rather than using iterative_hash_expr.
10674 2018-02-23 Carl Love <cel@us.ibm.com>
10676 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10677 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10679 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10680 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10681 P8V_BUILTIN_VEC_VUNSIGNED2.
10683 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10686 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10687 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10689 * lra-constraints.c (curr_insn_transform): Set up
10690 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10692 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10693 * lra-eliminations.c (spill_pseudos): Ditto.
10694 (process_insn_for_elimination): Ditto.
10695 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10696 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10697 LRA_NON_CLOBBERED_ALT.
10699 2018-02-22 Martin Sebor <msebor@redhat.com>
10701 PR tree-optimization/84480
10702 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10703 to maybe_diag_stxncpy_trunc. Call it.
10704 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10705 from gimple_fold_builtin_strcpy. Print inlining stack.
10706 (handle_builtin_stxncpy): Print inlining stack.
10707 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10709 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10712 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10713 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10714 and -fcheck-pointer-bounds are used together.
10715 (indirect_thunk_prefix): New enum.
10716 (indirect_thunk_need_prefix): New function.
10717 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10718 "_nt" instead of "_bnd" for NOTRACK prefix.
10719 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10720 (output_indirect_thunk_function): Likewise.
10722 (ix86_code_end): Update output_indirect_thunk_function calls.
10723 (ix86_output_indirect_branch_via_reg): Replace
10724 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10725 (ix86_output_indirect_branch_via_push): Likewise.
10726 (ix86_output_function_return): Likewise.
10727 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10728 incompatible with -fcf-protection=branch and
10729 -fcheck-pointer-bounds.
10731 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10734 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10735 Change gcc_assert call to output_operand_lossage.
10737 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10739 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10741 2018-02-22 DJ Delorie <dj@redhat.com>
10742 Sebastian Perta <sebastian.perta@renesas.com>
10743 Oleg Endo <olegendo@gcc.gnu.org>
10745 * config/rx/rx.c (rx_rtx_costs): New function.
10746 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10748 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10750 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10752 2018-02-22 Martin Liska <mliska@suse.cz>
10755 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10756 Add "native" as a possible value.
10758 2018-02-22 Martin Liska <mliska@suse.cz>
10761 * config/i386/i386.c (ix86_option_override_internal):
10762 Add "native" as a possible value for -march and -mtune.
10764 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10767 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10768 to all type variants.
10770 PR tree-optimization/84503
10771 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10772 width as info->bitpos + info->bitsize - start.
10773 (merged_store_group::merge_overlapping): Simplify width computation.
10774 (check_no_overlap): New function.
10775 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10776 start + width and last_order of the group, fail if check_no_overlap
10778 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10779 to group if check_no_overlap fails.
10781 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10783 * config/rs6000/altivec.md: Delete contraint arguments to
10784 define_expand, define_split, and define_peephole2, and in
10785 define_insn_and_split if always unused.
10786 * config/rs6000/darwin.md: Ditto.
10787 * config/rs6000/dfp.md: Ditto.
10788 * config/rs6000/rs6000.md: Ditto.
10789 * config/rs6000/sync.md: Ditto.
10790 * config/rs6000/vector.md: Ditto.
10791 * config/rs6000/vsx.md: Ditto.
10793 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10795 * config/rs6000/altivec.md: Write output control strings as braced
10796 blocks instead of double-quoted strings.
10797 * config/rs6000/darwin.md: Ditto.
10798 * config/rs6000/rs6000.md: Ditto.
10799 * config/rs6000/vector.md: Ditto.
10800 * config/rs6000/vsx.md: Ditto.
10802 2018-02-21 Jason Merrill <jason@redhat.com>
10804 PR c++/84314 - ICE with templates and fastcall attribute.
10805 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10807 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10809 * ipa-cp.c (determine_versionability): Fix comment typos.
10811 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10814 * ipa-cp.c (determine_versionability): Do not version functions caling
10817 2018-02-21 Martin Liska <mliska@suse.cz>
10820 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10821 Add "native" as a possible value.
10822 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10823 the macro when native cpu detection is available.
10825 2018-02-21 Martin Liska <mliska@suse.cz>
10828 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10829 Add "native" as a possible value.
10830 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10831 when native cpu detection is available.
10833 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10834 Martin Sebor <msebor@redhat.com>
10836 PR tree-optimization/84478
10837 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10839 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10840 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10841 support which is conservatively correct, for 2 only stay conservative
10842 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10843 argument to the 2 argument get_range_strlen, adjust 6 arg
10844 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10846 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10847 (gimple_fold_builtin_strlen): Pass true as last argument to
10850 2018-02-20 Martin Sebor <msebor@redhat.com>
10852 PR middle-end/84095
10853 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10854 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10855 (builtin_memref::builtin_memref): Factor out parts into
10856 set_base_and_offset and call it.
10858 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10860 PR middle-end/84406
10861 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10862 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10863 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10864 search at the associated MODE_INT.
10866 2018-02-20 Jeff Law <law@redhat.com>
10868 PR middle-end/82123
10869 PR tree-optimization/81592
10870 PR middle-end/79257
10871 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10872 for range data rather than using global data.
10873 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10874 range data rather than using global data.
10875 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10876 pass it to children as needed.
10877 (struct directive::fmtresult): Similarly.
10878 (struct directive::set_width): Similarly.
10879 (struct directive::set_precision): Similarly.
10880 (format_integer, format_directive, parse_directive): Similarly.
10881 (format_none): Accept unnamed vr_values parameter.
10882 (format_percent, format_floating, format_character): Similarly.
10883 (format_string, format_plain): Similarly.
10884 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10885 the EVRP range analyzer for range data rather than using global data.
10886 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10887 gimple-ssa-evrp-analyze.h
10888 (class sprintf_dom_walker): Add after_dom_children member function.
10889 Add evrp_range_analyzer member.
10890 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10891 range analyzer as needed.
10892 (sprintf_dom_walker::after_dom_children): New member function.
10893 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10895 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10896 (evrp_range_analyzer::pop_to_marker): Likewise.
10898 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10900 PR tree-optimization/84419
10901 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10902 with the required type if its current type is compatible but
10905 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10907 PR middle-end/82004
10908 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10909 after vectorization.
10911 2018-02-20 Martin Liska <mliska@suse.cz>
10914 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10915 possible values if we don't have a hint.
10917 2018-02-20 Martin Liska <mliska@suse.cz>
10921 * final.c (shorten_branches): Build align_tab array with one
10923 * opts.c (finish_options): Add alignment option limit check.
10924 (MAX_CODE_ALIGN): Likewise.
10925 (MAX_CODE_ALIGN_VALUE): Likewise.
10926 * doc/invoke.texi: Document maximum allowed option value for
10927 all -falign-* options.
10929 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10932 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10933 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10934 * var-tracking.c (emit_note_insn_var_location): Remove all references
10935 to NOTE_INSN_CALL_ARG_LOCATION.
10936 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10937 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10938 Use copy_rtx_if_shared.
10939 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10940 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10941 (dwarf2out_var_location): Remove handling of
10942 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10944 * final.c (final_scan_insn): Remove all references to
10945 NOTE_INSN_CALL_ARG_LOCATION.
10946 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10947 before dumping final insns.
10948 * except.c (emit_note_eh_region_end): Remove all references to
10949 NOTE_INSN_CALL_ARG_LOCATION.
10950 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10951 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10952 * config/arc/arc.c (hwloop_optimize): Likewise.
10953 * config/arm/arm.c (create_fix_barrier): Likewise.
10954 * config/s390/s390.c (s390_chunkify_start): Likewise.
10955 * config/sh/sh.c (find_barrier): Likewise.
10956 * config/i386/i386.c (rest_of_insert_endbranch,
10957 ix86_seh_fixup_eh_fallthru): Likewise.
10958 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10959 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10960 * config/frv/frv.c (frv_function_prologue): Likewise.
10961 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10963 (note_outside_basic_block_p): Remove all references to
10964 NOTE_INSN_CALL_ARG_LOCATION.
10965 * gengtype.c (adjust_field_rtx_def): Likewise.
10966 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10968 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10969 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10972 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10975 PR tree-optimization/84452
10976 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10977 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10980 2018-02-19 Martin Liska <mliska@suse.cz>
10983 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10985 2018-02-19 Martin Liska <mliska@suse.cz>
10986 Richard Sandiford <richard.sandiford@linaro.org>
10988 PR tree-optimization/82491
10989 * gimple-fold.c (get_base_constructor): Make earlier bail out
10992 2018-02-19 Carl Love <cel@us.ibm.com>
10994 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10995 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10997 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10998 P8V_BUILTIN_VEC_NEG.
11000 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11002 * config/rl78/rl78.md (movdf): New define expand.
11004 2018-02-19 Martin Liska <mliska@suse.cz>
11007 * doc/invoke.texi: Fix typo.
11008 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11010 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11012 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11013 handle rs6000_single_float and rs6000_double_float specially for
11016 2018-02-16 Jeff Law <law@redhat.com>
11018 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11019 (pop_regs): Use it.
11021 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11024 * ipa-inline.c (inline_small_functions): Fix a typo.
11026 2018-02-16 Nathan Sidwell <nathan@acm.org>
11028 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11030 2018-02-16 Carl Love <cel@us.ibm.com>
11032 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11033 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11034 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11035 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11036 expansion to P8V_BUILTIN_VEC_FLOAT2.
11038 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11040 PR rtl-optimization/70023
11041 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11042 src_regno into account.
11044 2018-02-16 Carl Love <cel@us.ibm.com>
11046 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11047 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11048 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11049 * config/rs6000/rs6000.c: Remove case statements for
11050 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11051 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11052 and P9V_BUILTIN_VEC_VINSERT4B.
11053 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11054 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11055 * config/rs6000/vsx.md:
11056 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11059 2018-02-16 Carl Love <cel@us.ibm.com>
11061 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11063 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11065 * config/rs6000/rs6000-c.c: Add the definitions for
11066 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11067 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11068 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11069 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11070 definition for insert4b and define insn *insert3b_internal.
11071 * doc/extend.texi: Add documentation for vec_extract4b.
11073 2018-02-16 Nathan Sidwell <nathan@acm.org>
11075 * doc/extend.texi (Backwards Compatibility): Mention friend
11076 injection. Note for-scope is deprecated.
11077 * doc/invoke.texi (-ffriend-injection): Deprecate.
11079 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11081 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11082 that moved to I2, also allow destinations that are a paradoxical
11083 subreg (instead of a normal reg).
11085 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11088 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11091 2018-02-16 Richard Biener <rguenther@suse.de>
11093 PR tree-optimization/84037
11094 PR tree-optimization/84016
11096 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11097 Adjust vec_construct for the fact we need additional higher latency
11098 128bit inserts for AVX256 and AVX512 vector builds.
11099 (ix86_add_stmt_cost): Scale vector construction cost for
11102 2018-02-16 Richard Biener <rguenther@suse.de>
11104 PR tree-optimization/84417
11105 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11106 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11107 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11109 2018-02-16 Martin Liska <mliska@suse.cz>
11112 * internal-fn.def (ASAN_CHECK): Set proper flags.
11113 (ASAN_MARK): Likewise.
11115 2018-02-16 Julia Koval <julia.koval@intel.com>
11117 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11118 from PTA_CANNONLAKE.
11120 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11123 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11124 Use ++iter rather than iter++ for std::list iterators.
11125 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11126 defer deleting them until all nodes in the forest are processed. Do
11127 free even leaf nodes. Change to_process into auto_vec.
11130 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11131 * vec.h (vec_default_construct): Use memset instead of placement new
11132 if BROKEN_VALUE_INITIALIZATION is defined.
11133 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11134 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11137 PR rtl-optimization/83723
11138 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11139 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11140 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11142 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11144 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11146 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11148 PR rtl-optimization/81443
11149 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11150 from inner REGs to paradoxical SUBREGs.
11152 2018-02-16 Richard Biener <rguenther@suse.de>
11154 PR tree-optimization/84399
11155 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11156 For operands we can analyze at their definition make sure we can
11157 analyze them at each use as well.
11159 2018-02-16 Richard Biener <rguenther@suse.de>
11161 PR tree-optimization/84190
11162 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11163 volatile accesses if the decl isn't volatile.
11165 2018-02-15 Jason Merrill <jason@redhat.com>
11167 PR c++/84314 - ICE with templates and fastcall attribute.
11168 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11169 TYPE_CANONICAL on an existing type.
11171 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11173 PR tree-optimization/84383
11174 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11175 dstoff nor call operand_equal_p if dstbase is NULL.
11177 PR tree-optimization/84334
11178 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11179 also a CONSTANT_CLASS_P, punt.
11181 2018-02-14 Jim Wilson <jimw@sifive.com>
11183 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11184 first SMALL_OPERAND check. New local min_second_step. Move assert
11185 to where locals are set. Add TARGET_RVC support.
11186 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11188 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11190 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11192 2018-02-14 Martin Sebor <msebor@redhat.com>
11194 PR tree-optimization/83698
11195 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11196 arrays constrain the offset range to their bounds.
11197 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11198 (builtin_access::overlap): Avoid setting the size of overlap if it's
11200 (maybe_diag_overlap): Also consider arrays when deciding what values
11201 of offsets to include in diagnostics.
11203 2018-02-14 Martin Sebor <msebor@redhat.com>
11206 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11207 that correspond to the kind of a declaration.
11209 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11212 * config/pa/pa.md: Load address of PIC label using the linkage table
11213 if the label is nonlocal.
11215 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11217 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11218 warning message if user requests -maltivec=be.
11219 * doc/invoke.texi: Document deprecation of -maltivec=be.
11221 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11224 * config/rs6000/rs6000-c.c: Update definitions for
11225 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11226 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11228 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11231 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11232 add _get_ssp intrinsics. Remove argument from
11233 __builtin_ia32_rdssp[d|q].
11234 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11235 * config/i386/i386-builtin.def: Remove argument from
11236 __builtin_ia32_rdssp[d|q].
11237 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11238 ix86_expand_special_args_builtin for _rdssp[d|q].
11239 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11240 Clear register before usage.
11241 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11242 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11244 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11246 PR tree-optimization/84357
11247 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11248 operand 1 of an ARRAY_REF too.
11250 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11253 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11254 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11256 (set_of_reg): New struct.
11257 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11258 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11259 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11261 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11262 Split into bitclr, bitset, bitinvert patterns if appropriate.
11263 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11264 use rx_fuse_in_memory_bitop.
11265 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11266 to named insn, correct maximum insn length.
11268 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11271 * machmode.def: Define a complex mode for PARTIAL_INT.
11272 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11274 * doc/rtl.texi: Document CSPImode.
11275 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11277 (msp430_hard_regno_nregs_with_padding): Likewise.
11279 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11282 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11284 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11286 PR rtl-optimization/84169
11287 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11288 we generated a parallel as new i3 and we split that to new i2 and i3
11289 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11290 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11291 those to i2, not i1. Partially rewrite this scan code.
11293 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11296 * stor-layout.c (place_field): For variable length fields, adjust
11297 offset_align afterwards not just based on the field's alignment,
11298 but also on the size.
11300 PR middle-end/84309
11301 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11302 of exps and logs in the use_exp2 case.
11304 2018-02-13 Jeff Law <law@redhat.com>
11306 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11307 entry for "vector".
11309 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11312 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11316 * common.opt (gas-loc-support, gas-locview-support): New.
11317 (ginline-points, ginternal-reset-location-views): New.
11318 * doc/invoke.texi: Document them. Use @itemx where intended.
11319 (gvariable-location-views): Adjust.
11320 * target.def (reset_location_view): New.
11321 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11322 (TARGET_RESET_LOCATION_VIEW): New.
11323 * doc/tm.texi: Rebuilt.
11324 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11325 (dwarf2out_default_as_locview_support): New.
11326 (output_asm_line_debug_info): Use option variables.
11327 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11328 (output_loc_list): Likewise.
11329 (add_high_low_attributes): Check option variables.
11330 Don't output entry view attribute in strict mode.
11331 (gen_inlined_subroutine_die): Check option variables.
11332 (dwarf2out_inline_entry): Likewise.
11333 (init_sections_and_labels): Likewise.
11334 (dwarf2out_early_finish): Likewise.
11335 (maybe_reset_location_view): New, from...
11336 (dwarf2out_var_location): ... here. Call it.
11337 * debug.h (dwarf2out_default_as_loc_support): Declare.
11338 (dwarf2out_default_as_locview_support): Declare.
11339 * hooks.c (hook_int_rtx_insn_0): New.
11340 * hooks.h (hook_int_rtx_insn_0): Declare.
11341 * toplev.c (process_options): Take -gas-loc-support and
11342 -gas-locview-support from dwarf2out. Enable
11343 -gvariable-location-views by default only with locview
11344 assembler support. Enable -ginternal-reset-location-views by
11345 default only if the target defines the corresponding hook.
11346 Enable -ginline-points by default if location views are
11347 enabled; force it disabled if statement frontiers are
11349 * tree-inline.c (expand_call_inline): Check option variables.
11350 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11352 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11354 PR tree-optimization/84321
11355 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11356 handling. Also check whether the anti-range contains any values
11357 that satisfy the mask; switch to a VR_RANGE if not.
11359 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11362 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11364 2018-02-13 Martin Jambor <mjambor@suse.cz>
11367 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11368 of call statements, also set location of a load to a temporary.
11370 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11372 * config/rl78/rl78.c (add_vector_labels): New function.
11373 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11374 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11375 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11376 which checks that no arguments are passed.
11377 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11378 * doc/extend.texi: Documentation for the new attribute.
11380 2018-02-13 Andreas Schwab <schwab@suse.de>
11382 * config/riscv/linux.h (CPP_SPEC): Define.
11384 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11387 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11388 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11389 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11390 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11391 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11392 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11393 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11394 temporarily for AES and PCLMUL builtins.
11396 PR tree-optimization/84339
11397 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11398 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11401 PR middle-end/84309
11402 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11403 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11404 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11406 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11408 * omp-simd-clone.h: New file.
11409 * omp-simd-clone.c: Include omp-simd-clone.h.
11410 (expand_simd_clones): No longer static.
11411 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11412 cgraph.h and omp-simd-clone.h.
11413 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11414 (vect_recog_widen_shift_pattern): Formatting fix.
11415 (vect_pattern_recog_1): Don't check optab for calls.
11418 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11419 operands[2] into a REG before using gen_lowpart on it.
11421 2018-02-12 Jeff Law <law@redhat.com>
11424 * config/sh/sh.c (find_barrier): Consider a sibling call
11427 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11428 successfully back substituting a reg.
11430 2018-02-12 Richard Biener <rguenther@suse.de>
11432 PR tree-optimization/84037
11433 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11434 parameter, move visited init to caller.
11435 (vect_slp_analyze_operations): Separate cost from validity
11436 check, initialize visited once for all instances.
11437 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11439 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11441 (vect_model_promotion_demotion_cost): Likewise.
11442 (vectorizable_bswap): Guard cost modeling with !slp_node
11443 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11445 (vectorizable_call): Likewise.
11446 (vectorizable_conversion): Likewise.
11447 (vectorizable_assignment): Likewise.
11448 (vectorizable_shift): Likewise.
11449 (vectorizable_operation): Likewise.
11450 (vectorizable_store): Likewise.
11451 (vectorizable_load): Likewise.
11452 (vectorizable_condition): Likewise.
11453 (vectorizable_comparison): Likewise.
11455 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11458 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11459 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11460 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11462 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11464 PR middle-end/83665
11465 * params.def (inline-min-speedup): Increase from 8 to 15.
11466 (max-inline-insns-auto): Decrease from 40 to 30.
11467 * ipa-split.c (consider_split): Add some buffer for function to
11468 be considered inlining candidate.
11469 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11472 2018-02-12 Richard Biener <rguenther@suse.de>
11474 PR tree-optimization/84037
11475 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11476 matched stmts if we cannot swap the non-matched ones.
11478 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11480 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11481 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11482 _mm_maskz_scalef_round_ss): New intrinsics.
11483 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11484 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11485 __builtin_ia32_scalefss_round): Remove.
11486 (__builtin_ia32_scalefsd_mask_round,
11487 __builtin_ia32_scalefss_mask_round): New intrinsics.
11488 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11489 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11490 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11491 "<round_constraint>")): Changed to ...
11492 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11493 "<round_scalar_constraint>")): ... this.
11494 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11495 %0, %1, %2<round_op3>}"): Changed to ...
11496 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11497 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11498 %2<round_scalar_mask_op3>}"): ... this.
11499 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11501 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11503 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11504 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11505 (_mm_maskz_sqrt_round_ss): New intrinsics.
11506 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11507 (__builtin_ia32_sqrtsd_mask_round)
11508 (__builtin_ia32_sqrtss_mask_round): New builtins.
11509 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11510 (__builtin_ia32_sqrtss_round): Remove.
11511 (__builtin_ia32_sqrtsd_mask_round)
11512 (__builtin_ia32_sqrtss_mask_round): New builtins.
11513 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11514 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11515 ((match_operand:VF_128 1 "vector_operand"
11516 "xBm,<round_constraint>")): Changed to ...
11517 ((match_operand:VF_128 1 "vector_operand"
11518 "xBm,<round_scalar_constraint>")): ... this.
11519 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11520 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11521 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11522 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11523 %<iptr>1<round_scalar_mask_op3>}): ... this.
11524 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11525 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11527 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11530 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11531 Cast vec_cmpeq result to correct type.
11532 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11533 Cast vec_cmpgt result to correct type.
11535 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11537 * final.c (final_scan_insn_1): Renamed from...
11538 (final_scan_insn): ... this. New wrapper, to recover
11539 seen from the outermost call in recursive ones.
11540 * config/sparc/sparc.c (output_return): Drop seen from call.
11541 (output_sibcall): Likewise.
11542 * config/visium/visium.c (output_branch): Likewise.
11544 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11546 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11549 2018-02-10 Alan Modra <amodra@gmail.com>
11552 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11553 Specify LR as an input.
11555 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11558 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11559 remove_member_access_dummy_vars): New functions.
11560 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11561 lower_omp_1, execute_lower_omp): Use them.
11563 PR rtl-optimization/84308
11564 * shrink-wrap.c (spread_components): Release todo vector.
11566 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11568 PR rtl-optimization/57193
11569 * ira-color.c (struct allocno_color_data): Add member
11570 conflict_allocno_hard_prefs.
11571 (update_conflict_allocno_hard_prefs): New.
11572 (bucket_allocno_compare_func): Add a preference based on
11573 conflict_allocno_hard_prefs.
11574 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11575 (color_allocnos): Remove a dead code. Initiate
11576 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11578 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11581 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11582 constraint from =wa to wa. Avoid a subreg on the output operand,
11583 instead use a pseudo and subreg it in a move.
11584 (p9_xxbrd_<mode>): Changed to ...
11585 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11586 (p9_xxbrd_v2df): New expander.
11587 (p9_xxbrw_<mode>): Changed to ...
11588 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11589 (p9_xxbrw_v4sf): New expander.
11591 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11593 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11595 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11598 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11599 multiply in 32-bit mode.
11600 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11601 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11604 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11606 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11607 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11608 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11609 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11611 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11614 * dwarf2out.c (is_trivial_indirect_ref): New function.
11615 (dwarf2out_late_global_decl): Do not generate a location
11616 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11617 and that are not defined in the current unit.
11619 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11621 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11622 instead of a libcall for UNORDERED.
11624 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11627 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11628 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11630 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11633 * config/s390/s390.c (s390_set_current_function): Invoke
11634 s390_indirect_branch_settings also if fndecl didn't change.
11636 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11638 * config/rs6000/rs6000.md (blockage): Set length to zero.
11640 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11642 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11644 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11647 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11648 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11652 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11653 PARALLEL incoming that failed vt_get_decl_and_offset check.
11655 PR middle-end/84237
11656 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11657 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11659 (get_variable_section): For decls in named .bss* sections pass true as
11660 second argument to bss_initializer_p.
11662 2018-02-09 Marek Polacek <polacek@redhat.com>
11663 Jakub Jelinek <jakub@redhat.com>
11666 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11667 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11668 Sync some changes from cxx_fold_indirect_ref.
11670 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11672 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11674 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11675 (BLOCK_INLINE_ENTRY_LABEL): New.
11676 (dwarf2out_var_location): Disregard inline entry markers.
11677 (inline_entry_data): New struct.
11678 (inline_entry_data_hasher): New hashtable type.
11679 (inline_entry_data_hasher::hash): New.
11680 (inline_entry_data_hasher::equal): New.
11681 (inline_entry_data_table): New variable.
11682 (add_high_low_attributes): Add DW_AT_entry_pc and
11683 DW_AT_GNU_entry_view attributes if a pending entry is found
11684 in inline_entry_data_table. Add old entry_pc attribute only
11685 if debug nonbinding markers are disabled.
11686 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11687 markers are enabled.
11688 (block_within_block_p, dwarf2out_inline_entry): New.
11689 (dwarf2out_finish): Check that no entries remained in
11690 inline_entry_data_table.
11691 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11692 (final_scan_insn, notice_source_line): Likewise.
11693 (rest_of_clean_state): Skip inline entry markers.
11694 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11696 * gimple.c (gimple_build_debug_inline_entry): New.
11697 * gimple.h (enum gimple_debug_subcode): Add
11698 GIMPLE_DEBUG_INLINE_ENTRY.
11699 (gimple_build_debug_inline_entry): Declare.
11700 (gimple_debug_inline_entry_p): New.
11701 (gimple_debug_nonbind_marker_p): Adjust.
11702 * insn-notes.def (INLINE_ENTRY): New.
11703 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11704 inline entry marker notes.
11705 (print_insn): Likewise.
11706 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11707 (INSN_DEBUG_MARKER_KIND): Likewise.
11708 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11709 * tree-inline.c (expand_call_inline): Build and insert
11710 debug_inline_entry stmt.
11711 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11712 inline entry blocks early, if nonbind markers are enabled.
11713 (dump_scope_block): Dump fragment info.
11714 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11715 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11716 (gimple_build_debug_inline_entry): New.
11717 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11718 Enable/disable inline entry points too.
11719 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11720 (DEBUG_INSN): Describe inline entry markers.
11722 * common.opt (gvariable-location-views): New.
11723 (gvariable-location-views=incompat5): New.
11724 * config.in: Rebuilt.
11725 * configure: Rebuilt.
11726 * configure.ac: Test assembler for view support.
11727 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11728 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11729 * dwarf2out.c (var_loc_view): New typedef.
11730 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11731 (dwarf2out_locviews_in_attribute): New.
11732 (dwarf2out_locviews_in_loclist): New.
11733 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11734 (enum dw_line_info_opcode): Add LI_adv_address.
11735 (struct dw_line_info_table): Add view.
11736 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11737 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11738 (zero_view_p): New variable.
11739 (ZERO_VIEW_P): New macro.
11740 (output_asm_line_debug_info): New.
11741 (struct var_loc_node): Add view.
11742 (add_AT_view_list, AT_loc_list): New.
11743 (add_var_loc_to_decl): Add view param. Test it against last.
11744 (new_loc_list): Add view params. Record them.
11745 (AT_loc_list_ptr): Handle loc and view lists.
11746 (view_list_to_loc_list_val_node): New.
11747 (print_dw_val): Handle dw_val_class_view_list.
11748 (size_of_die): Likewise.
11749 (value_format): Likewise.
11750 (loc_list_has_views): New.
11751 (gen_llsym): Set vl_symbol too.
11752 (maybe_gen_llsym, skip_loc_list_entry): New.
11753 (dwarf2out_maybe_output_loclist_view_pair): New.
11754 (output_loc_list): Output view list or entries too.
11755 (output_view_list_offset): New.
11756 (output_die): Handle dw_val_class_view_list.
11757 (output_dwarf_version): New.
11758 (output_compilation_unit_header): Use it.
11759 (output_skeleton_debug_sections): Likewise.
11760 (output_rnglists, output_line_info): Likewise.
11761 (output_pubnames, output_aranges): Update version comments.
11762 (output_one_line_info_table): Output view numbers in asm comments.
11763 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11764 Call maybe_gen_llsym.
11765 (loc_list_from_tree_1): Adjust.
11766 (add_AT_location_description): Create view list attribute if
11767 needed, check it's absent otherwise.
11768 (convert_cfa_to_fb_loc_list): Adjust.
11769 (maybe_emit_file): Call output_asm_line_debug_info for test.
11770 (dwarf2out_var_location): Reset views as needed. Precompute
11771 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11772 attribute. Set view.
11773 (new_line_info_table): Reset next view.
11774 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11775 (dwarf2out_source_line): Likewise. Output view resets and labels to
11776 the assembler, or select appropriate line info opcodes.
11777 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11778 (optimize_string_length): Catch it. Adjust.
11779 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11780 dw_val_class_view_list, and remove it if no longer needed.
11781 (hash_loc_list): Hash view numbers.
11782 (loc_list_hasher::equal): Compare them.
11783 (optimize_location_lists): Check whether a view list symbol is
11784 needed, and whether the locview attribute is present, and
11785 whether they match. Remove the locview attribute if no longer
11787 (index_location_lists): Call skip_loc_list_entry for test.
11788 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11789 Use output_dwarf_version.
11790 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11791 (struct dw_val_node): Add val_view_list.
11792 * final.c (SEEN_NEXT_VIEW): New.
11793 (set_next_view_needed): New.
11794 (clear_next_view_needed): New.
11795 (maybe_output_next_view): New.
11796 (final_start_function): Rename to...
11797 (final_start_function_1): ... this. Take pointer to FIRST,
11798 add SEEN parameter. Emit param bindings in the initial view.
11799 (final_start_function): Reintroduce SEEN-less interface.
11800 (final): Rename to...
11801 (final_1): ... this. Take SEEN parameter. Output final pending
11802 next view at the end.
11803 (final): Reintroduce seen-less interface.
11804 (final_scan_insn): Output pending next view before switching
11805 sections or ending a block. Mark the next view as needed when
11806 outputting variable locations. Notify debug backend of section
11807 changes, and of location view changes.
11808 (rest_of_handle_final): Adjust.
11809 * toplev.c (process_options): Autodetect value for debug variable
11810 location views option. Warn on incompat5 without -gdwarf-5.
11811 * doc/invoke.texi (gvariable-location-views): New.
11812 (gvariable-location-views=incompat5): New.
11813 (gno-variable-location-views): New.
11815 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11817 PR tree-optimization/84136
11818 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11819 that the result of find_edge is non-NULL.
11821 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11824 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11825 storing integer register in SImode. Fix cost of 256 and 512
11826 byte aligned SSE register store.
11828 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11830 * config/i386/i386.c (ix86_multiplication_cost): Fix
11831 multiplication cost for TARGET_AVX512DQ.
11833 2018-02-08 Marek Polacek <polacek@redhat.com>
11835 PR tree-optimization/84238
11836 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11839 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11841 PR tree-optimization/84265
11842 * tree-vect-stmts.c (vectorizable_store): Don't treat
11843 VMAT_CONTIGUOUS accesses as grouped.
11844 (vectorizable_load): Likewise.
11846 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11848 PR tree-optimization/81635
11849 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11850 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11851 (test_round_for_mask): New functions.
11852 (wide_int_cc_tests): Call test_round_for_mask.
11853 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11854 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11855 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11856 range returned by get_range_info.
11858 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11861 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11862 * symtab.c: Include builtins.h
11863 (symtab_node::output_to_lto_symbol_table_p): Move here
11864 from lto-streamer-out.c:output_symbol_p.
11865 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11866 (output_symbol_p): Move all logic to symtab.c
11867 (produce_symtab): Update.
11869 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11871 * config/s390/s390-opts.h (enum indirect_branch): Define.
11872 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11873 (s390_indirect_branch_via_thunk)
11874 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11875 (enum s390_indirect_branch_type): Define.
11876 * config/s390/s390.c (struct s390_frame_layout, struct
11877 machine_function): Remove.
11878 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11879 (indirect_branch_table_label_no, indirect_branch_table_name):
11881 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11882 (enum s390_indirect_branch_option): Define.
11883 (s390_return_addr_from_memory): New function.
11884 (s390_handle_string_attribute): New function.
11885 (s390_attribute_table): Add new attribute handler.
11886 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11887 (s390_indirect_branch_via_thunk): New function.
11888 (s390_indirect_branch_via_inline_thunk): New function.
11889 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11890 sibling call optimization for non z10 compiles.
11891 (s390_emit_call): Force indirect branch target to be a single
11892 register. Add r1 clobber for non-z10 compiles.
11893 (s390_emit_epilogue): Emit return jump via return_use expander.
11894 (s390_reorg): Handle JUMP_INSNs as execute targets.
11895 (s390_option_override_internal): Perform validity checks for the
11896 new command line options.
11897 (s390_indirect_branch_attrvalue): New function.
11898 (s390_indirect_branch_settings): New function.
11899 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11900 (s390_output_indirect_thunk_function): New function.
11901 (s390_code_end): Implement target hook.
11902 (s390_case_values_threshold): Implement target hook.
11903 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11905 * config/s390/s390.h (struct s390_frame_layout)
11906 (struct machine_function): Move here from s390.c.
11907 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11908 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11909 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11910 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11911 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11912 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11913 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11914 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11915 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11916 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11917 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11918 (mnemonic attribute): Add values which aren't recognized
11920 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11921 pattern for branch conversion. Fix mnemonic attribute.
11922 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11923 indirect branch via thunk if requested.
11924 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11925 ("*indirect_jump"): Disable for branch conversion using out of
11927 ("indirect_jump_via_thunk<mode>_z10")
11928 ("indirect_jump_via_thunk<mode>")
11929 ("indirect_jump_via_inlinethunk<mode>_z10")
11930 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11931 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11932 ("casesi_jump_via_inlinethunk<mode>_z10")
11933 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11934 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11935 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11936 ("*indirect2_jump"): Disable for branch conversion.
11937 ("casesi_jump"): Turn into expander and expand patterns for branch
11939 ("return_use"): New expander.
11940 ("*return"): Emit return via thunk and rename it to ...
11941 ("*return<mode>"): ... this one.
11942 * config/s390/s390.opt: Add new options and and enum for the
11945 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11947 * lra-constraints.c (match_reload): Unconditionally use
11948 gen_lowpart_SUBREG, rather than selecting between that
11949 and equivalent gen_rtx_SUBREG code.
11951 2018-02-08 Richard Biener <rguenther@suse.de>
11953 PR tree-optimization/84233
11954 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11955 changed flag instead of boguously re-using phi_inserted.
11957 2018-02-08 Martin Jambor <mjambor@suse.cz>
11959 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11960 static local variables.
11962 2018-02-08 Richard Biener <rguenther@suse.de>
11964 PR tree-optimization/84278
11965 * tree-vect-stmts.c (vectorizable_store): When looking for
11966 smaller vector types to perform grouped strided loads/stores
11967 make sure the mode is supported by the target.
11968 (vectorizable_load): Likewise.
11970 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11972 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11973 Increase LDP/STP opportunities by adding adjacent callee-saves.
11975 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11977 PR rtl-optimization/84068
11978 PR rtl-optimization/83459
11979 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11981 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11983 PR tree-optimization/84224
11984 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11985 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11986 non-zero arguments.
11988 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11991 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11992 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11993 count, remove one USE.
11995 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11997 * doc/install.texi (Configuration): Document the
11998 --with-long-double-format={ibm,ieee} PowerPC configuration
12002 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12003 Convert from define_expand to be define_insn_and_split. Rework
12004 float/double/_Float128 conversions to QI/HI/SImode to work with
12005 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12006 conversions to QI/HImode types did a store and then a load to
12007 truncate the value. For conversions to VSX registers, don't split
12008 the insn, instead emit the code directly. Use the code iterator
12009 any_fix to combine signed and unsigned conversions.
12010 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12011 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12012 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12013 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12014 (fix_<mode>di2_hw): Likewise.
12015 (fixuns_<mode>di2_hw): Likewise.
12016 (fix_<mode>si2_hw): Likewise.
12017 (fixuns_<mode>si2_hw): Likewise.
12018 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12019 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12020 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12021 fix<uns>_trunc<SFDF:mode>si2_p8.
12022 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12023 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12024 (fix<uns>_<mode>_mem): Likewise.
12025 (fctiw<u>z_<mode>_mem): Likewise.
12026 (fix<uns>_<mode>_mem): Likewise.
12027 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12028 the register allocator from doing a direct move to the GPRs to do
12029 a store, and instead use the ISA 3.0 store byte/half-word from
12030 vector register instruction. For IEEE 128-bit floating point,
12031 also optimize stores of 32-bit ints.
12032 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12034 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12036 * genextract.c (push_pathstr_operand): New function to support
12038 (walk_rtx): Call push_pathstr_operand.
12039 (print_path): Support [a-zA-Z].
12041 2018-02-07 Richard Biener <rguenther@suse.de>
12043 PR tree-optimization/84037
12044 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12045 (cse_and_gimplify_to_preheader): Declare.
12046 (vect_get_place_in_interleaving_chain): Likewise.
12047 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12049 (_loop_vec_info::~_loop_vec_info): Delete it.
12050 (cse_and_gimplify_to_preheader): New function.
12051 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12052 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12053 (vectorizable_load): Likewise. For grouped stores always base
12054 the IV on the first element.
12055 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12056 condition before gimplifying.
12058 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12060 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12061 *DIV_EXPR and *MOD_EXPR.
12063 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12066 * config/i386/i386.c (ix86_option_override_internal): Mask out
12067 the CF_SET bit when checking -fcf-protection.
12069 2018-02-07 Tom de Vries <tom@codesourcery.com>
12072 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12075 2018-02-07 Richard Biener <rguenther@suse.de>
12077 PR tree-optimization/84204
12078 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12081 PR tree-optimization/84205
12082 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12083 special-case isl_ast_op_zdiv_r.
12085 PR tree-optimization/84223
12086 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12087 Only add conditions from within the region.
12088 (gather_bbs::after_dom_children): Adjust.
12090 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12093 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12094 * config/avr/avr.md: Only post-reload split REG-REG moves if
12095 either register is GENERAL_REG_P.
12097 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12099 PR tree-optimization/84235
12100 * tree-ssa-scopedtables.c
12101 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12102 if the subtraction is performed in floating point type where NaNs are
12103 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12104 build 1. Formatting fix.
12106 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12109 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12110 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12111 and skip it regardless of bb boundaries. Use CALL_P macro,
12112 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12113 unnecessarily, formatting fix.
12115 2018-02-06 Michael Collison <michael.collison@arm.com>
12117 * config/arm/thumb2.md:
12118 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12119 (*thumb_mov_notscc): Ditto.
12121 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12124 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12125 unsigned_fix, not "s".
12127 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12129 * configure.ac (gcc_fn_eh_frame_ro): New function.
12130 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12131 correct .eh_frame permissions.
12132 * configure: Regenerate.
12134 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12136 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12137 irrelevant options.
12139 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12141 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12142 Display warning message for -mno-speculate-indirect-jumps.
12144 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12146 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12148 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12150 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12152 PR tree-optimization/84225
12153 * tree-eh.c (find_trapping_overflow): Only call
12154 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12156 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12159 * config/i386/i386.c: Reimplement the check of possible options
12160 -mibt/-mshstk conbination. Change error messages.
12161 * doc/invoke.texi: Fix a typo: remove extra '='.
12163 2018-02-06 Marek Polacek <polacek@redhat.com>
12165 PR tree-optimization/84228
12166 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12168 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12171 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12172 emitted arch directives.
12173 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12174 __ARM_FEATURE_COPROC before changing architectures.
12176 2018-02-06 Richard Biener <rguenther@suse.de>
12178 * config/i386/i386.c (print_reg): Fix typo.
12179 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12181 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12183 * configure: Regenerate.
12185 2018-02-05 Martin Sebor <msebor@redhat.com>
12187 PR tree-optimization/83369
12188 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12191 2018-02-05 Martin Liska <mliska@suse.cz>
12193 * doc/invoke.texi: Cherry-pick upstream r323995.
12195 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12197 * ira.c (ira_init_register_move_cost): Adjust comment.
12199 2018-02-05 Martin Liska <mliska@suse.cz>
12201 PR gcov-profile/84137
12202 * doc/gcov.texi: Fix typo in documentation.
12204 2018-02-05 Martin Liska <mliska@suse.cz>
12206 PR gcov-profile/83879
12207 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12208 using dlopen functionality.
12210 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12212 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12213 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12214 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12215 _mm_maskz_range_round_ss): New intrinsics.
12216 (__builtin_ia32_rangesd128_round)
12217 (__builtin_ia32_rangess128_round): Remove.
12218 (__builtin_ia32_rangesd128_mask_round,
12219 __builtin_ia32_rangess128_mask_round): New builtins.
12220 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12221 __builtin_ia32_rangess128_round): Remove.
12222 (__builtin_ia32_rangesd128_mask_round,
12223 __builtin_ia32_rangess128_mask_round): New builtins.
12224 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12225 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12226 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12227 "<round_saeonly_constraint>")): Changed to ...
12228 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12229 "<round_saeonly_scalar_constraint>")): ... this.
12230 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12231 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12232 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12233 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12234 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12236 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12238 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12240 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12241 Remove all values except native, 8540 and 8548.
12243 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12245 * config/i386/i386.c (ix86_output_function_return): Pass
12246 INVALID_REGNUM, instead of -1, as invalid register number to
12247 indirect_thunk_name and output_indirect_thunk.
12249 2018-02-02 Julia Koval <julia.koval@intel.com>
12251 * config.gcc: Add -march=icelake.
12252 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12253 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12254 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12255 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12256 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12257 (processor_target_table): Add icelake.
12258 (ix86_option_override_internal): Handle new PTAs.
12259 (get_builtin_code_for_version): Handle icelake.
12260 (M_INTEL_COREI7_ICELAKE): New.
12261 (fold_builtin_cpu): Handle icelake.
12262 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12263 * doc/invoke.texi: Add -march=icelake.
12265 2018-02-02 Julia Koval <julia.koval@intel.com>
12267 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12268 to wide_int_bitmask.
12269 * wide-int-bitmask.h: New.
12271 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12274 * config/i386/i386.md: Replace Pmode with word_mode in
12275 builtin_setjmp_setup and builtin_longjmp to support x32.
12277 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12281 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12283 (rs6000_supported_cpu_names): New static variable.
12284 (linux_cpu_translation_table): Likewise.
12285 (elf_platform) <cpu>: Define new static variable and use it.
12286 Translate kernel AT_PLATFORM name to canonical name if needed.
12287 Error if platform name is unknown.
12289 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12292 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12294 2018-02-01 Jeff Law <law@redhat.com>
12297 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12298 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12299 the scratch if RELEASE_VIA_POP is false.
12300 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12301 If we have to save a temporary register, decrement SIZE appropriately.
12302 Pass new arguments to release_scratch_register_on_entry.
12303 (ix86_adjust_stack_and_probe): Likewise.
12304 (ix86_emit_probe_stack_range): Pass new arguments to
12305 release_scratch_register_on_entry.
12307 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12309 PR rtl-optimization/84157
12310 * combine.c (change_zero_ext): Use REG_P predicate in
12311 front of HARD_REGISTER_P predicate.
12313 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12315 * config/avr/avr.c (avr_option_override): Move disabling of
12316 -fdelete-null-pointer-checks to...
12317 * common/config/avr/avr-common.c (avr_option_optimization_table):
12320 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12322 PR tree-optimization/81635
12323 * tree-data-ref.c (split_constant_offset_1): For types that
12324 wrap on overflow, try to use range info to prove that wrapping
12327 2018-02-01 Renlin Li <renlin.li@arm.com>
12330 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12331 TAILCALL_ADDR_REGS.
12332 (aarch64_register_move_cost): Likewise.
12333 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12334 TAILCALL_ADDR_REGS.
12335 (REG_CLASS_NAMES): Likewise.
12336 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12337 TAILCALL_ADDR_REGS. Remove IP registers.
12338 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12340 2018-02-01 Richard Biener <rguenther@suse.de>
12342 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12343 for specifying RPO order and allow NULL for that.
12344 * domwalk.c (dom_walker::dom_walker): Likewise.
12345 (dom_walker::walk): Handle NULL RPO order.
12346 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12348 (rewrite_update_dom_walker): Likewise.
12349 (mark_def_dom_walker): Likewise.
12351 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12353 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12354 (aarch64_maybe_expand_sve_subreg_move): Declare.
12355 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12356 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12358 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12359 that are semantically a reverse operation.
12360 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12361 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12362 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12364 (aarch64_can_change_mode_class): For big-endian, forbid changes
12365 between two SVE modes if they have different element sizes.
12367 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12369 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12370 the TImode handling for big-endian targets.
12372 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12374 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12375 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12377 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12378 Remove BSWAP handing for big-endian targets and use the form of
12379 LD1RQ appropariate for the mode.
12381 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12383 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12384 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12385 duplicated element.
12387 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12390 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12391 check for operands that need to go through aarch64_sve_reload_be.
12393 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12395 PR tree-optimization/81661
12396 PR tree-optimization/84117
12397 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12398 * tree-eh.c: Include gimplify.h.
12399 (find_trapping_overflow, replace_trapping_overflow,
12400 rewrite_to_non_trapping_overflow): New functions.
12401 * tree-vect-loop.c: Include tree-eh.h.
12402 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12403 * tree-data-ref.c: Include tree-eh.h.
12404 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12406 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12408 PR rtl-optimization/84123
12409 * combine.c (change_zero_ext): Check if hard register satisfies
12410 can_change_dest_mode before calling gen_lowpart_SUBREG.
12412 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12415 * ira.c (ira_init_register_move_cost): Remove assert.
12417 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12419 PR rtl-optimization/84071
12420 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12421 * doc/tm.texi: Regenerate.
12423 2018-01-31 Richard Biener <rguenther@suse.de>
12425 PR tree-optimization/84132
12426 * tree-data-ref.c (analyze_miv_subscript): Properly
12427 check whether evolution_function_is_affine_multivariate_p
12428 before calling gcd_of_steps_may_divide_p.
12430 2018-01-31 Julia Koval <julia.koval@intel.com>
12433 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12434 * config/i386/i386.md (rdpid_rex64) New.
12435 (rdpid): Make 32bit only.
12437 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12440 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12441 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12443 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12446 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12448 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12450 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12452 PR rtl-optimization/84071
12453 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12454 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12456 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12458 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12459 (arc_attribute_table): Add 'aux' attribute.
12460 (arc_in_small_data_p): Consider aux like variables.
12461 (arc_is_aux_reg_p): New function.
12462 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12463 (arc_get_aux_arg): New function.
12464 (prepare_move_operands): Handle aux-register access.
12465 (arc_handle_aux_attribute): New function.
12466 * doc/extend.texi (ARC Variable attributes): Add subsection.
12468 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12470 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12471 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12472 (arc_attribute_table): Add 'uncached' attribute.
12473 (arc_print_operand): Print '.di' flag for uncached memory
12475 (arc_in_small_data_p): Do not consider for small data the uncached
12477 (arc_is_uncached_mem_p): New function.
12478 * config/arc/predicates.md (compact_store_memory_operand): Check
12479 for uncached memory accesses.
12480 (nonvol_nonimm_operand): Likewise.
12481 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12483 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12486 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12487 falign-loops=): Add Optimization flag.
12489 2018-01-30 Jeff Law <law@redhat.com>
12492 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12493 INT_REGISTERS_SAVED. Check it prior to calling
12494 get_scratch_register_on_entry.
12495 (ix86_adjust_stack_and_probe): Similarly.
12496 (ix86_emit_probe_stack_range): Similarly.
12497 (ix86_expand_prologue): Corresponding changes.
12499 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12502 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12503 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12505 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12508 * lra-constraints.c (curr_insn_transform): Process AND in the
12511 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12513 PR rtl-optimization/83986
12514 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12515 dependence against last_pending_memory_flush in addition to
12516 pending_jump_insns.
12518 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12520 PR tree-optimization/81611
12521 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12524 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12527 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12530 2018-01-30 Richard Biener <rguenther@suse.de>
12531 Jakub Jelinek <jakub@redhat.com>
12533 PR tree-optimization/84111
12534 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12535 inner loops added during recursion, as they don't have up-to-date
12538 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12541 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12542 (can_inline_edge_by_limits_p): ... here.
12543 (can_early_inline_edge_p, check_callers,
12544 update_caller_keys, update_callee_keys, recursive_inlining,
12545 add_new_edges_to_heap, speculation_useful_p,
12546 inline_small_functions,
12547 inline_small_functions, flatten_function,
12548 inline_to_all_callers_1): Update.
12550 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12552 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12555 2018-01-30 Richard Biener <rguenther@suse.de>
12557 PR tree-optimization/83008
12558 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12559 invariant and constant vector uses in stmts when they need
12560 more than one stmt.
12562 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12565 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12566 * configure: Regenerate.
12568 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12570 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12572 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12573 Use gen_rtx_REG rather than gen_lowpart.
12575 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12577 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12578 rather than 0 when creating partial subregs.
12580 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12582 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12585 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12588 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12589 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12590 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12591 flags. This restores the settings used before the 2017-07-24.
12592 Turning off pre increment/decrement/modify allows IVOPTS to
12593 optimize DF/SF loops where the index is an int.
12595 2018-01-29 Richard Biener <rguenther@suse.de>
12596 Kelvin Nilsen <kelvin@gcc.gnu.org>
12599 * tree-vect-stmts.c (vectorizable_call): Don't call
12600 targetm.vectorize_builtin_md_vectorized_function if callee is
12603 2018-01-22 Carl Love <cel@us.ibm.com>
12605 * doc/extend.tex: Fix typo in second arg in
12606 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12608 2018-01-29 Richard Biener <rguenther@suse.de>
12610 PR tree-optimization/84086
12611 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12612 (flush_ssaname_freelist): When SSA names were released reset
12613 the SCEV hash table.
12615 2018-01-29 Richard Biener <rguenther@suse.de>
12617 PR tree-optimization/84057
12618 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12619 removed paths when removing edges.
12621 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12623 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12624 -mfunction-return=@var{choice}.
12626 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12628 PR diagnostic/84034
12629 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12630 Handle CR like TAB.
12631 (layout::print_source_line): Likewise.
12632 (test_get_line_width_without_trailing_whitespace): Add test cases.
12634 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12636 PR middle-end/84040
12637 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12640 2018-01-26 Jim Wilson <jimw@sifive.com>
12642 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12644 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12647 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12649 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12650 and CMP + SUB-immediate -> SUBS.
12652 2018-01-26 Martin Sebor <msebor@redhat.com>
12654 PR tree-optimization/83896
12655 * tree-ssa-strlen.c (get_string_len): Rename...
12656 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12657 Avoid assuming length is constant.
12658 (handle_char_store): Use HOST_WIDE_INT for string length.
12660 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12663 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12664 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12666 2018-01-26 Richard Biener <rguenther@suse.de>
12668 PR rtl-optimization/84003
12669 * dse.c (record_store): Only record redundant stores when
12670 the earlier store aliases at least all accesses the later one does.
12672 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12674 PR rtl-optimization/83985
12675 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12676 REG_CFA_RESTORE insns.
12677 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12678 REG_CFA_RESTORE insns here.
12681 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12682 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12684 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12686 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12687 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12688 (arc_init): Likewise.
12689 (arc_override_options): Likewise.
12690 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12692 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12694 * config/arc/arc.h (TARGET_DBNZ): Define.
12695 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12696 properly set the tune attribute.
12697 (dbnz): Use TARGET_DBNZ guard.
12698 * config/arc/arc.opt (mtune): Add core3 option.
12700 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12702 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12703 recognize new pic like addresses.
12704 (arc_delegitimize_address): Clean up.
12706 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12708 * config/arc/arc-arches.def: Option mrf16 valid for all
12710 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12711 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12712 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12713 * config/arc/arc-tables.opt: Regenerate.
12714 * config/arc/arc.c (arc_conditional_register_usage): Handle
12715 reduced register file case.
12716 (arc_file_start): Set must have build attributes.
12717 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12718 mrf16 option value.
12719 * config/arc/arc.opt (mrf16): Add new option.
12720 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12721 * config/arc/genmultilib.awk: Handle new mrf16 option.
12722 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12723 * config/arc/t-multilib: Regenerate.
12724 * doc/invoke.texi (ARC Options): Document mrf16 option.
12726 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12728 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12729 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12730 (arc_attribute_table): Add 'secure_call' attribute.
12731 (arc_print_operand): Print secure call operand.
12732 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12734 (arc_is_secure_call_p): New function. * config/arc/arc.md
12735 (call_i): Add support for sjli instruction.
12736 (call_value_i): Likewise.
12737 * config/arc/constraints.md (Csc): New constraint.
12739 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12740 John Eric Martin <John.Martin@emmicro-us.com>
12742 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12743 * config/arc/arc.c (_arc_jli_section): New struct.
12744 (arc_jli_section): New type.
12745 (rc_jli_sections): New static variable.
12746 (arc_handle_jli_attribute): New function.
12747 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12748 (arc_file_end): New function.
12749 (TARGET_ASM_FILE_END): Define.
12750 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12751 (arc_add_jli_section): New function.
12752 (jli_call_scan): Likewise.
12753 (arc_reorg): Call jli_call_scan.
12754 (arc_output_addsi): Remove 'S' from printing asm operand.
12755 (arc_is_jli_call_p): New function.
12756 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12758 (movhi_insn): Likewise.
12759 (movsi_insn): Likewise.
12760 (movsi_set_cc_insn): Likewise.
12761 (loadqi_update): Likewise.
12762 (load_zeroextendqisi_update): Likewise.
12763 (load_signextendqisi_update): Likewise.
12764 (loadhi_update): Likewise.
12765 (load_zeroextendhisi_update): Likewise.
12766 (load_signextendhisi_update): Likewise.
12767 (loadsi_update): Likewise.
12768 (loadsf_update): Likewise.
12769 (movsicc_insn): Likewise.
12770 (bset_insn): Likewise.
12771 (bxor_insn): Likewise.
12772 (bclr_insn): Likewise.
12773 (bmsk_insn): Likewise.
12774 (bicsi3_insn): Likewise.
12775 (cmpsi_cc_c_insn): Likewise.
12776 (movsi_ne): Likewise.
12777 (movsi_cond_exec): Likewise.
12778 (clrsbsi2): Likewise.
12779 (norm_f): Likewise.
12789 (call_i): Remove 'S' asm letter, add jli instruction.
12790 (call_value_i): Likewise.
12791 * config/arc/arc.op (mjli-always): New option.
12792 * config/arc/constraints.md (Cji): New constraint.
12793 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12795 (subsf3_fpx): Likewise.
12796 (mulsf3_fpx): Likewise.
12797 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12799 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12800 function attrbutes.
12801 * doc/invoke.texi (ARC): Document mjli-always option.
12803 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12805 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12806 avoid addition with 0 and use incw and decw where possible.
12808 2018-01-26 Richard Biener <rguenther@suse.de>
12810 PR tree-optimization/81082
12811 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12812 association if it requires casting to unsigned.
12813 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12814 from fold_plusminus_mult_expr to catch important cases late when
12815 range info is available.
12817 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12819 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12820 * configure.ac (hidden_linkonce): New test.
12821 * configure: Regenerate.
12822 * config.in: Regenerate.
12824 2018-01-26 Julia Koval <julia.koval@intel.com>
12826 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12827 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12828 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12829 _mm_mask_bitshuffle_epi64_mask): Fix type.
12830 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12831 USI_FTYPE_V4DI_V4DI_USI): Remove.
12832 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12833 __builtin_ia32_vpshufbitqmb256_mask,
12834 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12835 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12836 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12838 2018-01-26 Alan Modra <amodra@gmail.com>
12841 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12842 UNSPEC_VBPERMQ. Sort other unspecs.
12844 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12846 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12848 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12850 PR middle-end/83055
12851 * predict.c (drop_profile): Do not push/pop cfun; update also
12853 (handle_missing_profiles): Fix logic looking for zero profiles.
12855 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12857 PR middle-end/83977
12858 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12859 on functions with #pragma omp declare simd or functions with simd
12861 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12862 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12863 Remove trailing \n from warning_at calls.
12865 2018-01-25 Tom de Vries <tom@codesourcery.com>
12868 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12869 for neutered workers.
12871 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12874 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12875 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12877 2018-01-24 Jeff Law <law@redhat.com>
12880 * i386.c (get_probe_interval): Move to earlier point.
12881 (ix86_compute_frame_layout): If -fstack-clash-protection and
12882 the frame is larger than the probe interval, then use pushes
12883 to save registers rather than reg->mem moves.
12884 (ix86_expand_prologue): Remove conditional for int_registers_saved
12887 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12890 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12891 min/max for never referenced object.
12893 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12895 PR middle-end/83977
12896 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12898 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12899 attributes from DECL_ATTRIBUTES (decl) without affecting
12900 DECL_ATTRIBUTES (current_function_decl).
12901 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12902 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12904 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12906 PR tree-optimization/83979
12907 * fold-const.c (fold_comparison): Use constant_boolean_node
12908 instead of boolean_{true,false}_node.
12910 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12912 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12915 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12917 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12918 Simplify the clause that sets the length attribute.
12919 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12920 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12921 clause that sets the length attribute.
12922 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12924 2018-01-24 Tom de Vries <tom@codesourcery.com>
12927 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12928 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12929 Add strict parameter.
12930 (prevent_branch_around_nothing): Insert dummy insn between branch to
12931 label and label with no ptx insn inbetween.
12932 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12934 2018-01-24 Tom de Vries <tom@codesourcery.com>
12937 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12938 for neutered threads in warp.
12939 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12941 2018-01-24 Richard Biener <rguenther@suse.de>
12943 PR tree-optimization/83176
12944 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12947 2018-01-24 Richard Biener <rguenther@suse.de>
12949 PR tree-optimization/82819
12950 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12951 code generating pluses that are no-ops in the target precision.
12953 2018-01-24 Richard Biener <rguenther@suse.de>
12955 PR middle-end/84000
12956 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12958 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12960 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12961 to merge probabilities.
12962 * predict.c (probably_never_executed): Also mark as cold functions
12963 with global 0 profile and guessed local profile.
12964 * profile-count.c (profile_probability::combine_with_count): New
12966 * profile-count.h (profile_probability::operator*,
12967 profile_probability::operator*=, profile_probability::operator/,
12968 profile_probability::operator/=): Reduce precision to adjusted
12969 and set value to guessed on contradictory divisions.
12970 (profile_probability::combine_with_freq): Remove.
12971 (profile_probability::combine_wiht_count): Declare.
12972 (profile_count::force_nonzero):: Set to adjusted.
12973 (profile_count::probability_in):: Set quality to adjusted.
12974 * tree-ssa-tail-merge.c (replace_block_by): Use
12975 combine_with_count.
12977 2018-01-23 Andrew Waterman <andrew@sifive.com>
12978 Jim Wilson <jimw@sifive.com>
12980 * config/riscv/riscv.c (riscv_stack_boundary): New.
12981 (riscv_option_override): Set riscv_stack_boundary. Handle
12982 riscv_preferred_stack_boundary_arg.
12983 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12984 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12985 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12986 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12987 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12988 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12990 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12993 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12994 of struct ix86_frame.
12995 (ix86_expand_epilogue): Likewise. Add a local variable for
12996 the reg_save_offset field in struct ix86_frame.
12998 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13000 PR tree-optimization/82604
13001 * tree-loop-distribution.c (enum partition_kind): New enum item
13002 PKIND_PARTIAL_MEMSET.
13003 (partition_builtin_p): Support above new enum item.
13004 (generate_code_for_partition): Ditto.
13005 (compute_access_range): Differentiate cases that equality can be
13006 proven at all loops, the innermost loops or no loops.
13007 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13008 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13009 (finalize_partitions, distribute_loop): Don't fuse partition of
13010 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13011 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13012 parloop is enabled.
13014 2018-01-23 Martin Liska <mliska@suse.cz>
13016 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13017 order to ignore the predictor.
13018 (PRED_POLYMORPHIC_CALL): Likewise.
13019 (PRED_RECURSIVE_CALL): Likewise.
13021 2018-01-23 Martin Liska <mliska@suse.cz>
13023 * tree-profile.c (tree_profiling): Print function header to
13024 aware reader which function we are working on.
13025 * value-prof.c (gimple_find_values_to_profile): Do not print
13026 not interesting value histograms.
13028 2018-01-23 Martin Liska <mliska@suse.cz>
13030 * profile-count.h (enum profile_quality): Add
13031 profile_uninitialized as the first value. Do not number values
13032 as they are zero based.
13033 (profile_count::verify): Update sanity check.
13034 (profile_probability::verify): Likewise.
13036 2018-01-23 Nathan Sidwell <nathan@acm.org>
13038 * doc/invoke.texi (ffor-scope): Deprecate.
13040 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13042 PR tree-optimization/83510
13043 * domwalk.c (set_all_edges_as_executable): New function.
13044 (dom_walker::dom_walker): Convert bool param
13045 "skip_unreachable_blocks" to enum reachability. Move setup of
13046 edge flags to set_all_edges_as_executable and only do it when
13047 reachability is REACHABLE_BLOCKS.
13048 * domwalk.h (enum dom_walker::reachability): New enum.
13049 (dom_walker::dom_walker): Convert bool param
13050 "skip_unreachable_blocks" to enum reachability.
13051 (set_all_edges_as_executable): New decl.
13052 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13053 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13055 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13056 but converting true to REACHABLE_BLOCKS.
13057 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13059 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13060 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13061 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13063 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13064 if check_all_array_refs will be called.
13066 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13068 * tree.c (selftest::test_location_wrappers): Add more test
13071 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13073 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13074 (selftest::test_bit_in_range): Likewise.
13076 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13079 * doc/sourcebuild.texi (vect_float): Say that the selector
13080 only describes the situation when -funsafe-math-optimizations is on.
13081 (vect_float_strict): Document.
13083 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13085 PR tree-optimization/83965
13086 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13087 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13088 instead of checking only for a reduction.
13089 (vect_recog_widen_sum_pattern): Likewise.
13091 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13093 * predict.c (probably_never_executed): Only use precise profile info.
13094 (compute_function_frequency): Skip after inlining hack since we now
13095 have quality checking.
13097 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13099 * profile-count.h (profile_probability::very_unlikely,
13100 profile_probability::unlikely, profile_probability::even): Set
13101 precision to guessed.
13103 2018-01-23 Richard Biener <rguenther@suse.de>
13105 PR tree-optimization/83963
13106 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13107 Properly terminate dominator walk when crossing the exit edge not
13108 when visiting its source block.
13110 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13113 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13114 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13116 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13118 PR tree-optimization/83957
13119 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13120 semicolon after for body surrounded by braces.
13122 PR tree-optimization/83081
13123 * profile-count.h (profile_probability::split): New method.
13124 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13125 Use profile_probability::split.
13126 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13127 when splitting a single conditional jump into 2.
13129 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13131 PR tree-optimization/69452
13132 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13135 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13137 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13138 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13139 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13141 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13143 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13145 * config/rl78/rl78.md (movdi): New define_expand.
13146 * config/rl78/rl78.c (rl78_split_movdi): New function.
13148 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13151 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13153 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13154 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13155 128-bit to produce an UNSPEC move to get the double word with the
13156 signbit and then a shift directly to do signbit.
13157 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13158 implementation with a new version that just does either a direct
13159 move or a regular move. Move memory interface to separate insns.
13160 Move insns so they are next to the expander.
13161 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13162 with signbit move. Split big and little endian case.
13163 (signbit<mode>2_dm_mem_le): Likewise.
13164 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13165 (signbit<mode>2_dm2): Likewise.
13167 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13169 * config/rl78/rl78.md (anddi3): New define_expand.
13171 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13173 * config/rl78/rl78.md (umindi3): New define_expand.
13175 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13177 * config/rl78/rl78.md (smindi3): New define_expand.
13179 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13181 * config/rl78/rl78.md (smaxdi3): New define_expand.
13183 2018-01-22 Carl Love <cel@us.ibm.com>
13185 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13186 LVX_V1TI): Add macro expansion.
13187 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13188 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13189 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13190 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13191 Change check to determine if the instruction is a byte reversing
13192 entry. Fix typo in comment.
13193 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13194 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13195 Add def_builtin calls for new builtins.
13196 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13197 Add define_insn expansion.
13199 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13201 * config/rl78/rl78.md (umaxdi3): New define_expand.
13203 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13205 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13206 for non-QImode registers.
13208 2018-01-22 Richard Biener <rguenther@suse.de>
13210 PR tree-optimization/83963
13211 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13212 including the loop exit block.
13213 (scop_detection::merge_sese): Likewise.
13214 (scop_detection::add_scop): Do it here instead.
13216 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13218 * doc/sourcebuild.texi (arm_softfloat): Document.
13220 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13223 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13224 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13225 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13227 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13228 David Edelsohn <dje.gcc@gmail.com>
13231 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13232 Change "crset eq" to "crset 2".
13233 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13234 (*call_indirect_aix<mode>_nospec): Likewise.
13235 (*call_value_indirect_aix<mode>_nospec): Likewise.
13236 (*call_indirect_elfv2<mode>_nospec): Likewise.
13237 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13238 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13239 change assembly output from . to $.
13240 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13241 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13242 (*tablejump<mode>_internal1_nospec): Likewise.
13244 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13247 * config/sh/sh_optimize_sett_clrt.cc:
13248 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13250 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13252 PR tree-optimization/83940
13253 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13254 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13255 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13256 use it to pass back the definition type.
13257 (vect_check_store_rhs): Likewise rhs_dt_out.
13258 (vect_build_gather_load_calls): Add a mask_dt argument and use
13259 it instead of a call to vect_is_simple_use.
13260 (vectorizable_store): Update calls to vect_check_load_store_mask
13261 and vect_check_store_rhs. Use the dt returned by the latter instead
13262 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13263 instead of calls to vect_is_simple_use. Pass the scalar rather
13264 than the vector operand to vect_is_simple_use when handling
13265 second and subsequent copies of an rhs value.
13266 (vectorizable_load): Update calls to vect_check_load_store_mask
13267 and vect_build_gather_load_calls. Use the cached mask_dt and
13268 gs_info.offset_dt instead of calls to vect_is_simple_use.
13270 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13272 PR middle-end/83945
13273 * tree-emutls.c: Include gimplify.h.
13274 (lower_emutls_2): New function.
13275 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13276 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13277 it before further processing.
13280 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13281 UINTVAL (trueop1) instead of INTVAL (op1).
13283 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13287 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13288 INCOMING_FRAME_SP_OFFSET if not defined.
13289 (scan_trace): Add ENTRY argument. If true and
13290 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13291 emit a note to adjust the CFA offset.
13292 (create_cfi_notes): Adjust scan_trace callers.
13293 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13294 INCOMING_FRAME_SP_OFFSET in the CIE.
13295 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13296 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13298 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13299 * doc/tm.texi: Regenerated.
13301 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13303 PR rtl-optimization/83147
13304 * lra-constraints.c (remove_inheritance_pseudos): Use
13305 lra_substitute_pseudo_within_insn.
13307 2018-01-19 Tom de Vries <tom@codesourcery.com>
13308 Cesar Philippidis <cesar@codesourcery.com>
13311 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13313 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13316 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13317 spaces for function labels.
13319 2018-01-19 Martin Liska <mliska@suse.cz>
13321 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13322 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13323 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13324 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13325 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13326 (PRED_CONST_RETURN): Change from 69 to 65.
13327 (PRED_NULL_RETURN): Change from 91 to 71.
13328 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13329 (PRED_LOOP_GUARD): Change from 66 to 73.
13331 2018-01-19 Martin Liska <mliska@suse.cz>
13333 * predict.c (predict_insn_def): Add new assert.
13334 (struct branch_predictor): Change type to signed integer.
13335 (test_prediction_value_range): Amend test to cover
13336 PROB_UNINITIALIZED.
13337 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13338 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13339 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13340 (PRED_LOOP_IV_COMPARE): Likewise.
13341 * predict.h (PROB_UNINITIALIZED): Define new constant.
13343 2018-01-19 Martin Liska <mliska@suse.cz>
13345 * predict.c (dump_prediction): Add new format for
13346 analyze_brprob.py script which is enabled with -details
13348 * profile-count.h (precise_p): New function.
13350 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13352 PR tree-optimization/83922
13353 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13354 there are no statements that need masking.
13355 (vect_active_double_reduction_p): New function.
13356 (vect_analyze_loop_operations): Use it when handling phis that
13357 are not in the loop header.
13359 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13361 PR tree-optimization/83914
13362 * tree-vect-loop.c (vectorizable_induction): Don't convert
13363 init_expr or apply the peeling adjustment for inductions
13364 that are nested within the vectorized loop.
13366 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13368 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13371 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13375 * function.h (gimplify_parameters): Add gimple_seq * argument.
13376 * function.c: Include gimple.h and options.h.
13377 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13378 for the added local temporaries if needed.
13379 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13380 if there are any parameter cleanups, wrap whole body into a
13381 try/finally with the cleanups.
13383 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13386 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13387 Use GET_MODE_CLASS for scalar floating point.
13389 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13393 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13394 Fix call of call_cgraph_insertion_hooks.
13396 2018-01-18 Martin Sebor <msebor@redhat.com>
13398 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13400 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13403 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13406 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13409 * common.opt: (-ffile-prefix-map): New option.
13410 * opts.c (common_handle_option): Defer it.
13411 * opts-global.c (handle_common_deferred_options): Handle it.
13412 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13413 * file-prefix-map.h: New file.
13414 (remap_debug_filename, add_debug_prefix_map): ...here.
13415 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13416 * final.c (debug_prefix_map, add_debug_prefix_map
13417 remap_debug_filename): Move to...
13418 * file-prefix-map.c: New file.
13419 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13420 generalize, get rid of alloca(), use strrchr() instead of strchr().
13421 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13422 Implement in terms of add_prefix_map().
13423 (remap_macro_filename, remap_debug_filename): Implement in term of
13425 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13426 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13427 * dbxout.c: Include file-prefix-map.h.
13428 * varasm.c: Likewise.
13429 * vmsdbgout.c: Likewise.
13430 * xcoffout.c: Likewise.
13431 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13432 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13433 * doc/invoke.texi (-ffile-prefix-map): Document.
13434 (-fdebug-prefix-map): Update description.
13436 2018-01-18 Martin Liska <mliska@suse.cz>
13438 * config/i386/i386.c (indirect_thunk_name): Document that also
13440 (output_indirect_thunk): Document why both instructions
13441 (pause and lfence) are generated.
13443 2018-01-18 Richard Biener <rguenther@suse.de>
13445 PR tree-optimization/83887
13446 * graphite-scop-detection.c
13447 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13448 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13449 (scop_detection::merge_sese): Re-implement with a flood-fill
13450 algorithm that properly finds a SESE region if it exists.
13452 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13455 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13456 pointer_diff optimizations use view_convert instead of convert.
13458 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13460 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13461 Generate different code for -mno-speculate-indirect-jumps.
13462 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13463 (*call_indirect_aix<mode>): Disable for
13464 -mno-speculate-indirect-jumps.
13465 (*call_indirect_aix<mode>_nospec): New define_insn.
13466 (*call_value_indirect_aix<mode>): Disable for
13467 -mno-speculate-indirect-jumps.
13468 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13469 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13470 -mno-speculate-indirect-jumps.
13471 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13473 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13475 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13476 long double type, set the flags for noting the default long double
13477 type, even if we don't pass or return a long double type.
13479 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13482 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13485 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13487 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13488 support for merge[hl].
13489 (fold_mergehl_helper): New helper function.
13490 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13491 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13492 (altivec_vmrglw_direct): Add xxmrglw insn.
13494 2018-01-17 Andrew Waterman <andrew@sifive.com>
13496 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13497 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13499 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13502 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13503 call the lto_location_cache before reading the
13504 DECL_SOURCE_LOCATION of the types.
13506 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13507 Richard Sandiford <richard.sandiford@linaro.org>
13509 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13510 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13511 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13512 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13513 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13515 * config/aarch64/constraints.md (aarch64_movti_operand):
13517 * config/aarch64/predicates.md (Uti): Add new constraint.
13519 2018-01-17 Carl Love <cel@us.ibm.com>
13521 * config/rs6000/vsx.md (define_expand xl_len_r,
13522 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13523 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13525 (define_expand, define_insn): Move the shift left from the
13526 define_insn to the define_expand for lxvl and stxvl instructions.
13527 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13528 and XL_LEN_R definitions to PURE.
13530 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13532 * config/i386/i386.c (indirect_thunk_name): Declare regno
13533 as unsigned int. Compare regno with INVALID_REGNUM.
13534 (output_indirect_thunk): Ditto.
13535 (output_indirect_thunk_function): Ditto.
13536 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13537 in the call to output_indirect_thunk_function.
13539 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13541 PR middle-end/83884
13542 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13543 rather than the size of inner_type to determine the stack slot size
13544 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13546 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13549 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13552 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13554 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13555 endian Linux systems to optionally enable multilibs for selecting
13556 the long double type if the user configured an explicit type.
13557 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13558 have no long double multilibs if not defined.
13559 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13560 warn if the user used -mabi={ieee,ibm}longdouble and we built
13561 multilibs for long double.
13562 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13563 appropriate multilib option.
13564 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13566 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13567 for building long double multilibs.
13568 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13570 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13572 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13575 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13577 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13580 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13583 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13586 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13588 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13589 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13591 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13593 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13594 different rtl trees depending on TARGET_64BIT.
13595 (rs6000_gen_lvx): Likewise.
13597 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13599 * config/visium/visium.md (nop): Tweak comment.
13600 (hazard_nop): Likewise.
13602 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13604 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13605 -mspeculate-indirect-jumps.
13606 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13607 for -mno-speculate-indirect-jumps.
13608 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13609 (*call_value_indirect_elfv2<mode>): Disable for
13610 -mno-speculate-indirect-jumps.
13611 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13612 (indirect_jump): Emit different RTL for
13613 -mno-speculate-indirect-jumps.
13614 (*indirect_jump<mode>): Disable for
13615 -mno-speculate-indirect-jumps.
13616 (*indirect_jump<mode>_nospec): New define_insn.
13617 (tablejump): Emit different RTL for
13618 -mno-speculate-indirect-jumps.
13619 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13620 (tablejumpsi_nospec): New define_expand.
13621 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13622 (tablejumpdi_nospec): New define_expand.
13623 (*tablejump<mode>_internal1): Disable for
13624 -mno-speculate-indirect-jumps.
13625 (*tablejump<mode>_internal1_nospec): New define_insn.
13626 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13629 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13631 * caller-save.c (insert_save): Drop unnecessary parameter. All
13634 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13635 Richard Biener <rguenth@suse.de>
13638 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13639 return early, inline manually is_gimple_sizepos. Make sure if we
13640 call gimplify_expr we don't end up with a gimple constant.
13641 * tree.c (variably_modified_type_p): Don't return true for
13642 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13643 * gimplify.h (is_gimple_sizepos): Remove.
13645 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13647 PR tree-optimization/83857
13648 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13649 vectorizable_live_operation for pure SLP statements.
13650 (vectorizable_live_operation): Handle PHIs.
13652 2018-01-16 Richard Biener <rguenther@suse.de>
13654 PR tree-optimization/83867
13655 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13656 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13658 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13661 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13662 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13663 If off is not INTEGER_CST, issue a may not be aligned warning
13664 rather than isn't aligned. Use isn%'t rather than isn't.
13665 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13667 <case MULT_EXPR>: Improve the case when bottom and one of the
13668 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13669 operand, in that case check if the other operand is multiple of
13670 bottom divided by the INTEGER_CST operand.
13672 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13675 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13676 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13677 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13678 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13679 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13680 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13681 (pa_function_arg_size): New function.
13683 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13685 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13686 in a separate statement.
13688 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13690 PR tree-optimization/83847
13691 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13692 group gathers and scatters.
13694 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13696 PR rtl-optimization/86620
13697 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13699 PR rtl-optimization/83213
13700 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13701 to last if both are JUMP_INSNs.
13703 PR tree-optimization/83843
13704 * gimple-ssa-store-merging.c
13705 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13706 store_immediate_info for bswap/nop orig_stores.
13708 2018-01-15 Andrew Waterman <andrew@sifive.com>
13710 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13712 <UDIV>: Increase cost if !TARGET_DIV.
13714 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13716 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13717 (define_attr "cr_logical_3op"): New.
13718 (cceq_ior_compare): Adjust.
13719 (cceq_ior_compare_complement): Adjust.
13720 (*cceq_rev_compare): Adjust.
13721 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13722 (is_cracked_insn): Adjust.
13723 (insn_must_be_first_in_group): Adjust.
13724 * config/rs6000/40x.md: Adjust.
13725 * config/rs6000/440.md: Adjust.
13726 * config/rs6000/476.md: Adjust.
13727 * config/rs6000/601.md: Adjust.
13728 * config/rs6000/603.md: Adjust.
13729 * config/rs6000/6xx.md: Adjust.
13730 * config/rs6000/7450.md: Adjust.
13731 * config/rs6000/7xx.md: Adjust.
13732 * config/rs6000/8540.md: Adjust.
13733 * config/rs6000/cell.md: Adjust.
13734 * config/rs6000/e300c2c3.md: Adjust.
13735 * config/rs6000/e500mc.md: Adjust.
13736 * config/rs6000/e500mc64.md: Adjust.
13737 * config/rs6000/e5500.md: Adjust.
13738 * config/rs6000/e6500.md: Adjust.
13739 * config/rs6000/mpc.md: Adjust.
13740 * config/rs6000/power4.md: Adjust.
13741 * config/rs6000/power5.md: Adjust.
13742 * config/rs6000/power6.md: Adjust.
13743 * config/rs6000/power7.md: Adjust.
13744 * config/rs6000/power8.md: Adjust.
13745 * config/rs6000/power9.md: Adjust.
13746 * config/rs6000/rs64.md: Adjust.
13747 * config/rs6000/titan.md: Adjust.
13749 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13751 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13752 ix86_indirect_branch_register logic.
13754 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13756 * config/i386/constraints.md (Bs): Update
13757 ix86_indirect_branch_register check. Don't check
13758 ix86_indirect_branch_register with GOT_memory_operand.
13760 * config/i386/predicates.md (GOT_memory_operand): Don't check
13761 ix86_indirect_branch_register here.
13762 (GOT32_symbol_operand): Likewise.
13764 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13766 * config/i386/predicates.md (constant_call_address_operand):
13767 Rewrite ix86_indirect_branch_register logic.
13768 (sibcall_insn_operand): Likewise.
13770 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13772 * config/i386/constraints.md (Bs): Replace
13773 ix86_indirect_branch_thunk_register with
13774 ix86_indirect_branch_register.
13776 * config/i386/i386.md (indirect_jump): Likewise.
13777 (tablejump): Likewise.
13778 (*sibcall_memory): Likewise.
13779 (*sibcall_value_memory): Likewise.
13780 Peepholes of indirect call and jump via memory: Likewise.
13781 * config/i386/i386.opt: Likewise.
13782 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13783 (GOT_memory_operand): Likewise.
13784 (call_insn_operand): Likewise.
13785 (sibcall_insn_operand): Likewise.
13786 (GOT32_symbol_operand): Likewise.
13788 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13790 PR middle-end/83837
13791 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13792 type rather than type addr's type points to.
13793 (expand_omp_atomic_mutex): Likewise.
13794 (expand_omp_atomic): Likewise.
13796 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13799 * config/i386/i386.c (output_indirect_thunk_function): Use
13800 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13801 for __x86_return_thunk.
13803 2018-01-15 Richard Biener <rguenther@suse.de>
13805 PR middle-end/83850
13806 * expmed.c (extract_bit_field_1): Fix typo.
13808 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13811 * config/arm/iterators.md (VF): New mode iterator.
13812 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13813 Remove integer-related logic from pattern.
13814 (neon_vabd<mode>_3): Likewise.
13816 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13818 PR middle-end/82694
13819 * common.opt (fstrict-overflow): No longer an alias.
13820 (fwrapv-pointer): New option.
13821 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13822 also for pointer types based on flag_wrapv_pointer.
13823 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13824 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13825 opts->x_flag_wrapv got set.
13826 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13827 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13828 POINTER_TYPE_OVERFLOW_UNDEFINED.
13829 * match.pd: Likewise in address comparison pattern.
13830 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13832 2018-01-15 Richard Biener <rguenther@suse.de>
13835 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13836 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13837 Reset type names to their identifier if their TYPE_DECL doesn't
13838 have linkage (and thus is used for ODR and devirt).
13839 (save_debug_info_for_decl): Remove.
13840 (save_debug_info_for_type): Likewise.
13841 (add_tree_to_fld_list): Adjust.
13842 * tree-pretty-print.c (dump_generic_node): Make dumping of
13843 type names more robust.
13845 2018-01-15 Richard Biener <rguenther@suse.de>
13847 * BASE-VER: Bump to 8.0.1.
13849 2018-01-14 Martin Sebor <msebor@redhat.com>
13852 * builtins.c (check_access): Avoid warning when the no-warning bit
13855 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13857 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13858 * ira-color (allocno_hard_regs_compare): Likewise.
13860 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13863 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13864 Use .pushsection/.popsection.
13866 2018-01-14 Martin Sebor <msebor@redhat.com>
13869 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13871 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13873 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13874 entry from extra_headers.
13875 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13876 extra_headers, make the list bitwise identical to the i?86-*-* one.
13878 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13880 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13881 -mcmodel=large with -mindirect-branch=thunk,
13882 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13883 -mfunction-return=thunk-extern.
13884 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13885 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13886 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13888 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13890 * config/i386/i386.c (print_reg): Print the name of the full
13891 integer register without '%'.
13892 (ix86_print_operand): Handle 'V'.
13893 * doc/extend.texi: Document 'V' modifier.
13895 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13897 * config/i386/constraints.md (Bs): Disallow memory operand for
13898 -mindirect-branch-register.
13900 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13901 (GOT_memory_operand): Likewise.
13902 (call_insn_operand): Likewise.
13903 (sibcall_insn_operand): Likewise.
13904 (GOT32_symbol_operand): Likewise.
13905 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13906 for -mindirect-branch-register.
13907 (tablejump): Likewise.
13908 (*sibcall_memory): Likewise.
13909 (*sibcall_value_memory): Likewise.
13910 Disallow peepholes of indirect call and jump via memory for
13911 -mindirect-branch-register.
13912 (*call_pop): Replace m with Bw.
13913 (*call_value_pop): Likewise.
13914 (*sibcall_pop_memory): Replace m with Bs.
13915 * config/i386/i386.opt (mindirect-branch-register): New option.
13916 * doc/invoke.texi: Document -mindirect-branch-register option.
13918 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13920 * config/i386/i386-protos.h (ix86_output_function_return): New.
13921 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13922 set function_return_type.
13923 (indirect_thunk_name): Add ret_p to indicate thunk for function
13925 (output_indirect_thunk_function): Pass false to
13926 indirect_thunk_name.
13927 (ix86_output_indirect_branch_via_reg): Likewise.
13928 (ix86_output_indirect_branch_via_push): Likewise.
13929 (output_indirect_thunk_function): Create alias for function
13930 return thunk if regno < 0.
13931 (ix86_output_function_return): New function.
13932 (ix86_handle_fndecl_attribute): Handle function_return.
13933 (ix86_attribute_table): Add function_return.
13934 * config/i386/i386.h (machine_function): Add
13935 function_return_type.
13936 * config/i386/i386.md (simple_return_internal): Use
13937 ix86_output_function_return.
13938 (simple_return_internal_long): Likewise.
13939 * config/i386/i386.opt (mfunction-return=): New option.
13940 (indirect_branch): Mention -mfunction-return=.
13941 * doc/extend.texi: Document function_return function attribute.
13942 * doc/invoke.texi: Document -mfunction-return= option.
13944 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13946 * config/i386/i386-opts.h (indirect_branch): New.
13947 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13948 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13949 with local indirect jump when converting indirect call and jump.
13950 (ix86_set_indirect_branch_type): New.
13951 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13952 (indirectlabelno): New.
13953 (indirect_thunk_needed): Likewise.
13954 (indirect_thunk_bnd_needed): Likewise.
13955 (indirect_thunks_used): Likewise.
13956 (indirect_thunks_bnd_used): Likewise.
13957 (INDIRECT_LABEL): Likewise.
13958 (indirect_thunk_name): Likewise.
13959 (output_indirect_thunk): Likewise.
13960 (output_indirect_thunk_function): Likewise.
13961 (ix86_output_indirect_branch_via_reg): Likewise.
13962 (ix86_output_indirect_branch_via_push): Likewise.
13963 (ix86_output_indirect_branch): Likewise.
13964 (ix86_output_indirect_jmp): Likewise.
13965 (ix86_code_end): Call output_indirect_thunk_function if needed.
13966 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13968 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13969 (ix86_attribute_table): Add indirect_branch.
13970 * config/i386/i386.h (machine_function): Add indirect_branch_type
13971 and has_local_indirect_jump.
13972 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13974 (tablejump): Likewise.
13975 (*indirect_jump): Use ix86_output_indirect_jmp.
13976 (*tablejump_1): Likewise.
13977 (simple_return_indirect_internal): Likewise.
13978 * config/i386/i386.opt (mindirect-branch=): New option.
13979 (indirect_branch): New.
13982 (thunk-inline): Likewise.
13983 (thunk-extern): Likewise.
13984 * doc/extend.texi: Document indirect_branch function attribute.
13985 * doc/invoke.texi: Document -mindirect-branch= option.
13987 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13990 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13992 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13994 * ipa-inline.c (want_inline_small_function_p): Return false if
13995 inlining has already failed with CIF_FINAL_ERROR.
13996 (update_caller_keys): Call want_inline_small_function_p before
13998 (update_callee_keys): Likewise.
14000 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14002 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14004 (rs6000_quadword_masked_address_p): Likewise.
14005 (quad_aligned_load_p): Likewise.
14006 (quad_aligned_store_p): Likewise.
14007 (const_load_sequence_p): Add comment to describe the outer-most loop.
14008 (mimic_memory_attributes_and_flags): New function.
14009 (rs6000_gen_stvx): Likewise.
14010 (replace_swapped_aligned_store): Likewise.
14011 (rs6000_gen_lvx): Likewise.
14012 (replace_swapped_aligned_load): Likewise.
14013 (replace_swapped_load_constant): Capitalize argument name in
14014 comment describing this function.
14015 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14016 and stores that access quad-word aligned addresses and replace
14017 with stvx or lvx instructions when appropriate.
14018 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14019 New function prototype.
14020 (rs6000_quadword_masked_address_p): Likewise.
14021 (rs6000_gen_lvx): Likewise.
14022 (rs6000_gen_stvx): Likewise.
14023 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14024 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14025 when memory address is aligned.
14026 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14027 this split to select lvx instruction when memory address is aligned.
14028 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14029 instruction when memory address is aligned.
14030 (*vsx_le_perm_load_v16qi): Likewise.
14031 (four unnamed splitters): Modify to select the stvx instruction
14032 when memory is aligned.
14034 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14036 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14037 which appears in the queue multiple times.
14039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14040 Alan Hayward <alan.hayward@arm.com>
14041 David Sherwood <david.sherwood@arm.com>
14043 * tree-vectorizer.h (vec_lower_bound): New structure.
14044 (_loop_vec_info): Add check_nonzero and lower_bounds.
14045 (LOOP_VINFO_CHECK_NONZERO): New macro.
14046 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14047 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14048 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14049 fields. Make seg_len the distance travelled, not including the
14051 (dr_direction_indicator): Declare.
14052 (dr_zero_step_indicator): Likewise.
14053 (dr_known_forward_stride_p): Likewise.
14054 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14056 (runtime_alias_check_p): Allow runtime alias checks with
14058 (operator ==): Compare access_size and align.
14059 (prune_runtime_alias_test_list): Rework for new distinction between
14060 the access_size and seg_len.
14061 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14063 (get_segment_min_max): New function.
14064 (create_intersect_range_checks): Use it.
14065 (dr_step_indicator): New function.
14066 (dr_direction_indicator): Likewise.
14067 (dr_zero_step_indicator): Likewise.
14068 (dr_known_forward_stride_p): Likewise.
14069 * tree-loop-distribution.c (data_ref_segment_size): Return
14070 DR_STEP * (niters - 1).
14071 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14073 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14074 (vect_preserves_scalar_order_p): New function, split out from...
14075 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14076 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14077 (vect_vfa_access_size): New function.
14078 (vect_vfa_align): Likewise.
14079 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14080 (dump_lower_bound): New function.
14081 (vect_check_lower_bound): Likewise.
14082 (vect_small_gap_p): Likewise.
14083 (vectorizable_with_step_bound_p): Likewise.
14084 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14085 depencies if the vectorization factor is 1. Convert the checks
14086 for nonzero steps into checks on the bounds of DR_STEP. Try using
14087 a bunds check for variable steps if the minimum required step is
14088 relatively small. Update calls to the dr_with_seg_len
14089 constructor and to vect_compile_time_alias.
14090 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14092 (vect_loop_versioning): Call it.
14093 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14095 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14098 Alan Hayward <alan.hayward@arm.com>
14099 David Sherwood <david.sherwood@arm.com>
14101 * doc/sourcebuild.texi (vect_scatter_store): Document.
14102 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14104 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14106 * genopinit.c (main): Add supports_vec_scatter_store and
14107 supports_vec_scatter_store_cached to target_optabs.
14108 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14109 IFN_MASK_SCATTER_STORE.
14110 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14112 * internal-fn.h (internal_store_fn_p): Declare.
14113 (internal_fn_stored_value_index): Likewise.
14114 * internal-fn.c (scatter_store_direct): New macro.
14115 (expand_scatter_store_optab_fn): New function.
14116 (direct_scatter_store_optab_supported_p): New macro.
14117 (internal_store_fn_p): New function.
14118 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14119 IFN_MASK_SCATTER_STORE.
14120 (internal_fn_mask_index): Likewise.
14121 (internal_fn_stored_value_index): New function.
14122 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14123 for scatter stores.
14124 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14125 * optabs-query.c (supports_vec_scatter_store_p): New function.
14126 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14127 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14128 true for scatter stores.
14129 (vect_gather_scatter_fn_p): Handle scatter stores too.
14130 (vect_check_gather_scatter): Consider using scatter stores if
14131 supports_vec_scatter_store_p.
14132 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14133 scatter stores too.
14134 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14135 internal_fn_stored_value_index.
14136 (check_load_store_masking): Handle scatter stores too.
14137 (vect_get_store_rhs): Make public.
14138 (vectorizable_call): Use internal_store_fn_p.
14139 (vectorizable_store): Handle scatter store internal functions.
14140 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14141 when deciding whether the end of the group has been reached.
14142 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14143 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14144 (mask_scatter_store<mode>): New insns.
14146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14147 Alan Hayward <alan.hayward@arm.com>
14148 David Sherwood <david.sherwood@arm.com>
14150 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14151 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14152 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14154 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14155 Use vect_truncate_gather_scatter_offset if we can't treat the
14156 operation as a normal gather load or scatter store.
14157 (get_group_load_store_type): Take the gather_scatter_info
14158 as argument. Try using a gather load or scatter store for
14159 single-element groups.
14160 (get_load_store_type): Update calls to get_group_load_store_type
14161 and vect_use_strided_gather_scatters_p.
14163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14164 Alan Hayward <alan.hayward@arm.com>
14165 David Sherwood <david.sherwood@arm.com>
14167 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14168 optional tree argument.
14169 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14171 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14172 but continue to use the current value as a fallback.
14173 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14174 to compare the updates.
14175 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14176 (get_load_store_type): Use it when handling a strided access.
14177 (vect_get_strided_load_store_ops): New function.
14178 (vect_get_data_ptr_increment): Likewise.
14179 (vectorizable_load): Handle strided gather loads. Always pass
14180 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14183 Alan Hayward <alan.hayward@arm.com>
14184 David Sherwood <david.sherwood@arm.com>
14186 * doc/md.texi (gather_load@var{m}): Document.
14187 (mask_gather_load@var{m}): Likewise.
14188 * genopinit.c (main): Add supports_vec_gather_load and
14189 supports_vec_gather_load_cached to target_optabs.
14190 * optabs-tree.c (init_tree_optimization_optabs): Use
14191 ggc_cleared_alloc to allocate target_optabs.
14192 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14193 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14195 * internal-fn.h (internal_load_fn_p): Declare.
14196 (internal_gather_scatter_fn_p): Likewise.
14197 (internal_fn_mask_index): Likewise.
14198 (internal_gather_scatter_fn_supported_p): Likewise.
14199 * internal-fn.c (gather_load_direct): New macro.
14200 (expand_gather_load_optab_fn): New function.
14201 (direct_gather_load_optab_supported_p): New macro.
14202 (direct_internal_fn_optab): New function.
14203 (internal_load_fn_p): Likewise.
14204 (internal_gather_scatter_fn_p): Likewise.
14205 (internal_fn_mask_index): Likewise.
14206 (internal_gather_scatter_fn_supported_p): Likewise.
14207 * optabs-query.c (supports_at_least_one_mode_p): New function.
14208 (supports_vec_gather_load_p): Likewise.
14209 * optabs-query.h (supports_vec_gather_load_p): Declare.
14210 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14211 and memory_type field.
14212 (NUM_PATTERNS): Bump to 15.
14213 * tree-vect-data-refs.c: Include internal-fn.h.
14214 (vect_gather_scatter_fn_p): New function.
14215 (vect_describe_gather_scatter_call): Likewise.
14216 (vect_check_gather_scatter): Try using internal functions for
14217 gather loads. Recognize existing calls to a gather load function.
14218 (vect_analyze_data_refs): Consider using gather loads if
14219 supports_vec_gather_load_p.
14220 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14221 (vect_get_gather_scatter_offset_type): Likewise.
14222 (vect_convert_mask_for_vectype): Likewise.
14223 (vect_add_conversion_to_patterm): Likewise.
14224 (vect_try_gather_scatter_pattern): Likewise.
14225 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14226 (vect_vect_recog_func_ptrs): Add it.
14227 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14228 internal_fn_mask_index and internal_gather_scatter_fn_p.
14229 (check_load_store_masking): Take the gather_scatter_info as an
14230 argument and handle gather loads.
14231 (vect_get_gather_scatter_ops): New function.
14232 (vectorizable_call): Check internal_load_fn_p.
14233 (vectorizable_load): Likewise. Handle gather load internal
14235 (vectorizable_store): Update call to check_load_store_masking.
14236 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14237 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14238 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14239 (aarch64_gather_scale_operand_d): New predicates.
14240 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14241 (mask_gather_load<mode>): New insns.
14243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14244 Alan Hayward <alan.hayward@arm.com>
14245 David Sherwood <david.sherwood@arm.com>
14247 * optabs.def (fold_left_plus_optab): New optab.
14248 * doc/md.texi (fold_left_plus_@var{m}): Document.
14249 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14250 * internal-fn.c (fold_left_direct): Define.
14251 (expand_fold_left_optab_fn): Likewise.
14252 (direct_fold_left_optab_supported_p): Likewise.
14253 * fold-const-call.c (fold_const_fold_left): New function.
14254 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14255 * tree-parloops.c (valid_reduction_p): New function.
14256 (gather_scalar_reductions): Use it.
14257 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14258 (vect_finish_replace_stmt): Declare.
14259 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14260 (needs_fold_left_reduction_p): New function, split out from...
14261 (vect_is_simple_reduction): ...here. Accept reductions that
14262 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14263 (vect_force_simple_reduction): Also store the reduction type in
14264 the assignment's STMT_VINFO_REDUC_TYPE.
14265 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14266 (merge_with_identity): New function.
14267 (vect_expand_fold_left): Likewise.
14268 (vectorize_fold_left_reduction): Likewise.
14269 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14270 scalar phi in place for it. Check for target support and reject
14271 cases that would reassociate the operation. Defer the transform
14272 phase to vectorize_fold_left_reduction.
14273 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14274 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14275 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14279 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14280 call to ifc_temp_var.
14282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14283 Alan Hayward <alan.hayward@arm.com>
14284 David Sherwood <david.sherwood@arm.com>
14286 * target.def (legitimize_address_displacement): Take the original
14287 offset as a poly_int.
14288 * targhooks.h (default_legitimize_address_displacement): Update
14290 * targhooks.c (default_legitimize_address_displacement): Likewise.
14291 * doc/tm.texi: Regenerate.
14292 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14293 as an argument, moving assert of ad->disp == ad->disp_term to...
14294 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14295 Try calling targetm.legitimize_address_displacement before expanding
14296 the address rather than afterwards, and adjust for the new interface.
14297 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14298 Match the new hook interface. Handle SVE addresses.
14299 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14300 new hook interface.
14302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14304 * Makefile.in (OBJS): Add early-remat.o.
14305 * target.def (select_early_remat_modes): New hook.
14306 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14307 * doc/tm.texi: Regenerate.
14308 * targhooks.h (default_select_early_remat_modes): Declare.
14309 * targhooks.c (default_select_early_remat_modes): New function.
14310 * timevar.def (TV_EARLY_REMAT): New timevar.
14311 * passes.def (pass_early_remat): New pass.
14312 * tree-pass.h (make_pass_early_remat): Declare.
14313 * early-remat.c: New file.
14314 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14316 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14319 Alan Hayward <alan.hayward@arm.com>
14320 David Sherwood <david.sherwood@arm.com>
14322 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14323 vfm1 with a bound_epilog parameter.
14324 (vect_do_peeling): Update calls accordingly, and move the prologue
14325 call earlier in the function. Treat the base bound_epilog as 0 for
14326 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14327 this base when peeling for gaps.
14328 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14329 with fully-masked loops.
14330 (vect_estimate_min_profitable_iters): Handle the single peeled
14331 iteration in that case.
14333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14334 Alan Hayward <alan.hayward@arm.com>
14335 David Sherwood <david.sherwood@arm.com>
14337 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14338 single-element interleaving even if the size is not a power of 2.
14339 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14340 accesses for single-element interleaving if the group size is
14343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14344 Alan Hayward <alan.hayward@arm.com>
14345 David Sherwood <david.sherwood@arm.com>
14347 * doc/md.texi (fold_extract_last_@var{m}): Document.
14348 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14349 * optabs.def (fold_extract_last_optab): New optab.
14350 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14351 * internal-fn.c (fold_extract_direct): New macro.
14352 (expand_fold_extract_optab_fn): Likewise.
14353 (direct_fold_extract_optab_supported_p): Likewise.
14354 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14355 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14356 EXTRACT_LAST_REDUCTION.
14357 (get_initial_def_for_reduction): Do not create an initial vector
14358 for EXTRACT_LAST_REDUCTION reductions.
14359 (vectorizable_reduction): Leave the scalar phi in place for
14360 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14361 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14362 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14363 transform phase to vectorizable_condition.
14364 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14366 (vect_finish_stmt_generation): ...here.
14367 (vect_finish_replace_stmt): New function.
14368 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14369 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14371 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14374 Alan Hayward <alan.hayward@arm.com>
14375 David Sherwood <david.sherwood@arm.com>
14377 * doc/md.texi (extract_last_@var{m}): Document.
14378 * optabs.def (extract_last_optab): New optab.
14379 * internal-fn.def (EXTRACT_LAST): New internal function.
14380 * internal-fn.c (cond_unary_direct): New macro.
14381 (expand_cond_unary_optab_fn): Likewise.
14382 (direct_cond_unary_optab_supported_p): Likewise.
14383 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14384 loops using EXTRACT_LAST.
14385 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14386 (extract_last_<mode>): ...this optab.
14387 (vec_extract<mode><Vel>): Update accordingly.
14389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14390 Alan Hayward <alan.hayward@arm.com>
14391 David Sherwood <david.sherwood@arm.com>
14393 * target.def (empty_mask_is_expensive): New hook.
14394 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14395 * doc/tm.texi: Regenerate.
14396 * targhooks.h (default_empty_mask_is_expensive): Declare.
14397 * targhooks.c (default_empty_mask_is_expensive): New function.
14398 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14399 if the target says that empty masks are expensive.
14400 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14402 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14405 Alan Hayward <alan.hayward@arm.com>
14406 David Sherwood <david.sherwood@arm.com>
14408 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14409 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14410 (vect_use_loop_mask_for_alignment_p): New function.
14411 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14412 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14413 niters_skip argument. Make sure that the first niters_skip elements
14414 of the first iteration are inactive.
14415 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14416 Update call to vect_set_loop_masks_directly.
14417 (get_misalign_in_elems): New function, split out from...
14418 (vect_gen_prolog_loop_niters): ...here.
14419 (vect_update_init_of_dr): Take a code argument that specifies whether
14420 the adjustment should be added or subtracted.
14421 (vect_update_init_of_drs): Likewise.
14422 (vect_prepare_for_masked_peels): New function.
14423 (vect_do_peeling): Skip prologue peeling if we're using a mask
14424 instead. Update call to vect_update_inits_of_drs.
14425 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14427 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14428 alignment. Do not include the number of peeled iterations in
14429 the minimum threshold in that case.
14430 (vectorizable_induction): Adjust the start value down by
14431 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14432 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14433 Take the number of skipped iterations into account when calculating
14435 * tree-vect-stmts.c (vect_gen_while_not): New function.
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/sourcebuild.texi (vect_fully_masked): Document.
14442 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14443 default value to 0.
14444 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14446 (vect_analyze_loop_2): ...here. Don't check the vectorization
14447 factor against the number of loop iterations if the loop is
14450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14451 Alan Hayward <alan.hayward@arm.com>
14452 David Sherwood <david.sherwood@arm.com>
14454 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14455 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14456 (dump_groups): Update accordingly.
14457 (iv_use::mem_type): New member variable.
14458 (address_p): New function.
14459 (record_use): Add a mem_type argument and initialize the new
14461 (record_group_use): Add a mem_type argument. Use address_p.
14462 Remove obsolete null checks of base_object. Update call to record_use.
14463 (find_interesting_uses_op): Update call to record_group_use.
14464 (find_interesting_uses_cond): Likewise.
14465 (find_interesting_uses_address): Likewise.
14466 (get_mem_type_for_internal_fn): New function.
14467 (find_address_like_use): Likewise.
14468 (find_interesting_uses_stmt): Try find_address_like_use before
14469 calling find_interesting_uses_op.
14470 (addr_offset_valid_p): Use the iv mem_type field as the type
14471 of the addressed memory.
14472 (add_autoinc_candidates): Likewise.
14473 (get_address_cost): Likewise.
14474 (split_small_address_groups_p): Use address_p.
14475 (split_address_groups): Likewise.
14476 (add_iv_candidate_for_use): Likewise.
14477 (autoinc_possible_for_pair): Likewise.
14478 (rewrite_groups): Likewise.
14479 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14480 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14481 (get_alias_ptr_type_for_ptr_address): New function.
14482 (rewrite_use_address): Rewrite address uses in calls that were
14483 identified by find_address_like_use.
14485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14486 Alan Hayward <alan.hayward@arm.com>
14487 David Sherwood <david.sherwood@arm.com>
14489 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14491 * gimple-expr.h (is_gimple_addressable: Likewise.
14492 * gimple-expr.c (is_gimple_address): Likewise.
14493 * internal-fn.c (expand_call_mem_ref): New function.
14494 (expand_mask_load_optab_fn): Use it.
14495 (expand_mask_store_optab_fn): Likewise.
14497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14498 Alan Hayward <alan.hayward@arm.com>
14499 David Sherwood <david.sherwood@arm.com>
14501 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14502 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14503 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14504 (cond_umax@var{mode}): Document.
14505 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14506 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14507 (cond_umin_optab, cond_umax_optab): New optabs.
14508 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14509 (COND_IOR, COND_XOR): New internal functions.
14510 * internal-fn.h (get_conditional_internal_fn): Declare.
14511 * internal-fn.c (cond_binary_direct): New macro.
14512 (expand_cond_binary_optab_fn): Likewise.
14513 (direct_cond_binary_optab_supported_p): Likewise.
14514 (get_conditional_internal_fn): New function.
14515 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14516 Cope with reduction statements that are vectorized as calls rather
14518 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14519 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14520 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14521 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14522 (UNSPEC_COND_EOR): New unspecs.
14523 (optab): Add mappings for them.
14524 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14525 (sve_int_op, sve_fp_op): New int attributes.
14527 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14528 Alan Hayward <alan.hayward@arm.com>
14529 David Sherwood <david.sherwood@arm.com>
14531 * optabs.def (while_ult_optab): New optab.
14532 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14533 * internal-fn.def (WHILE_ULT): New internal function.
14534 * internal-fn.h (direct_internal_fn_supported_p): New override
14535 that takes two types as argument.
14536 * internal-fn.c (while_direct): New macro.
14537 (expand_while_optab_fn): New function.
14538 (convert_optab_supported_p): Likewise.
14539 (direct_while_optab_supported_p): New macro.
14540 * wide-int.h (wi::udiv_ceil): New function.
14541 * tree-vectorizer.h (rgroup_masks): New structure.
14542 (vec_loop_masks): New typedef.
14543 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14544 and fully_masked_p.
14545 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14546 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14547 (vect_max_vf): New function.
14548 (slpeel_make_loop_iterate_ntimes): Delete.
14549 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14550 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14551 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14552 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14553 internal-fn.h, stor-layout.h and optabs-query.h.
14554 (vect_set_loop_mask): New function.
14555 (add_preheader_seq): Likewise.
14556 (add_header_seq): Likewise.
14557 (interleave_supported_p): Likewise.
14558 (vect_maybe_permute_loop_masks): Likewise.
14559 (vect_set_loop_masks_directly): Likewise.
14560 (vect_set_loop_condition_masked): Likewise.
14561 (vect_set_loop_condition_unmasked): New function, split out from
14562 slpeel_make_loop_iterate_ntimes.
14563 (slpeel_make_loop_iterate_ntimes): Rename to..
14564 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14565 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14566 (vect_do_peeling): Update call accordingly.
14567 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14569 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14570 mask_compare_type, can_fully_mask_p and fully_masked_p.
14571 (release_vec_loop_masks): New function.
14572 (_loop_vec_info): Use it to free the loop masks.
14573 (can_produce_all_loop_masks_p): New function.
14574 (vect_get_max_nscalars_per_iter): Likewise.
14575 (vect_verify_full_masking): Likewise.
14576 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14577 retries, and free the mask rgroups before retrying. Check loop-wide
14578 reasons for disallowing fully-masked loops. Make the final decision
14579 about whether use a fully-masked loop or not.
14580 (vect_estimate_min_profitable_iters): Do not assume that peeling
14581 for the number of iterations will be needed for fully-masked loops.
14582 (vectorizable_reduction): Disable fully-masked loops.
14583 (vectorizable_live_operation): Likewise.
14584 (vect_halve_mask_nunits): New function.
14585 (vect_double_mask_nunits): Likewise.
14586 (vect_record_loop_mask): Likewise.
14587 (vect_get_loop_mask): Likewise.
14588 (vect_transform_loop): Handle the case in which the final loop
14589 iteration might handle a partial vector. Call vect_set_loop_condition
14590 instead of slpeel_make_loop_iterate_ntimes.
14591 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14592 (check_load_store_masking): New function.
14593 (prepare_load_store_mask): Likewise.
14594 (vectorizable_store): Handle fully-masked loops.
14595 (vectorizable_load): Likewise.
14596 (supportable_widening_operation): Use vect_halve_mask_nunits for
14598 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14599 (vect_gen_while): New function.
14600 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14601 (aarch64_uqdec<mode>): New insn.
14603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14604 Alan Hayward <alan.hayward@arm.com>
14605 David Sherwood <david.sherwood@arm.com>
14607 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14608 (reduc_xor_scal_optab): New optabs.
14609 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14610 (reduc_xor_scal_@var{m}): Document.
14611 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14612 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14613 internal functions.
14614 * fold-const-call.c (fold_const_call): Handle them.
14615 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14616 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14617 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14618 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14619 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14620 (UNSPEC_XORV): New unspecs.
14621 (optab): Add entries for them.
14622 (BITWISEV): New int iterator.
14623 (bit_reduc_op): New int attributes.
14625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14626 Alan Hayward <alan.hayward@arm.com>
14627 David Sherwood <david.sherwood@arm.com>
14629 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14630 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14631 * optabs.def (vec_shl_insert_optab): New optab.
14632 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14633 (duplicate_and_interleave): Likewise.
14634 * tree-vect-loop.c: Include internal-fn.h.
14635 (neutral_op_for_slp_reduction): New function, split out from
14636 get_initial_defs_for_reduction.
14637 (get_initial_def_for_reduction): Handle option 2 for variable-length
14638 vectors by loading the neutral value into a vector and then shifting
14639 the initial value into element 0.
14640 (get_initial_defs_for_reduction): Replace the code argument with
14641 the neutral value calculated by neutral_op_for_slp_reduction.
14642 Use gimple_build_vector for constant-length vectors.
14643 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14644 but the first group_size elements have a neutral value.
14645 Use duplicate_and_interleave otherwise.
14646 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14647 Update call to get_initial_defs_for_reduction. Handle SLP
14648 reductions for variable-length vectors by creating one vector
14649 result for each scalar result, with the elements associated
14650 with other scalar results stubbed out with the neutral value.
14651 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14652 Require IFN_VEC_SHL_INSERT for double reductions on
14653 variable-length vectors, or SLP reductions that have
14654 a neutral value. Require can_duplicate_and_interleave_p
14655 support for variable-length unchained SLP reductions if there
14656 is no neutral value, such as for MIN/MAX reductions. Also require
14657 the number of vector elements to be a multiple of the number of
14658 SLP statements when doing variable-length unchained SLP reductions.
14659 Update call to vect_create_epilog_for_reduction.
14660 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14661 and remove initial values.
14662 (duplicate_and_interleave): Make public.
14663 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14664 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14667 Alan Hayward <alan.hayward@arm.com>
14668 David Sherwood <david.sherwood@arm.com>
14670 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14671 (can_duplicate_and_interleave_p): New function.
14672 (vect_get_and_check_slp_defs): Take the vector of statements
14673 rather than just the current one. Remove excess parentheses.
14674 Restriction rejectinon of vect_constant_def and vect_external_def
14675 for variable-length vectors to boolean types, or types for which
14676 can_duplicate_and_interleave_p is false.
14677 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14678 (duplicate_and_interleave): New function.
14679 (vect_get_constant_vectors): Use gimple_build_vector for
14680 constant-length vectors and suitable variable-length constant
14681 vectors. Use duplicate_and_interleave for other variable-length
14682 vectors. Don't defer the update when inserting new statements.
14684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14685 Alan Hayward <alan.hayward@arm.com>
14686 David Sherwood <david.sherwood@arm.com>
14688 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14689 min_profitable_iters doesn't go negative.
14691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14692 Alan Hayward <alan.hayward@arm.com>
14693 David Sherwood <david.sherwood@arm.com>
14695 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14696 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14697 * optabs.def (vec_mask_load_lanes_optab): New optab.
14698 (vec_mask_store_lanes_optab): Likewise.
14699 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14700 (MASK_STORE_LANES): Likewise.
14701 * internal-fn.c (mask_load_lanes_direct): New macro.
14702 (mask_store_lanes_direct): Likewise.
14703 (expand_mask_load_optab_fn): Handle masked operations.
14704 (expand_mask_load_lanes_optab_fn): New macro.
14705 (expand_mask_store_optab_fn): Handle masked operations.
14706 (expand_mask_store_lanes_optab_fn): New macro.
14707 (direct_mask_load_lanes_optab_supported_p): Likewise.
14708 (direct_mask_store_lanes_optab_supported_p): Likewise.
14709 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14711 (vect_load_lanes_supported): Likewise.
14712 * tree-vect-data-refs.c (strip_conversion): New function.
14713 (can_group_stmts_p): Likewise.
14714 (vect_analyze_data_ref_accesses): Use it instead of checking
14715 for a pair of assignments.
14716 (vect_store_lanes_supported): Take a masked_p parameter.
14717 (vect_load_lanes_supported): Likewise.
14718 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14719 vect_store_lanes_supported and vect_load_lanes_supported.
14720 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14721 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14722 parameter. Don't allow gaps for masked accesses.
14723 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14724 and vect_load_lanes_supported.
14725 (get_load_store_type): Take a masked_p parameter and update
14726 call to get_group_load_store_type.
14727 (vectorizable_store): Update call to get_load_store_type.
14728 Handle IFN_MASK_STORE_LANES.
14729 (vectorizable_load): Update call to get_load_store_type.
14730 Handle IFN_MASK_LOAD_LANES.
14732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14733 Alan Hayward <alan.hayward@arm.com>
14734 David Sherwood <david.sherwood@arm.com>
14736 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14738 * config/aarch64/aarch64-protos.h
14739 (aarch64_sve_struct_memory_operand_p): Declare.
14740 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14741 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14742 (VPRED, vpred): Handle SVE structure modes.
14743 * config/aarch64/constraints.md (Utx): New constraint.
14744 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14745 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14746 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14747 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14748 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14749 structure modes. Split into pieces after RA.
14750 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14751 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14753 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14754 SVE structure modes.
14755 (aarch64_classify_address): Likewise.
14756 (sizetochar): Move earlier in file.
14757 (aarch64_print_operand): Handle SVE register lists.
14758 (aarch64_array_mode): New function.
14759 (aarch64_sve_struct_memory_operand_p): Likewise.
14760 (TARGET_ARRAY_MODE): Redefine.
14762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14763 Alan Hayward <alan.hayward@arm.com>
14764 David Sherwood <david.sherwood@arm.com>
14766 * target.def (array_mode): New target hook.
14767 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14768 * doc/tm.texi: Regenerate.
14769 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14770 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14771 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14772 targetm.array_mode.
14773 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14777 Alan Hayward <alan.hayward@arm.com>
14778 David Sherwood <david.sherwood@arm.com>
14780 * fold-const.c (fold_binary_loc): Check the argument types
14781 rather than the result type when testing for a vector operation.
14783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14785 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14786 * doc/tm.texi: Regenerate.
14788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14789 Alan Hayward <alan.hayward@arm.com>
14790 David Sherwood <david.sherwood@arm.com>
14792 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14793 (sve): Document new AArch64 extension.
14794 * doc/md.texi (w): Extend the description of the AArch64
14795 constraint to include SVE vectors.
14796 (Upl, Upa): Document new AArch64 predicate constraints.
14797 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14799 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14800 (msve-vector-bits=): New option.
14801 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14802 SVE when these are disabled.
14803 (sve): New extension.
14804 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14805 modes. Adjust their number of units based on aarch64_sve_vg.
14806 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14807 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14808 aarch64_addr_query_type.
14809 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14810 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14811 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14812 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14813 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14814 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14815 (aarch64_simd_imm_zero_p): Delete.
14816 (aarch64_check_zero_based_sve_index_immediate): Declare.
14817 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14818 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14819 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14820 (aarch64_sve_float_mul_immediate_p): Likewise.
14821 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14822 rather than an rtx.
14823 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14824 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14825 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14826 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14827 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14828 (aarch64_regmode_natural_size): Likewise.
14829 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14830 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14832 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14833 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14834 for VG and the SVE predicate registers.
14835 (V_ALIASES): Add a "z"-prefixed alias.
14836 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14837 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14838 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14839 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14840 (REG_CLASS_NAMES): Add entries for them.
14841 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14842 and the predicate registers.
14843 (aarch64_sve_vg): Declare.
14844 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14845 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14846 (REGMODE_NATURAL_SIZE): Define.
14847 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14849 * config/aarch64/aarch64.c: Include cfgrtl.h.
14850 (simd_immediate_info): Add a constructor for series vectors,
14851 and an associated step field.
14852 (aarch64_sve_vg): New variable.
14853 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14854 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14855 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14856 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14857 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14858 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14859 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14860 (aarch64_get_mask_mode): New functions.
14861 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14862 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14863 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14864 predicate modes and predicate registers. Explicitly restrict
14865 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14866 to store a vector mode if it is recognized by
14867 aarch64_classify_vector_mode.
14868 (aarch64_regmode_natural_size): New function.
14869 (aarch64_hard_regno_caller_save_mode): Return the original mode
14871 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14872 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14873 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14874 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14876 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14877 does not overlap dest if the function is frame-related. Handle
14879 (aarch64_split_add_offset): New function.
14880 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14881 them aarch64_add_offset.
14882 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14883 and update call to aarch64_sub_sp.
14884 (aarch64_add_cfa_expression): New function.
14885 (aarch64_expand_prologue): Pass extra temporary registers to the
14886 functions above. Handle the case in which we need to emit new
14887 DW_CFA_expressions for registers that were originally saved
14888 relative to the stack pointer, but now have to be expressed
14889 relative to the frame pointer.
14890 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14892 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14893 IP0 and IP1 values for SVE frames.
14894 (aarch64_expand_vec_series): New function.
14895 (aarch64_expand_sve_widened_duplicate): Likewise.
14896 (aarch64_expand_sve_const_vector): Likewise.
14897 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14898 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14899 into the register, rather than emitting a SET directly.
14900 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14901 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14902 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14903 (offset_9bit_signed_scaled_p): New functions.
14904 (aarch64_replicate_bitmask_imm): New function.
14905 (aarch64_bitmask_imm): Use it.
14906 (aarch64_cannot_force_const_mem): Reject expressions involving
14907 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14908 (aarch64_classify_index): Handle SVE indices, by requiring
14909 a plain register index with a scale that matches the element size.
14910 (aarch64_classify_address): Handle SVE addresses. Assert that
14911 the mode of the address is VOIDmode or an integer mode.
14912 Update call to aarch64_classify_symbol.
14913 (aarch64_classify_symbolic_expression): Update call to
14914 aarch64_classify_symbol.
14915 (aarch64_const_vec_all_in_range_p): New function.
14916 (aarch64_print_vector_float_operand): Likewise.
14917 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14918 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14919 and the FP immediates 1.0 and 0.5.
14920 (aarch64_print_address_internal): Handle SVE addresses.
14921 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14922 (aarch64_regno_regclass): Handle predicate registers.
14923 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14925 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14926 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14927 (aarch64_convert_sve_vector_bits): New function.
14928 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14929 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14930 rather than an rtx.
14931 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14932 Handle SVE vector and predicate modes. Accept VL-based constants
14933 that need only one temporary register, and VL offsets that require
14934 no temporary registers.
14935 (aarch64_conditional_register_usage): Mark the predicate registers
14936 as fixed if SVE isn't available.
14937 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14938 Return true for SVE vector and predicate modes.
14939 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14940 rather than an unsigned int. Handle SVE modes.
14941 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14943 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14945 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14946 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14947 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14948 (aarch64_sve_float_mul_immediate_p): New functions.
14949 (aarch64_sve_valid_immediate): New function.
14950 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14951 Explicitly reject structure modes. Check for INDEX constants.
14952 Handle PTRUE and PFALSE constants.
14953 (aarch64_check_zero_based_sve_index_immediate): New function.
14954 (aarch64_simd_imm_zero_p): Delete.
14955 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14956 vector modes. Accept constants in the range of CNT[BHWD].
14957 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14958 ask for an Advanced SIMD mode.
14959 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14960 (aarch64_simd_vector_alignment): Handle SVE predicates.
14961 (aarch64_vectorize_preferred_vector_alignment): New function.
14962 (aarch64_simd_vector_alignment_reachable): Use it instead of
14964 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14965 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14967 (MAX_VECT_LEN): Delete.
14968 (expand_vec_perm_d): Add a vec_flags field.
14969 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14970 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14971 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14973 (aarch64_evpc_rev): Rename to...
14974 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14975 (aarch64_evpc_rev_global): New function.
14976 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14977 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14979 (aarch64_evpc_sve_tbl): New function.
14980 (aarch64_expand_vec_perm_const_1): Update after rename of
14981 aarch64_evpc_rev. Handle SVE permutes too, trying
14982 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14983 than aarch64_evpc_tbl.
14984 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14985 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14986 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14987 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14988 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14989 (aarch64_expand_sve_vcond): New functions.
14990 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14991 of aarch64_vector_mode_p.
14992 (aarch64_dwarf_poly_indeterminate_value): New function.
14993 (aarch64_compute_pressure_classes): Likewise.
14994 (aarch64_can_change_mode_class): Likewise.
14995 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14996 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14997 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14998 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14999 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15000 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15001 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15002 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15004 (Dn, Dl, Dr): Accept const as well as const_vector.
15005 (Dz): Likewise. Compare against CONST0_RTX.
15006 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15007 of "vector" where appropriate.
15008 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15009 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15010 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15011 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15012 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15013 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15014 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15015 (v_int_equiv): Extend to SVE modes.
15016 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15018 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15019 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15020 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15021 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15022 (SVE_COND_FP_CMP): New int iterators.
15023 (perm_hilo): Handle the new unpack unspecs.
15024 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15026 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15027 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15028 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15029 (aarch64_equality_operator, aarch64_constant_vector_operand)
15030 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15031 (aarch64_sve_nonimmediate_operand): Likewise.
15032 (aarch64_sve_general_operand): Likewise.
15033 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15034 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15035 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15036 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15037 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15038 (aarch64_sve_float_arith_immediate): Likewise.
15039 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15040 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15041 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15042 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15043 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15044 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15045 (aarch64_sve_float_arith_operand): Likewise.
15046 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15047 (aarch64_sve_float_mul_operand): Likewise.
15048 (aarch64_sve_vec_perm_operand): Likewise.
15049 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15050 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15051 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15052 as well as const_vector.
15053 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15054 in file. Use CONST0_RTX and CONSTM1_RTX.
15055 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15056 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15057 Use aarch64_simd_imm_zero.
15058 * config/aarch64/aarch64-sve.md: New file.
15059 * config/aarch64/aarch64.md: Include it.
15060 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15061 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15062 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15063 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15064 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15065 (sve): New attribute.
15066 (enabled): Disable instructions with the sve attribute unless
15068 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15069 aarch64_expand_mov_immediate.
15070 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15071 CNT[BHSD] immediates.
15072 (movti): Split CONST_POLY_INT moves into two halves.
15073 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15074 Split additions that need a temporary here if the destination
15075 is the stack pointer.
15076 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15077 (*add<mode>3_poly_1): New instruction.
15078 (set_clobber_cc): New expander.
15080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15082 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15083 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15084 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15085 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15086 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15087 Change innermode from fixed_mode_size to machine_mode.
15088 (simplify_subreg): Update call accordingly. Handle a constant-sized
15089 subreg of a variable-length CONST_VECTOR.
15091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15092 Alan Hayward <alan.hayward@arm.com>
15093 David Sherwood <david.sherwood@arm.com>
15095 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15096 (add_offset_to_base): New function, split out from...
15097 (create_mem_ref): ...here. When handling a scale other than 1,
15098 check first whether the address is valid without the offset.
15099 Add it into the base if so, leaving the index and scale as-is.
15101 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15104 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15105 fold_for_warn before checking if arg2 is INTEGER_CST.
15107 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15109 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15110 (store_multiple_operation): Delete.
15111 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15112 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15113 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15114 guarded by TARGET_STRING.
15115 (rs6000_output_load_multiple): Delete.
15116 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15117 OPTION_MASK_STRING / TARGET_STRING handling.
15118 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15119 (const rs6000_opt_masks) <"string">: Change mask to 0.
15120 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15121 (MASK_STRING): Delete.
15122 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15124 (load_multiple): Delete.
15131 (store_multiple): Delete.
15138 (movmemsi_8reg): Delete.
15139 (corresponding unnamed define_insn): Delete.
15140 (movmemsi_6reg): Delete.
15141 (corresponding unnamed define_insn): Delete.
15142 (movmemsi_4reg): Delete.
15143 (corresponding unnamed define_insn): Delete.
15144 (movmemsi_2reg): Delete.
15145 (corresponding unnamed define_insn): Delete.
15146 (movmemsi_1reg): Delete.
15147 (corresponding unnamed define_insn): Delete.
15148 * config/rs6000/rs6000.opt (mno-string): New.
15149 (mstring): Replace by deprecation warning stub.
15150 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15152 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15154 * regrename.c (regrename_do_replace): If replacing the same
15155 reg multiple times, try to reuse last created gen_raw_REG.
15158 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15159 main to workaround a bug in GDB.
15161 2018-01-12 Tom de Vries <tom@codesourcery.com>
15164 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15166 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15168 PR rtl-optimization/80481
15169 * ira-color.c (get_cap_member): New function.
15170 (allocnos_conflict_by_live_ranges_p): Use it.
15171 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15172 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15174 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15177 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15178 (*saddl_se_1): Ditto.
15179 (*ssubsi_1): Ditto.
15180 (*ssubl_se_1): Ditto.
15182 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15184 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15185 rather than wi::to_widest for DR_INITs.
15186 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15187 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15188 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15190 (vect_analyze_group_access_1): Note that here.
15192 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15194 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15195 polynomial type sizes.
15197 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15199 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15200 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15201 (gimple_add_tmp_var): Likewise.
15203 2018-01-12 Martin Liska <mliska@suse.cz>
15205 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15206 (gimple_alloc_sizes): Likewise.
15207 (dump_gimple_statistics): Use PRIu64 in printf format.
15208 * gimple.h: Change uint64_t to int.
15210 2018-01-12 Martin Liska <mliska@suse.cz>
15212 * tree-core.h: Use uint64_t instead of int.
15213 * tree.c (tree_node_counts): Likewise.
15214 (tree_node_sizes): Likewise.
15215 (dump_tree_statistics): Use PRIu64 in printf format.
15217 2018-01-12 Martin Liska <mliska@suse.cz>
15219 * Makefile.in: As qsort_chk is implemented in vec.c, add
15220 vec.o to linkage of gencfn-macros.
15221 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15222 passing the info to record_node_allocation_statistics.
15223 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15225 * ggc-common.c (struct ggc_usage): Add operator== and use
15226 it in operator< and compare function.
15227 * mem-stats.h (struct mem_usage): Likewise.
15228 * vec.c (struct vec_usage): Remove operator< and compare
15229 function. Can be simply inherited.
15231 2018-01-12 Martin Jambor <mjambor@suse.cz>
15234 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15235 * tree-ssa-math-opts.c: Include domwalk.h.
15236 (convert_mult_to_fma_1): New function.
15237 (fma_transformation_info): New type.
15238 (fma_deferring_state): Likewise.
15239 (cancel_fma_deferring): New function.
15240 (result_of_phi): Likewise.
15241 (last_fma_candidate_feeds_initial_phi): Likewise.
15242 (convert_mult_to_fma): Added deferring logic, split actual
15243 transformation to convert_mult_to_fma_1.
15244 (math_opts_dom_walker): New type.
15245 (math_opts_dom_walker::after_dom_children): New method, body moved
15246 here from pass_optimize_widening_mul::execute, added deferring logic
15248 (pass_optimize_widening_mul::execute): Moved most of code to
15249 math_opts_dom_walker::after_dom_children.
15250 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15251 * config/i386/i386.c (ix86_option_override_internal): Added
15252 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15254 2018-01-12 Richard Biener <rguenther@suse.de>
15257 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15258 inline instance vars.
15260 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15263 * config/rx/rx.c (rx_is_restricted_memory_address):
15264 Handle SUBREG case.
15266 2018-01-12 Richard Biener <rguenther@suse.de>
15268 PR tree-optimization/80846
15269 * target.def (split_reduction): New target hook.
15270 * targhooks.c (default_split_reduction): New function.
15271 * targhooks.h (default_split_reduction): Declare.
15272 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15273 target requests first reduce vectors by combining low and high
15275 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15276 (get_vectype_for_scalar_type_and_size): Export.
15277 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15278 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15279 * doc/tm.texi: Regenerate.
15280 * config/i386/i386.c (ix86_split_reduction): Implement
15281 TARGET_VECTORIZE_SPLIT_REDUCTION.
15283 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15286 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15287 in PIC mode except for TARGET_VXWORKS_RTP.
15288 * config/sparc/sparc.c: Include cfgrtl.h.
15289 (TARGET_INIT_PIC_REG): Define.
15290 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15291 (sparc_pic_register_p): New predicate.
15292 (sparc_legitimate_address_p): Use it.
15293 (sparc_legitimize_pic_address): Likewise.
15294 (sparc_delegitimize_address): Likewise.
15295 (sparc_mode_dependent_address_p): Likewise.
15296 (gen_load_pcrel_sym): Remove 4th parameter.
15297 (load_got_register): Adjust call to above. Remove obsolete stuff.
15298 (sparc_expand_prologue): Do not call load_got_register here.
15299 (sparc_flat_expand_prologue): Likewise.
15300 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15301 (sparc_use_pseudo_pic_reg): New function.
15302 (sparc_init_pic_reg): Likewise.
15303 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15304 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15306 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15308 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15309 Add item for branch_cost.
15311 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15313 PR rtl-optimization/83565
15314 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15315 not extend the result to a larger mode for rotate operations.
15316 (num_sign_bit_copies1): Likewise.
15318 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15321 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15323 Use values-Xc.o for -pedantic.
15324 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15326 2018-01-12 Martin Liska <mliska@suse.cz>
15329 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15331 (possible_polymorphic_call_targets): Use it.
15332 (ipa_devirt): Likewise.
15334 2018-01-12 Martin Liska <mliska@suse.cz>
15336 * profile-count.h (enum profile_quality): Use 0 as invalid
15337 enum value of profile_quality.
15339 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15341 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15342 -mext-string options.
15344 2018-01-12 Richard Biener <rguenther@suse.de>
15346 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15347 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15348 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15350 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15352 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15354 * configure.ac (--with-long-double-format): Add support for the
15355 configuration option to change the default long double format on
15357 * config.gcc (powerpc*-linux*-*): Likewise.
15358 * configure: Regenerate.
15359 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15360 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15361 used without modification.
15363 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15365 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15366 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15367 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15368 MISC_BUILTIN_SPEC_BARRIER.
15369 (rs6000_init_builtins): Likewise.
15370 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15372 (speculation_barrier): New define_insn.
15373 * doc/extend.texi: Document __builtin_speculation_barrier.
15375 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15378 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15379 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15380 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15382 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15383 integral modes instead of "ss" and "sd".
15384 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15385 vectors with 32-bit and 64-bit elements.
15386 (vecdupssescalarmodesuffix): New mode attribute.
15387 (vec_dup<mode>): Use it.
15389 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15392 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15393 frame if argument is passed on stack.
15395 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15398 * ree.c (combine_reaching_defs): Optimize also
15399 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15400 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15402 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15404 PR middle-end/83189
15405 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15407 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15409 PR middle-end/83718
15410 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15411 after they are computed.
15413 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15415 PR tree-optimization/83695
15416 * gimple-loop-linterchange.cc
15417 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15418 reset cached scev information after interchange.
15419 (pass_linterchange::execute): Remove call to scev_reset_htab.
15421 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15423 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15424 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15425 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15426 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15427 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15428 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15429 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15430 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15431 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15432 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15433 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15434 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15435 (V_lane_reg): Likewise.
15436 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15438 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15439 (vfmal_lane_low<mode>_intrinsic,
15440 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15441 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15442 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15443 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15444 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15445 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15447 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15449 * config/arm/arm-cpus.in (fp16fml): New feature.
15450 (ALL_SIMD): Add fp16fml.
15451 (armv8.2-a): Add fp16fml as an option.
15452 (armv8.3-a): Likewise.
15453 (armv8.4-a): Add fp16fml as part of fp16.
15454 * config/arm/arm.h (TARGET_FP16FML): Define.
15455 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15457 * config/arm/arm-modes.def (V2HF): Define.
15458 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15459 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15460 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15461 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15462 vfmsl_low, vfmsl_high): New set of builtins.
15463 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15464 (vfml_op): New code attribute.
15465 (VFMLHALVES): New int iterator.
15466 (VFML, VFMLSEL): New mode attributes.
15467 (V_reg): Define mapping for V2HF.
15468 (V_hi, V_lo): New mode attributes.
15469 (VF_constraint): Likewise.
15470 (vfml_half, vfml_half_selector): New int attributes.
15471 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15473 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15474 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15476 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15477 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15478 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15479 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15481 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15482 Document new effective target and option set.
15484 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15486 * config/arm/arm-cpus.in (armv8_4): New feature.
15487 (ARMv8_4a): New fgroup.
15488 (armv8.4-a): New arch.
15489 * config/arm/arm-tables.opt: Regenerate.
15490 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15491 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15492 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15493 Add matching rules for -march=armv8.4-a and extensions.
15494 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15496 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15499 * config/rx/rx.md (BW): New mode attribute.
15500 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15502 2018-01-11 Richard Biener <rguenther@suse.de>
15504 PR tree-optimization/83435
15505 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15506 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15507 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15509 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15510 Alan Hayward <alan.hayward@arm.com>
15511 David Sherwood <david.sherwood@arm.com>
15513 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15515 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15516 (aarch64_print_address_internal): Use it to check for a zero offset.
15518 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15519 Alan Hayward <alan.hayward@arm.com>
15520 David Sherwood <david.sherwood@arm.com>
15522 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15523 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15524 Return a poly_int64 rather than a HOST_WIDE_INT.
15525 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15526 rather than a HOST_WIDE_INT.
15527 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15528 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15529 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15530 final_offset from HOST_WIDE_INT to poly_int64.
15531 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15532 to_constant when getting the number of units in an Advanced SIMD
15534 (aarch64_builtin_vectorized_function): Check for a constant number
15536 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15538 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15539 attribute instead of GET_MODE_NUNITS.
15540 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15541 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15542 GET_MODE_SIZE for fixed-size registers.
15543 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15544 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15545 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15546 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15547 (aarch64_print_operand, aarch64_print_address_internal)
15548 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15549 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15550 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15551 Handle polynomial GET_MODE_SIZE.
15552 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15553 wider than SImode without modification.
15554 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15555 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15556 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15557 passing and returning SVE modes.
15558 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15559 rather than GEN_INT.
15560 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15561 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15562 (aarch64_allocate_and_probe_stack_space): Likewise.
15563 (aarch64_layout_frame): Cope with polynomial offsets.
15564 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15565 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15566 polynomial offsets.
15567 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15568 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15569 poly_int64 rather than a HOST_WIDE_INT.
15570 (aarch64_get_separate_components, aarch64_process_components)
15571 (aarch64_expand_prologue, aarch64_expand_epilogue)
15572 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15573 (aarch64_anchor_offset): New function, split out from...
15574 (aarch64_legitimize_address): ...here.
15575 (aarch64_builtin_vectorization_cost): Handle polynomial
15576 TYPE_VECTOR_SUBPARTS.
15577 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15579 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15580 number of elements from the PARALLEL rather than the mode.
15581 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15582 rather than GET_MODE_BITSIZE.
15583 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15584 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15585 (aarch64_expand_vec_perm_const_1): Handle polynomial
15586 d->perm.length () and d->perm elements.
15587 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15588 Apply to_constant to d->perm elements.
15589 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15590 polynomial CONST_VECTOR_NUNITS.
15591 (aarch64_move_pointer): Take amount as a poly_int64 rather
15593 (aarch64_progress_pointer): Avoid temporary variable.
15594 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15595 the mode attribute instead of GET_MODE.
15597 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15598 Alan Hayward <alan.hayward@arm.com>
15599 David Sherwood <david.sherwood@arm.com>
15601 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15602 x exists before using it.
15603 (aarch64_add_constant_internal): Rename to...
15604 (aarch64_add_offset_1): ...this. Replace regnum with separate
15605 src and dest rtxes. Handle the case in which they're different,
15606 including when the offset is zero. Replace scratchreg with an rtx.
15607 Use 2 additions if there is no spare register into which we can
15608 move a 16-bit constant.
15609 (aarch64_add_constant): Delete.
15610 (aarch64_add_offset): Replace reg with separate src and dest
15611 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15612 Use aarch64_add_offset_1.
15613 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15614 an rtx rather than an int. Take the delta as a poly_int64
15615 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15616 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15617 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15618 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15619 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15620 and aarch64_add_sp.
15621 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15622 aarch64_add_constant.
15624 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15626 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15627 Use scalar_float_mode.
15629 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15631 * config/aarch64/aarch64-simd.md
15632 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15633 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15634 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15635 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15636 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15637 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15638 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15639 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15640 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15641 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15643 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15646 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15647 targ_options->x_arm_arch_string is non NULL.
15649 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15651 * config/aarch64/aarch64.h
15652 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15654 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15657 * expmed.c (emit_store_flag_force): Swap if const op0
15658 and change VOIDmode to mode of op0.
15660 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15662 PR rtl-optimization/83761
15663 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15664 than bytes to mode_for_size.
15666 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15668 PR middle-end/83189
15669 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15670 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15673 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15675 PR middle-end/83575
15676 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15677 when in layout mode.
15678 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15679 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15682 2018-01-10 Michael Collison <michael.collison@arm.com>
15684 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15685 * config/aarch64/aarch64-option-extension.def: Add
15686 AARCH64_OPT_EXTENSION of 'fp16fml'.
15687 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15688 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15689 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15690 * config/aarch64/constraints.md (Ui7): New constraint.
15691 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15692 (VFMLA_SEL_W): Ditto.
15695 (VFMLA16_LOW): New int iterator.
15696 (VFMLA16_HIGH): Ditto.
15697 (UNSPEC_FMLAL): New unspec.
15698 (UNSPEC_FMLSL): Ditto.
15699 (UNSPEC_FMLAL2): Ditto.
15700 (UNSPEC_FMLSL2): Ditto.
15701 (f16mac): New code attribute.
15702 * config/aarch64/aarch64-simd-builtins.def
15703 (aarch64_fmlal_lowv2sf): Ditto.
15704 (aarch64_fmlsl_lowv2sf): Ditto.
15705 (aarch64_fmlalq_lowv4sf): Ditto.
15706 (aarch64_fmlslq_lowv4sf): Ditto.
15707 (aarch64_fmlal_highv2sf): Ditto.
15708 (aarch64_fmlsl_highv2sf): Ditto.
15709 (aarch64_fmlalq_highv4sf): Ditto.
15710 (aarch64_fmlslq_highv4sf): Ditto.
15711 (aarch64_fmlal_lane_lowv2sf): Ditto.
15712 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15713 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15714 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15715 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15716 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15717 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15718 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15719 (aarch64_fmlal_lane_highv2sf): Ditto.
15720 (aarch64_fmlsl_lane_highv2sf): Ditto.
15721 (aarch64_fmlal_laneq_highv2sf): Ditto.
15722 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15723 (aarch64_fmlalq_lane_highv4sf): Ditto.
15724 (aarch64_fmlsl_lane_highv4sf): Ditto.
15725 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15726 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15727 * config/aarch64/aarch64-simd.md:
15728 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15729 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15730 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15731 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15732 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15733 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15734 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15735 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15736 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15737 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15738 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15739 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15740 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15741 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15742 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15743 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15744 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15745 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15746 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15747 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15748 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15749 (vfmlsl_low_u32): Ditto.
15750 (vfmlalq_low_u32): Ditto.
15751 (vfmlslq_low_u32): Ditto.
15752 (vfmlal_high_u32): Ditto.
15753 (vfmlsl_high_u32): Ditto.
15754 (vfmlalq_high_u32): Ditto.
15755 (vfmlslq_high_u32): Ditto.
15756 (vfmlal_lane_low_u32): Ditto.
15757 (vfmlsl_lane_low_u32): Ditto.
15758 (vfmlal_laneq_low_u32): Ditto.
15759 (vfmlsl_laneq_low_u32): Ditto.
15760 (vfmlalq_lane_low_u32): Ditto.
15761 (vfmlslq_lane_low_u32): Ditto.
15762 (vfmlalq_laneq_low_u32): Ditto.
15763 (vfmlslq_laneq_low_u32): Ditto.
15764 (vfmlal_lane_high_u32): Ditto.
15765 (vfmlsl_lane_high_u32): Ditto.
15766 (vfmlal_laneq_high_u32): Ditto.
15767 (vfmlsl_laneq_high_u32): Ditto.
15768 (vfmlalq_lane_high_u32): Ditto.
15769 (vfmlslq_lane_high_u32): Ditto.
15770 (vfmlalq_laneq_high_u32): Ditto.
15771 (vfmlslq_laneq_high_u32): Ditto.
15772 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15773 (AARCH64_FL_FOR_ARCH8_4): New.
15774 (AARCH64_ISA_F16FML): New ISA flag.
15775 (TARGET_F16FML): New feature flag for fp16fml.
15776 (doc/invoke.texi): Document new fp16fml option.
15778 2018-01-10 Michael Collison <michael.collison@arm.com>
15780 * config/aarch64/aarch64-builtins.c:
15781 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15782 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15783 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15784 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15785 (AARCH64_ISA_SHA3): New ISA flag.
15786 (TARGET_SHA3): New feature flag for sha3.
15787 * config/aarch64/iterators.md (sha512_op): New int attribute.
15788 (CRYPTO_SHA512): New int iterator.
15789 (UNSPEC_SHA512H): New unspec.
15790 (UNSPEC_SHA512H2): Ditto.
15791 (UNSPEC_SHA512SU0): Ditto.
15792 (UNSPEC_SHA512SU1): Ditto.
15793 * config/aarch64/aarch64-simd-builtins.def
15794 (aarch64_crypto_sha512hqv2di): New builtin.
15795 (aarch64_crypto_sha512h2qv2di): Ditto.
15796 (aarch64_crypto_sha512su0qv2di): Ditto.
15797 (aarch64_crypto_sha512su1qv2di): Ditto.
15798 (aarch64_eor3qv8hi): Ditto.
15799 (aarch64_rax1qv2di): Ditto.
15800 (aarch64_xarqv2di): Ditto.
15801 (aarch64_bcaxqv8hi): Ditto.
15802 * config/aarch64/aarch64-simd.md:
15803 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15804 (aarch64_crypto_sha512su0qv2di): Ditto.
15805 (aarch64_crypto_sha512su1qv2di): Ditto.
15806 (aarch64_eor3qv8hi): Ditto.
15807 (aarch64_rax1qv2di): Ditto.
15808 (aarch64_xarqv2di): Ditto.
15809 (aarch64_bcaxqv8hi): Ditto.
15810 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15811 (vsha512h2q_u64): Ditto.
15812 (vsha512su0q_u64): Ditto.
15813 (vsha512su1q_u64): Ditto.
15814 (veor3q_u16): Ditto.
15815 (vrax1q_u64): Ditto.
15816 (vxarq_u64): Ditto.
15817 (vbcaxq_u16): Ditto.
15818 * config/arm/types.md (crypto_sha512): New type attribute.
15819 (crypto_sha3): Ditto.
15820 (doc/invoke.texi): Document new sha3 option.
15822 2018-01-10 Michael Collison <michael.collison@arm.com>
15824 * config/aarch64/aarch64-builtins.c:
15825 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15826 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15827 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15828 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15829 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15830 (AARCH64_ISA_SM4): New ISA flag.
15831 (TARGET_SM4): New feature flag for sm4.
15832 * config/aarch64/aarch64-simd-builtins.def
15833 (aarch64_sm3ss1qv4si): Ditto.
15834 (aarch64_sm3tt1aq4si): Ditto.
15835 (aarch64_sm3tt1bq4si): Ditto.
15836 (aarch64_sm3tt2aq4si): Ditto.
15837 (aarch64_sm3tt2bq4si): Ditto.
15838 (aarch64_sm3partw1qv4si): Ditto.
15839 (aarch64_sm3partw2qv4si): Ditto.
15840 (aarch64_sm4eqv4si): Ditto.
15841 (aarch64_sm4ekeyqv4si): Ditto.
15842 * config/aarch64/aarch64-simd.md:
15843 (aarch64_sm3ss1qv4si): Ditto.
15844 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15845 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15846 (aarch64_sm4eqv4si): Ditto.
15847 (aarch64_sm4ekeyqv4si): Ditto.
15848 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15849 (sm3part_op): Ditto.
15850 (CRYPTO_SM3TT): Ditto.
15851 (CRYPTO_SM3PART): Ditto.
15852 (UNSPEC_SM3SS1): New unspec.
15853 (UNSPEC_SM3TT1A): Ditto.
15854 (UNSPEC_SM3TT1B): Ditto.
15855 (UNSPEC_SM3TT2A): Ditto.
15856 (UNSPEC_SM3TT2B): Ditto.
15857 (UNSPEC_SM3PARTW1): Ditto.
15858 (UNSPEC_SM3PARTW2): Ditto.
15859 (UNSPEC_SM4E): Ditto.
15860 (UNSPEC_SM4EKEY): Ditto.
15861 * config/aarch64/constraints.md (Ui2): New constraint.
15862 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15863 * config/arm/types.md (crypto_sm3): New type attribute.
15864 (crypto_sm4): Ditto.
15865 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15866 (vsm3tt1aq_u32): Ditto.
15867 (vsm3tt1bq_u32): Ditto.
15868 (vsm3tt2aq_u32): Ditto.
15869 (vsm3tt2bq_u32): Ditto.
15870 (vsm3partw1q_u32): Ditto.
15871 (vsm3partw2q_u32): Ditto.
15872 (vsm4eq_u32): Ditto.
15873 (vsm4ekeyq_u32): Ditto.
15874 (doc/invoke.texi): Document new sm4 option.
15876 2018-01-10 Michael Collison <michael.collison@arm.com>
15878 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15879 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15880 (AARCH64_FL_FOR_ARCH8_4): New.
15881 (AARCH64_FL_V8_4): New flag.
15882 (doc/invoke.texi): Document new armv8.4-a option.
15884 2018-01-10 Michael Collison <michael.collison@arm.com>
15886 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15887 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15888 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15889 * config/aarch64/aarch64-option-extension.def: Add
15890 AARCH64_OPT_EXTENSION of 'sha2'.
15891 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15892 (crypto): Disable sha2 and aes if crypto disabled.
15893 (crypto): Enable aes and sha2 if enabled.
15894 (simd): Disable sha2 and aes if simd disabled.
15895 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15897 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15898 (TARGET_SHA2): New feature flag for sha2.
15899 (TARGET_AES): New feature flag for aes.
15900 * config/aarch64/aarch64-simd.md:
15901 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15902 conditional on TARGET_AES.
15903 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15904 (aarch64_crypto_sha1hsi): Make pattern conditional
15906 (aarch64_crypto_sha1hv4si): Ditto.
15907 (aarch64_be_crypto_sha1hv4si): Ditto.
15908 (aarch64_crypto_sha1su1v4si): Ditto.
15909 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15910 (aarch64_crypto_sha1su0v4si): Ditto.
15911 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15912 (aarch64_crypto_sha256su0v4si): Ditto.
15913 (aarch64_crypto_sha256su1v4si): Ditto.
15914 (doc/invoke.texi): Document new aes and sha2 options.
15916 2018-01-10 Martin Sebor <msebor@redhat.com>
15918 PR tree-optimization/83781
15919 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15922 2018-01-11 Martin Sebor <msebor@gmail.com>
15923 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15925 PR tree-optimization/83501
15926 PR tree-optimization/81703
15928 * tree-ssa-strlen.c (get_string_cst): Rename...
15929 (get_string_len): ...to this. Handle global constants.
15930 (handle_char_store): Adjust.
15932 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15933 Jim Wilson <jimw@sifive.com>
15935 * config/riscv/riscv-protos.h (riscv_output_return): New.
15936 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15937 (riscv_attribute_table, riscv_output_return),
15938 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15939 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15940 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15942 (riscv_expand_prologue): Add early return for naked function.
15943 (riscv_expand_epilogue): Likewise.
15944 (riscv_function_ok_for_sibcall): Return false for naked function.
15945 (riscv_set_current_function): New.
15946 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15947 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15948 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15949 * doc/extend.texi (RISC-V Function Attributes): New.
15951 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15953 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15954 check for 128-bit long double before checking TCmode.
15955 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15956 128-bit long doubles before checking TFmode or TCmode.
15957 (FLOAT128_IBM_P): Likewise.
15959 2018-01-10 Martin Sebor <msebor@redhat.com>
15961 PR tree-optimization/83671
15962 * builtins.c (c_strlen): Unconditionally return zero for the empty
15964 Use -Warray-bounds for warnings.
15965 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15966 for non-constant array indices with COMPONENT_REF, arrays of
15967 arrays, and pointers to arrays.
15968 (gimple_fold_builtin_strlen): Determine and set length range for
15969 non-constant character arrays.
15971 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15973 PR middle-end/81897
15974 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15977 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15979 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15981 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15984 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15985 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15986 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15987 indexed_or_indirect_operand predicate.
15988 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15989 (*vsx_le_perm_load_v8hi): Likewise.
15990 (*vsx_le_perm_load_v16qi): Likewise.
15991 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15992 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15993 (*vsx_le_perm_store_v8hi): Likewise.
15994 (*vsx_le_perm_store_v16qi): Likewise.
15995 (eight unnamed splitters): Likewise.
15997 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15999 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16000 * config/rs6000/emmintrin.h: Likewise.
16001 * config/rs6000/mmintrin.h: Likewise.
16002 * config/rs6000/xmmintrin.h: Likewise.
16004 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16007 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16009 * tree.c (tree_nop_conversion): Return true for location wrapper
16011 (maybe_wrap_with_location): New function.
16012 (selftest::check_strip_nops): New function.
16013 (selftest::test_location_wrappers): New function.
16014 (selftest::tree_c_tests): Call it.
16015 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16016 (maybe_wrap_with_location): New decl.
16017 (EXPR_LOCATION_WRAPPER_P): New macro.
16018 (location_wrapper_p): New inline function.
16019 (tree_strip_any_location_wrapper): New inline function.
16021 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16024 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16025 stack_realign_offset for the largest alignment of stack slot
16027 (ix86_find_max_used_stack_alignment): New function.
16028 (ix86_finalize_stack_frame_flags): Use it. Set
16029 max_used_stack_alignment if we don't realign stack.
16030 * config/i386/i386.h (machine_function): Add
16031 max_used_stack_alignment.
16033 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16035 * config/arm/arm.opt (-mbranch-cost): New option.
16036 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16039 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16042 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16043 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16045 2018-01-10 Richard Biener <rguenther@suse.de>
16048 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16049 early out so it also covers the case where we have a non-NULL
16052 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16054 PR tree-optimization/83753
16055 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16056 for non-strided grouped accesses if the number of elements is 1.
16058 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16061 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16062 * i386.h (TARGET_USE_GATHER): Define.
16063 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16065 2018-01-10 Martin Liska <mliska@suse.cz>
16068 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16069 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16071 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16072 CLEANUP_NO_PARTITIONING is not set.
16074 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16076 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16077 for vectors, as a partial revert of r254296.
16078 * rtl.h (const_vec_p): Delete.
16079 (const_vec_duplicate_p): Don't test for vector CONSTs.
16080 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16081 * expmed.c (make_tree): Likewise.
16084 * common.md (E, F): Use CONSTANT_P instead of checking for
16086 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16087 checking for CONST_VECTOR.
16089 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16091 PR middle-end/83575
16092 * predict.c (force_edge_cold): Handle in more sane way edges
16093 with no prediction.
16095 2018-01-09 Carl Love <cel@us.ibm.com>
16097 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16099 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16100 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16101 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16102 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16103 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16104 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16105 * config/rs6000/rs6000-protos.h: Add extern defition for
16106 rs6000_generate_float2_double_code.
16107 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16109 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16110 (float2_v2df): Add define_expand.
16112 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16115 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16116 op_mode in the force_to_mode call.
16118 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16120 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16121 instead of checking each element individually.
16122 (aarch64_evpc_uzp): Likewise.
16123 (aarch64_evpc_zip): Likewise.
16124 (aarch64_evpc_ext): Likewise.
16125 (aarch64_evpc_rev): Likewise.
16126 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16127 instead of checking each element individually. Return true without
16129 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16130 whether all selected elements come from the same input, instead of
16131 checking each element individually. Remove calls to gen_rtx_REG,
16132 start_sequence and end_sequence and instead assert that no rtl is
16135 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16137 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16138 order of HIGH and CONST checks.
16140 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16142 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16143 if the destination isn't an SSA_NAME.
16145 2018-01-09 Richard Biener <rguenther@suse.de>
16147 PR tree-optimization/83668
16148 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16150 (canonicalize_loop_form): ... here, renamed from ...
16151 (canonicalize_loop_closed_ssa_form): ... this and amended to
16152 swap successor edges for loop exit blocks to make us use
16153 the RPO order we need for initial schedule generation.
16155 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16157 PR tree-optimization/64811
16158 * match.pd: When optimizing comparisons with Inf, avoid
16159 introducing or losing exceptions from comparisons with NaN.
16161 2018-01-09 Martin Liska <mliska@suse.cz>
16164 * asan.c (shadow_mem_size): Add gcc_assert.
16166 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16168 Don't save registers in main().
16171 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16172 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16173 * config/avr/avr.c (avr_set_current_function): Don't error if
16174 naked, OS_task or OS_main are specified at the same time.
16175 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16177 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16179 * common/config/avr/avr-common.c (avr_option_optimization_table):
16180 Switch on -mmain-is-OS_task for optimizing compilations.
16182 2018-01-09 Richard Biener <rguenther@suse.de>
16184 PR tree-optimization/83572
16185 * graphite.c: Include cfganal.h.
16186 (graphite_transform_loops): Connect infinite loops to exit
16187 and remove fake edges at the end.
16189 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16191 * ipa-inline.c (edge_badness): Revert accidental checkin.
16193 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16196 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16197 symbols; not inline clones.
16199 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16202 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16203 hard registers. Formatting fixes.
16205 PR preprocessor/83722
16206 * gcc.c (try_generate_repro): Pass
16207 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16208 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16211 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16212 Kito Cheng <kito.cheng@gmail.com>
16214 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16215 (riscv_leaf_function_p): Delete.
16216 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16218 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16220 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16222 (do_ifelse): New function.
16223 (do_isel): New function.
16224 (do_sub3): New function.
16225 (do_add3): New function.
16226 (do_load_mask_compare): New function.
16227 (do_overlap_load_compare): New function.
16228 (expand_compare_loop): New function.
16229 (expand_block_compare): Call expand_compare_loop() when appropriate.
16230 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16231 option description.
16232 (-mblock-compare-inline-loop-limit): New option.
16234 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16237 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16238 Reverse order of second and third operands in first alternative.
16239 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16240 of first and second elements in UNSPEC_VPERMR vector.
16241 (altivec_expand_vec_perm_le): Likewise.
16243 2018-01-08 Jeff Law <law@redhat.com>
16245 PR rtl-optimizatin/81308
16246 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16247 (process_switch): If group_case_labels makes a change, then set
16249 (pass_convert_switch::execute): If a switch is converted, then
16250 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16252 PR rtl-optimization/81308
16253 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16256 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16258 PR target/83663 - Revert r255946
16259 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16260 generation for cases where splatting a value is not useful.
16261 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16262 across a vec_duplicate and a paradoxical subreg forming a vector
16263 mode to a vec_concat.
16265 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16267 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16268 -march=armv8.3-a variants.
16269 * config/arm/t-multilib: Likewise.
16270 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16272 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16274 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16276 (cceq_ior_compare_complement): Give it a name so I can use it, and
16277 change boolean_or_operator predicate to boolean_operator so it can
16278 be used to generate a crand.
16279 (eqne): New code iterator.
16280 (bd/bd_neg): New code_attrs.
16281 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16282 a single define_insn.
16283 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16284 decrement (bdnzt/bdnzf/bdzt/bdzf).
16285 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16286 with the new names of the branch decrement patterns, and added the
16287 names of the branch decrement conditional patterns.
16289 2018-01-08 Richard Biener <rguenther@suse.de>
16291 PR tree-optimization/83563
16292 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16295 2018-01-08 Richard Biener <rguenther@suse.de>
16297 PR middle-end/83713
16298 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16300 2018-01-08 Richard Biener <rguenther@suse.de>
16302 PR tree-optimization/83685
16303 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16304 references to abnormals.
16306 2018-01-08 Richard Biener <rguenther@suse.de>
16309 * dwarf2out.c (output_indirect_strings): Handle empty
16310 skeleton_debug_str_hash.
16311 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16313 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16315 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16316 (emit_store_direct): Likewise.
16317 (arc_trampoline_adjust_address): Likewise.
16318 (arc_asm_trampoline_template): New function.
16319 (arc_initialize_trampoline): Use asm_trampoline_template.
16320 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16321 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16322 * config/arc/arc.md (flush_icache): Delete pattern.
16324 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16326 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16327 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16330 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16333 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16334 by not USED_FOR_TARGET.
16335 (make_pass_resolve_sw_modes): Likewise.
16337 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16339 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16342 2018-01-08 Richard Biener <rguenther@suse.de>
16344 PR middle-end/83580
16345 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16347 2018-01-08 Richard Biener <rguenther@suse.de>
16349 PR middle-end/83517
16350 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16352 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16354 PR middle-end/81897
16355 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16356 basic blocks with a small number of successors.
16357 (convert_control_dep_chain_into_preds): Improve handling of
16359 (dump_predicates): Split apart into...
16360 (dump_pred_chain): ...here...
16361 (dump_pred_info): ...and here.
16362 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16363 (can_chain_union_be_invalidated_p): Improve check for invalidation
16365 (uninit_uses_cannot_happen): Avoid unnecessary if
16366 convert_control_dep_chain_into_preds yielded nothing.
16368 2018-01-06 Martin Sebor <msebor@redhat.com>
16370 PR tree-optimization/83640
16371 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16372 subtracting negative offset from size.
16373 (builtin_access::overlap): Adjust offset bounds of the access to fall
16374 within the size of the object if possible.
16376 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16378 PR rtl-optimization/83699
16379 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16380 extract_bit_field_as_subreg to cases in which the extracted
16381 value is also a vector.
16383 * lra-constraints.c (process_alt_operands): Test for the equivalence
16384 substitutions when detecting a possible reload cycle.
16386 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16389 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16390 by default if flag_selective_schedling{,2}. Formatting fixes.
16392 PR rtl-optimization/83682
16393 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16394 if it has non-VECTOR_MODE element mode.
16395 (vec_duplicate_p): Likewise.
16397 PR middle-end/83694
16398 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16399 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16401 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16404 * config/i386/i386-builtin.def
16405 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16406 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16407 Require also OPTION_MASK_ISA_AVX512F in addition to
16408 OPTION_MASK_ISA_GFNI.
16409 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16410 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16411 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16412 to OPTION_MASK_ISA_GFNI.
16413 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16414 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16415 OPTION_MASK_ISA_AVX512BW.
16416 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16417 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16418 addition to OPTION_MASK_ISA_GFNI.
16419 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16420 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16421 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16422 to OPTION_MASK_ISA_GFNI.
16423 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16424 a requirement for all ISAs rather than any of them with a few
16426 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16428 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16429 bitmasks to be enabled with 3 exceptions, instead of requiring any
16430 enabled ISA with lots of exceptions.
16431 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16432 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16433 Change avx512bw in isa attribute to avx512f.
16434 * config/i386/sgxintrin.h: Add license boilerplate.
16435 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16436 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16437 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16438 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16440 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16441 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16442 temporarily sse2 rather than sse if not enabled already.
16445 * config/i386/sse.md (VI248_VLBW): Rename to ...
16446 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16447 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16448 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16449 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16450 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16451 mode iterator instead of VI248_VLBW.
16453 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16455 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16456 (record_modified): Skip clobbers; add debug output.
16457 (param_change_prob): Use sreal frequencies.
16459 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16461 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16462 punt for user-aligned variables.
16464 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16466 * tree-chrec.c (chrec_contains_symbols): Return true for
16469 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16472 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16473 of (x|y) == x for BICS pattern.
16475 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16477 PR tree-optimization/83605
16478 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16479 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16482 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16484 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16485 * config/epiphany/rtems.h: New file.
16487 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16488 Uros Bizjak <ubizjak@gmail.com>
16491 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16492 QIreg_operand instead of register_operand predicate.
16493 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16494 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16495 comments instead of -fmitigate[-_]rop.
16497 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16500 * cgraphunit.c (symbol_table::compile): Switch to text_section
16501 before calling assembly_start debug hook.
16502 * run-rtl-passes.c (run_rtl_passes): Likewise.
16505 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16507 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16508 range_int_cst_p rather than !symbolic_range_p before calling
16509 extract_range_from_multiplicative_op_1.
16511 2018-01-04 Jeff Law <law@redhat.com>
16513 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16514 redundant test in assertion.
16516 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16518 * doc/rtl.texi: Document machine_mode wrapper classes.
16520 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16522 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16523 using tree_to_uhwi.
16525 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16527 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16528 the VEC_PERM_EXPR fold to fail.
16530 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16533 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16534 to switched_sections.
16536 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16539 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16540 test for d.testing.
16542 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16545 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16546 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16548 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16551 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16552 is BLKmode and bitpos not zero or mode change is needed.
16554 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16557 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16560 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16563 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16564 instead of MULT rtx. Update all corresponding splitters.
16565 (*saddl_se): Ditto.
16566 (*ssub<modesuffix>): Ditto.
16567 (*ssubl_se): Ditto.
16568 (*cmp_sadd_di): Update split patterns.
16569 (*cmp_sadd_si): Ditto.
16570 (*cmp_sadd_sidi): Ditto.
16571 (*cmp_ssub_di): Ditto.
16572 (*cmp_ssub_si): Ditto.
16573 (*cmp_ssub_sidi): Ditto.
16574 * config/alpha/predicates.md (const23_operand): New predicate.
16575 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16576 Look for ASHIFT, not MULT inner operand.
16577 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16579 2018-01-04 Martin Liska <mliska@suse.cz>
16581 PR gcov-profile/83669
16582 * gcov.c (output_intermediate_file): Add version to intermediate
16584 * doc/gcov.texi: Document new field 'version' in intermediate
16585 file format. Fix location of '-k' option of gcov command.
16587 2018-01-04 Martin Liska <mliska@suse.cz>
16590 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16592 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16594 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16596 2018-01-03 Martin Sebor <msebor@redhat.com>
16598 PR tree-optimization/83655
16599 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16600 checking calls with invalid arguments.
16602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16604 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16605 (vectorizable_mask_load_store): Delete.
16606 (vectorizable_call): Return false for masked loads and stores.
16607 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16608 instead of gimple_assign_rhs1.
16609 (vectorizable_load): Handle IFN_MASK_LOAD.
16610 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16614 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16616 (vectorizable_mask_load_store): ...here.
16617 (vectorizable_load): ...and here.
16619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16621 * tree-vect-stmts.c (vect_build_all_ones_mask)
16622 (vect_build_zero_merge_argument): New functions, split out from...
16623 (vectorizable_load): ...here.
16625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16627 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16629 (vectorizable_mask_load_store): ...here.
16630 (vectorizable_store): ...and here.
16632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16634 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16636 (vectorizable_mask_load_store): ...here.
16638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16640 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16641 (vect_model_store_cost): Take a vec_load_store_type instead of a
16643 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16644 (vect_model_store_cost): Take a vec_load_store_type instead of a
16646 (vectorizable_mask_load_store): Update accordingly.
16647 (vectorizable_store): Likewise.
16648 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16652 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16653 IFN_MASK_LOAD calls here rather than...
16654 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16657 Alan Hayward <alan.hayward@arm.com>
16658 David Sherwood <david.sherwood@arm.com>
16660 * expmed.c (extract_bit_field_1): For vector extracts,
16661 fall back to extract_bit_field_as_subreg if vec_extract
16664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16665 Alan Hayward <alan.hayward@arm.com>
16666 David Sherwood <david.sherwood@arm.com>
16668 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16669 they are variable or constant sized.
16670 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16671 slots for constant-sized data.
16673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16674 Alan Hayward <alan.hayward@arm.com>
16675 David Sherwood <david.sherwood@arm.com>
16677 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16678 handling COND_EXPRs with boolean comparisons, try to find a better
16679 basis for the mask type than the boolean itself.
16681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16683 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16684 is calculated and how it can be overridden.
16685 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16686 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16688 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16692 Alan Hayward <alan.hayward@arm.com>
16693 David Sherwood <david.sherwood@arm.com>
16695 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16696 Remove the mode argument.
16697 (aarch64_simd_valid_immediate): Remove the mode and inverse
16699 * config/aarch64/iterators.md (bitsize): New iterator.
16700 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16701 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16702 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16703 aarch64_simd_valid_immediate.
16704 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16705 (aarch64_reg_or_bic_imm): Likewise.
16706 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16707 with an insn_type enum and msl with a modifier_type enum.
16708 Replace element_width with a scalar_mode. Change the shift
16709 to unsigned int. Add constructors for scalar_float_mode and
16710 scalar_int_mode elements.
16711 (aarch64_vect_float_const_representable_p): Delete.
16712 (aarch64_can_const_movi_rtx_p)
16713 (aarch64_simd_scalar_immediate_valid_for_move)
16714 (aarch64_simd_make_constant): Update call to
16715 aarch64_simd_valid_immediate.
16716 (aarch64_advsimd_valid_immediate_hs): New function.
16717 (aarch64_advsimd_valid_immediate): Likewise.
16718 (aarch64_simd_valid_immediate): Remove mode and inverse
16719 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16720 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16721 and aarch64_float_const_representable_p on the result.
16722 (aarch64_output_simd_mov_immediate): Remove mode argument.
16723 Update call to aarch64_simd_valid_immediate and use of
16724 simd_immediate_info.
16725 (aarch64_output_scalar_simd_mov_immediate): Update call
16728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16729 Alan Hayward <alan.hayward@arm.com>
16730 David Sherwood <david.sherwood@arm.com>
16732 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16733 (mode_nunits): Likewise CONST_MODE_NUNITS.
16734 * machmode.def (ADJUST_NUNITS): Document.
16735 * genmodes.c (mode_data::need_nunits_adj): New field.
16736 (blank_mode): Update accordingly.
16737 (adj_nunits): New variable.
16738 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16740 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16741 listed in adj_nunits.
16742 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16743 listed in adj_nunits. Don't emit case statements for such modes.
16744 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16745 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16746 nothing if adj_nunits is nonnull.
16747 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16748 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16749 (emit_mode_fbit): Update use of print_maybe_const_decl.
16750 (emit_move_size): Likewise. Treat the array as non-const
16752 (emit_mode_adjustments): Handle adj_nunits.
16754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16756 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16757 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16758 (VECTOR_MODES): Use it.
16759 (make_vector_modes): Take the prefix as an argument.
16761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16762 Alan Hayward <alan.hayward@arm.com>
16763 David Sherwood <david.sherwood@arm.com>
16765 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16766 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16767 for MODE_VECTOR_BOOL.
16768 * machmode.def (VECTOR_BOOL_MODE): Document.
16769 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16770 (make_vector_bool_mode): New function.
16771 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16773 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16774 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16776 * stor-layout.c (int_mode_for_mode): Likewise.
16777 * tree.c (build_vector_type_for_mode): Likewise.
16778 * varasm.c (output_constant_pool_2): Likewise.
16779 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16780 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16781 for MODE_VECTOR_BOOL.
16782 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16783 of mode class checks.
16784 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16785 instead of a list of mode class checks.
16786 (expand_vector_scalar_condition): Likewise.
16787 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16790 Alan Hayward <alan.hayward@arm.com>
16791 David Sherwood <david.sherwood@arm.com>
16793 * machmode.h (mode_size): Change from unsigned short to
16795 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16796 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16797 or if measurement_type is not polynomial.
16798 (fixed_size_mode::includes_p): Check for constant-sized modes.
16799 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16800 return a poly_uint16 rather than an unsigned short.
16801 (emit_mode_size): Change the type of mode_size from unsigned short
16802 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16803 (emit_mode_adjustments): Cope with polynomial vector sizes.
16804 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16806 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16808 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16809 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16810 * caller-save.c (setup_save_areas): Likewise.
16811 (replace_reg_with_saved_mem): Likewise.
16812 * calls.c (emit_library_call_value_1): Likewise.
16813 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16814 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16815 (gen_lowpart_for_combine): Likewise.
16816 * convert.c (convert_to_integer_1): Likewise.
16817 * cse.c (equiv_constant, cse_insn): Likewise.
16818 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16819 (cselib_subst_to_values): Likewise.
16820 * dce.c (word_dce_process_block): Likewise.
16821 * df-problems.c (df_word_lr_mark_ref): Likewise.
16822 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16823 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16824 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16825 (rtl_for_decl_location): Likewise.
16826 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16827 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16828 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16829 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16830 (expand_expr_real_1): Likewise.
16831 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16832 (pad_below): Likewise.
16833 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16834 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16835 * ira.c (get_subreg_tracking_sizes): Likewise.
16836 * ira-build.c (ira_create_allocno_objects): Likewise.
16837 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16838 (ira_sort_regnos_for_alter_reg): Likewise.
16839 * ira-costs.c (record_operand_costs): Likewise.
16840 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16841 (resolve_simple_move): Likewise.
16842 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16843 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16844 (lra_constraints): Likewise.
16845 (CONST_POOL_OK_P): Reject variable-sized modes.
16846 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16847 (add_pseudo_to_slot, lra_spill): Likewise.
16848 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16849 * optabs-query.c (get_best_extraction_insn): Likewise.
16850 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16851 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16852 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16853 * recog.c (offsettable_address_addr_space_p): Likewise.
16854 * regcprop.c (maybe_mode_change): Likewise.
16855 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16856 * regrename.c (build_def_use): Likewise.
16857 * regstat.c (dump_reg_info): Likewise.
16858 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16859 (find_reloads, find_reloads_subreg_address): Likewise.
16860 * reload1.c (eliminate_regs_1): Likewise.
16861 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16862 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16863 (simplify_binary_operation_1, simplify_subreg): Likewise.
16864 * targhooks.c (default_function_arg_padding): Likewise.
16865 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16866 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16867 (verify_gimple_assign_ternary): Likewise.
16868 * tree-inline.c (estimate_move_cost): Likewise.
16869 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16870 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16871 (get_address_cost_ainc): Likewise.
16872 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16873 (vect_supportable_dr_alignment): Likewise.
16874 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16875 (vectorizable_reduction): Likewise.
16876 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16877 (vectorizable_operation, vectorizable_load): Likewise.
16878 * tree.c (build_same_sized_truth_vector_type): Likewise.
16879 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16880 * var-tracking.c (emit_note_insn_var_location): Likewise.
16881 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16882 (ADDR_VEC_ALIGN): Likewise.
16884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16885 Alan Hayward <alan.hayward@arm.com>
16886 David Sherwood <david.sherwood@arm.com>
16888 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16890 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16891 or if measurement_type is polynomial.
16892 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16893 * combine.c (make_extraction): Likewise.
16894 * dse.c (find_shift_sequence): Likewise.
16895 * dwarf2out.c (mem_loc_descriptor): Likewise.
16896 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16897 (extract_bit_field, extract_low_bits): Likewise.
16898 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16899 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16900 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16901 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16902 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16903 * reload.c (find_reloads): Likewise.
16904 * reload1.c (alter_reg): Likewise.
16905 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16906 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16907 * tree-if-conv.c (predicate_mem_writes): Likewise.
16908 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16909 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16910 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16911 * valtrack.c (dead_debug_insert_temp): Likewise.
16912 * varasm.c (mergeable_constant_section): Likewise.
16913 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16916 Alan Hayward <alan.hayward@arm.com>
16917 David Sherwood <david.sherwood@arm.com>
16919 * expr.c (expand_assignment): Cope with polynomial mode sizes
16920 when assigning to a CONCAT.
16922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16923 Alan Hayward <alan.hayward@arm.com>
16924 David Sherwood <david.sherwood@arm.com>
16926 * machmode.h (mode_precision): Change from unsigned short to
16928 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16930 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16931 or if measurement_type is not polynomial.
16932 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16933 in which the mode is already known to be a scalar_int_mode.
16934 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16935 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16937 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16938 for GET_MODE_PRECISION.
16939 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16940 for GET_MODE_PRECISION.
16941 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16943 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16944 (expand_field_assignment, make_extraction): Likewise.
16945 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16946 (get_last_value): Likewise.
16947 * convert.c (convert_to_integer_1): Likewise.
16948 * cse.c (cse_insn): Likewise.
16949 * expr.c (expand_expr_real_1): Likewise.
16950 * lra-constraints.c (simplify_operand_subreg): Likewise.
16951 * optabs-query.c (can_atomic_load_p): Likewise.
16952 * optabs.c (expand_atomic_load): Likewise.
16953 (expand_atomic_store): Likewise.
16954 * ree.c (combine_reaching_defs): Likewise.
16955 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16956 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16957 * tree.h (type_has_mode_precision_p): Likewise.
16958 * ubsan.c (instrument_si_overflow): Likewise.
16960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16961 Alan Hayward <alan.hayward@arm.com>
16962 David Sherwood <david.sherwood@arm.com>
16964 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16965 polynomial numbers of units.
16966 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16967 (valid_vector_subparts_p): New function.
16968 (build_vector_type): Remove temporary shim and take the number
16969 of units as a poly_uint64 rather than an int.
16970 (build_opaque_vector_type): Take the number of units as a
16971 poly_uint64 rather than an int.
16972 * tree.c (build_vector_from_ctor): Handle polynomial
16973 TYPE_VECTOR_SUBPARTS.
16974 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16975 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16976 (build_vector_from_val): If the number of units is variable,
16977 use build_vec_duplicate_cst for constant operands and
16978 VEC_DUPLICATE_EXPR otherwise.
16979 (make_vector_type): Remove temporary is_constant ().
16980 (build_vector_type, build_opaque_vector_type): Take the number of
16981 units as a poly_uint64 rather than an int.
16982 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16984 * cfgexpand.c (expand_debug_expr): Likewise.
16985 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16986 (store_constructor, expand_expr_real_1): Likewise.
16987 (const_scalar_mask_from_tree): Likewise.
16988 * fold-const-call.c (fold_const_reduction): Likewise.
16989 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16990 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16991 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16992 (fold_relational_const): Likewise.
16993 (native_interpret_vector): Likewise. Change the size from an
16994 int to an unsigned int.
16995 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16996 TYPE_VECTOR_SUBPARTS.
16997 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16998 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16999 duplicating a non-constant operand into a variable-length vector.
17000 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17001 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17002 * ipa-icf.c (sem_variable::equals): Likewise.
17003 * match.pd: Likewise.
17004 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17005 * print-tree.c (print_node): Likewise.
17006 * stor-layout.c (layout_type): Likewise.
17007 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17008 * tree-cfg.c (verify_gimple_comparison): Likewise.
17009 (verify_gimple_assign_binary): Likewise.
17010 (verify_gimple_assign_ternary): Likewise.
17011 (verify_gimple_assign_single): Likewise.
17012 * tree-pretty-print.c (dump_generic_node): Likewise.
17013 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17014 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17015 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17016 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17017 (vect_shift_permute_load_chain): Likewise.
17018 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17019 (expand_vector_condition, optimize_vector_constructor): Likewise.
17020 (lower_vec_perm, get_compute_type): Likewise.
17021 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17022 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17023 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17024 (vect_recog_mask_conversion_pattern): Likewise.
17025 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17026 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17027 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17028 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17029 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17030 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17031 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17032 (supportable_widening_operation): Likewise.
17033 (supportable_narrowing_operation): Likewise.
17034 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17036 * varasm.c (output_constant): Likewise.
17038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17039 Alan Hayward <alan.hayward@arm.com>
17040 David Sherwood <david.sherwood@arm.com>
17042 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17043 so that both the length == 3 and length != 3 cases set up their
17044 own permute vectors. Add comments explaining why we know the
17045 number of elements is constant.
17046 (vect_permute_load_chain): Likewise.
17048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17049 Alan Hayward <alan.hayward@arm.com>
17050 David Sherwood <david.sherwood@arm.com>
17052 * machmode.h (mode_nunits): Change from unsigned char to
17054 (ONLY_FIXED_SIZE_MODES): New macro.
17055 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17056 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17057 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17059 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17060 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17061 or if measurement_type is not polynomial.
17062 * genmodes.c (ZERO_COEFFS): New macro.
17063 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17065 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17066 Use ZERO_COEFFS when emitting initializers.
17067 * data-streamer.h (bp_pack_poly_value): New function.
17068 (bp_unpack_poly_value): Likewise.
17069 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17070 for GET_MODE_NUNITS.
17071 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17072 for GET_MODE_NUNITS.
17073 * tree.c (make_vector_type): Remove temporary shim and make
17074 the real function take the number of units as a poly_uint64
17075 rather than an int.
17076 (build_vector_type_for_mode): Handle polynomial nunits.
17077 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17078 * emit-rtl.c (const_vec_series_p_1): Likewise.
17079 (gen_rtx_CONST_VECTOR): Likewise.
17080 * fold-const.c (test_vec_duplicate_folding): Likewise.
17081 * genrecog.c (validate_pattern): Likewise.
17082 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17083 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17084 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17085 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17086 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17087 * rtlanal.c (subreg_get_info): Likewise.
17088 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17089 (vect_grouped_load_supported): Likewise.
17090 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17091 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17092 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17093 (simplify_const_unary_operation, simplify_binary_operation_1)
17094 (simplify_const_binary_operation, simplify_ternary_operation)
17095 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17096 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17097 instead of CONST_VECTOR_NUNITS.
17098 * varasm.c (output_constant_pool_2): Likewise.
17099 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17100 explicit-encoded elements in the XVEC for variable-length vectors.
17102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17104 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17107 Alan Hayward <alan.hayward@arm.com>
17108 David Sherwood <david.sherwood@arm.com>
17110 * coretypes.h (fixed_size_mode): Declare.
17111 (fixed_size_mode_pod): New typedef.
17112 * builtins.h (target_builtins::x_apply_args_mode)
17113 (target_builtins::x_apply_result_mode): Change type to
17114 fixed_size_mode_pod.
17115 * builtins.c (apply_args_size, apply_result_size, result_vector)
17116 (expand_builtin_apply_args_1, expand_builtin_apply)
17117 (expand_builtin_return): Update accordingly.
17119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17121 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17122 * cselib.c (cselib_hash_rtx): Likewise.
17123 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17124 CONST_VECTOR encoding.
17126 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17127 Jeff Law <law@redhat.com>
17130 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17131 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17132 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17133 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17136 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17137 explicitly probe *sp in a noreturn function if there were any callee
17138 register saves or frame pointer is needed.
17140 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17143 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17144 BLKmode for ternary, binary or unary expressions.
17147 * var-tracking.c (delete_vta_debug_insn): New inline function.
17148 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17149 insns from get_insns () to NULL instead of each bb separately.
17150 Use delete_vta_debug_insn. No longer static.
17151 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17152 delete_vta_debug_insns callers.
17153 * rtl.h (delete_vta_debug_insns): Declare.
17154 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17155 instead of variable_tracking_main.
17157 2018-01-03 Martin Sebor <msebor@redhat.com>
17159 PR tree-optimization/83603
17160 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17161 arguments past the endof the argument list in functions declared
17162 without a prototype.
17163 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17164 Avoid checking when arguments are null.
17166 2018-01-03 Martin Sebor <msebor@redhat.com>
17169 * doc/extend.texi (attribute const): Fix a typo.
17170 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17171 issuing -Wsuggest-attribute for void functions.
17173 2018-01-03 Martin Sebor <msebor@redhat.com>
17175 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17176 offset_int::from instead of wide_int::to_shwi.
17177 (maybe_diag_overlap): Remove assertion.
17178 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17179 * gimple-ssa-sprintf.c (format_directive): Same.
17180 (parse_directive): Same.
17181 (sprintf_dom_walker::compute_format_length): Same.
17182 (try_substitute_return_value): Same.
17184 2018-01-03 Jeff Law <law@redhat.com>
17186 PR middle-end/83654
17187 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17188 non-constant residual for zero at runtime and avoid probing in
17189 that case. Reorganize code for trailing problem to mirror handling
17192 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17194 PR tree-optimization/83501
17195 * tree-ssa-strlen.c (get_string_cst): New.
17196 (handle_char_store): Call get_string_cst.
17198 2018-01-03 Martin Liska <mliska@suse.cz>
17200 PR tree-optimization/83593
17201 * tree-ssa-strlen.c: Include tree-cfg.h.
17202 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17203 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17204 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17206 (strlen_dom_walker::before_dom_children): Call
17207 gimple_purge_dead_eh_edges. Dump tranformation with details
17209 (strlen_dom_walker::before_dom_children): Update call by adding
17210 new argument cleanup_eh.
17211 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17213 2018-01-03 Martin Liska <mliska@suse.cz>
17216 * cif-code.def (VARIADIC_THUNK): New enum value.
17217 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17220 2018-01-03 Jan Beulich <jbeulich@suse.com>
17222 * sse.md (mov<mode>_internal): Tighten condition for when to use
17223 vmovdqu<ssescalarsize> for TI and OI modes.
17225 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17227 Update copyright years.
17229 2018-01-03 Martin Liska <mliska@suse.cz>
17232 * ipa-visibility.c (function_and_variable_visibility): Skip
17233 functions with noipa attribure.
17235 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17237 * gcc.c (process_command): Update copyright notice dates.
17238 * gcov-dump.c (print_version): Ditto.
17239 * gcov.c (print_version): Ditto.
17240 * gcov-tool.c (print_version): Ditto.
17241 * gengtype.c (create_file): Ditto.
17242 * doc/cpp.texi: Bump @copying's copyright year.
17243 * doc/cppinternals.texi: Ditto.
17244 * doc/gcc.texi: Ditto.
17245 * doc/gccint.texi: Ditto.
17246 * doc/gcov.texi: Ditto.
17247 * doc/install.texi: Ditto.
17248 * doc/invoke.texi: Ditto.
17250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17252 * vector-builder.h (vector_builder::m_full_nelts): Change from
17253 unsigned int to poly_uint64.
17254 (vector_builder::full_nelts): Update prototype accordingly.
17255 (vector_builder::new_vector): Likewise.
17256 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17257 (vector_builder::operator ==): Likewise.
17258 (vector_builder::finalize): Likewise.
17259 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17260 Take the number of elements as a poly_uint64 rather than an
17262 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17263 from unsigned int to poly_uint64.
17264 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17265 (vec_perm_indices::new_vector): Likewise.
17266 (vec_perm_indices::length): Likewise.
17267 (vec_perm_indices::nelts_per_input): Likewise.
17268 (vec_perm_indices::input_nelts): Likewise.
17269 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17270 number of elements per input as a poly_uint64 rather than an
17271 unsigned int. Use the original encoding for variable-length
17272 vectors, rather than clamping each individual element.
17273 For the second and subsequent elements in each pattern,
17274 clamp the step and base before clamping their sum.
17275 (vec_perm_indices::series_p): Handle polynomial element counts.
17276 (vec_perm_indices::all_in_range_p): Likewise.
17277 (vec_perm_indices_to_tree): Likewise.
17278 (vec_perm_indices_to_rtx): Likewise.
17279 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17280 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17281 (tree_vector_builder::new_binary_operation): Handle polynomial
17282 element counts. Return false if we need to know the number
17283 of elements at compile time.
17284 * fold-const.c (fold_vec_perm): Punt if the number of elements
17285 isn't known at compile time.
17287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17289 * vec-perm-indices.h (vec_perm_builder): Change element type
17290 from HOST_WIDE_INT to poly_int64.
17291 (vec_perm_indices::element_type): Update accordingly.
17292 (vec_perm_indices::clamp): Handle polynomial element_types.
17293 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17294 (vec_perm_indices::all_in_range_p): Likewise.
17295 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17297 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17298 polynomial vec_perm_indices element types.
17299 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17300 * fold-const.c (fold_vec_perm): Likewise.
17301 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17302 * tree-vect-generic.c (lower_vec_perm): Likewise.
17303 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17304 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17305 element type to HOST_WIDE_INT.
17307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17308 Alan Hayward <alan.hayward@arm.com>
17309 David Sherwood <david.sherwood@arm.com>
17311 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17312 rather than an int. Use plus_constant.
17313 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17314 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17317 Alan Hayward <alan.hayward@arm.com>
17318 David Sherwood <david.sherwood@arm.com>
17320 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17321 a HOST_WIDE_INT to a poly_int64.
17323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17324 Alan Hayward <alan.hayward@arm.com>
17325 David Sherwood <david.sherwood@arm.com>
17327 * calls.c (load_register_parameters): Cope with polynomial
17328 mode sizes. Require a constant size for BLKmode parameters
17329 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17330 forces a parameter to be padded at the lsb end in order to
17331 fill a complete number of words, require the parameter size
17332 to be ordered wrt UNITS_PER_WORD.
17334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17335 Alan Hayward <alan.hayward@arm.com>
17336 David Sherwood <david.sherwood@arm.com>
17338 * reload1.c (spill_stack_slot_width): Change element type
17339 from unsigned int to poly_uint64_pod.
17340 (alter_reg): Treat mode sizes as polynomial.
17342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17343 Alan Hayward <alan.hayward@arm.com>
17344 David Sherwood <david.sherwood@arm.com>
17346 * reload.c (complex_word_subreg_p): New function.
17347 (reload_inner_reg_of_subreg, push_reload): Use it.
17349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17350 Alan Hayward <alan.hayward@arm.com>
17351 David Sherwood <david.sherwood@arm.com>
17353 * lra-constraints.c (process_alt_operands): Reject matched
17354 operands whose sizes aren't ordered.
17355 (match_reload): Refer to this check here.
17357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17358 Alan Hayward <alan.hayward@arm.com>
17359 David Sherwood <david.sherwood@arm.com>
17361 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17362 that the mode size is in the set {1, 2, 4, 8, 16}.
17364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17365 Alan Hayward <alan.hayward@arm.com>
17366 David Sherwood <david.sherwood@arm.com>
17368 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17369 Use plus_constant instead of gen_rtx_PLUS.
17371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17372 Alan Hayward <alan.hayward@arm.com>
17373 David Sherwood <david.sherwood@arm.com>
17375 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17376 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17377 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17378 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17379 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17380 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17381 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17382 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17383 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17384 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17386 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17387 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17388 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17389 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17390 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17391 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17392 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17393 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17394 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17395 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17397 * expr.c (emit_move_resolve_push): Treat the input and result
17398 of PUSH_ROUNDING as a poly_int64.
17399 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17400 (emit_push_insn): Likewise.
17401 * lra-eliminations.c (mark_not_eliminable): Likewise.
17402 * recog.c (push_operand): Likewise.
17403 * reload1.c (elimination_effects): Likewise.
17404 * rtlanal.c (nonzero_bits1): Likewise.
17405 * calls.c (store_one_arg): Likewise. Require the padding to be
17406 known at compile time.
17408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17409 Alan Hayward <alan.hayward@arm.com>
17410 David Sherwood <david.sherwood@arm.com>
17412 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17413 Use plus_constant instead of gen_rtx_PLUS.
17415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17416 Alan Hayward <alan.hayward@arm.com>
17417 David Sherwood <david.sherwood@arm.com>
17419 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17420 rather than an int.
17422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17423 Alan Hayward <alan.hayward@arm.com>
17424 David Sherwood <david.sherwood@arm.com>
17426 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17427 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17428 via stack temporaries. Treat the mode size as polynomial too.
17430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17431 Alan Hayward <alan.hayward@arm.com>
17432 David Sherwood <david.sherwood@arm.com>
17434 * expr.c (expand_expr_real_2): When handling conversions involving
17435 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17436 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17437 as a poly_uint64 too.
17439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17440 Alan Hayward <alan.hayward@arm.com>
17441 David Sherwood <david.sherwood@arm.com>
17443 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17446 Alan Hayward <alan.hayward@arm.com>
17447 David Sherwood <david.sherwood@arm.com>
17449 * combine.c (can_change_dest_mode): Handle polynomial
17450 REGMODE_NATURAL_SIZE.
17451 * expmed.c (store_bit_field_1): Likewise.
17452 * expr.c (store_constructor): Likewise.
17453 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17454 and polynomial REGMODE_NATURAL_SIZE.
17455 (gen_lowpart_common): Likewise.
17456 * reginfo.c (record_subregs_of_mode): Likewise.
17457 * rtlanal.c (read_modify_subreg_p): Likewise.
17459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17460 Alan Hayward <alan.hayward@arm.com>
17461 David Sherwood <david.sherwood@arm.com>
17463 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17464 numbers of elements.
17466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17467 Alan Hayward <alan.hayward@arm.com>
17468 David Sherwood <david.sherwood@arm.com>
17470 * match.pd: Cope with polynomial numbers of vector elements.
17472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17473 Alan Hayward <alan.hayward@arm.com>
17474 David Sherwood <david.sherwood@arm.com>
17476 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17477 in a POINTER_PLUS_EXPR.
17479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17480 Alan Hayward <alan.hayward@arm.com>
17481 David Sherwood <david.sherwood@arm.com>
17483 * omp-simd-clone.c (simd_clone_subparts): New function.
17484 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17485 (ipa_simd_modify_function_body): Likewise.
17487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17488 Alan Hayward <alan.hayward@arm.com>
17489 David Sherwood <david.sherwood@arm.com>
17491 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17492 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17493 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17494 (expand_vector_condition, vector_element): Likewise.
17495 (subparts_gt): New function.
17496 (get_compute_type): Use subparts_gt.
17497 (count_type_subparts): Delete.
17498 (expand_vector_operations_1): Use subparts_gt instead of
17499 count_type_subparts.
17501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17502 Alan Hayward <alan.hayward@arm.com>
17503 David Sherwood <david.sherwood@arm.com>
17505 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17506 (vect_compile_time_alias): ...this new function. Do the calculation
17507 on poly_ints rather than trees.
17508 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
17516 (vect_schedule_slp_instance): Likewise.
17518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17519 Alan Hayward <alan.hayward@arm.com>
17520 David Sherwood <david.sherwood@arm.com>
17522 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17523 constant and extern definitions for variable-length vectors.
17524 (vect_get_constant_vectors): Note that the number of units
17525 is known to be constant.
17527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17528 Alan Hayward <alan.hayward@arm.com>
17529 David Sherwood <david.sherwood@arm.com>
17531 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17532 of units as polynomial. Choose between WIDE and NARROW based
17535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17536 Alan Hayward <alan.hayward@arm.com>
17537 David Sherwood <david.sherwood@arm.com>
17539 * tree-vect-stmts.c (simd_clone_subparts): New function.
17540 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17543 Alan Hayward <alan.hayward@arm.com>
17544 David Sherwood <david.sherwood@arm.com>
17546 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17547 vectors as polynomial. Use build_index_vector for
17548 IFN_GOMP_SIMD_LANE.
17550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17551 Alan Hayward <alan.hayward@arm.com>
17552 David Sherwood <david.sherwood@arm.com>
17554 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17555 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17556 for variable-length vectors.
17557 (vectorizable_mask_load_store): Treat the number of units as
17558 polynomial, asserting that it is constant if the condition has
17559 already been enforced.
17560 (vectorizable_store, vectorizable_load): Likewise.
17562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17563 Alan Hayward <alan.hayward@arm.com>
17564 David Sherwood <david.sherwood@arm.com>
17566 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17567 of units as polynomial. Punt if we can't tell at compile time
17568 which vector contains the final result.
17570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17571 Alan Hayward <alan.hayward@arm.com>
17572 David Sherwood <david.sherwood@arm.com>
17574 * tree-vect-loop.c (vectorizable_induction): Treat the number
17575 of units as polynomial. Punt on SLP inductions. Use an integer
17576 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17577 cast of such a series for variable-length floating-point
17580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17581 Alan Hayward <alan.hayward@arm.com>
17582 David Sherwood <david.sherwood@arm.com>
17584 * tree.h (build_index_vector): Declare.
17585 * tree.c (build_index_vector): New function.
17586 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17587 of units as polynomial, forcibly converting it to a constant if
17588 vectorizable_reduction has already enforced the condition.
17589 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17590 to create a {1,2,3,...} vector.
17591 (vectorizable_reduction): Treat the number of units as polynomial.
17592 Choose vectype_in based on the largest scalar element size rather
17593 than the smallest number of units. Enforce the restrictions
17596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17597 Alan Hayward <alan.hayward@arm.com>
17598 David Sherwood <david.sherwood@arm.com>
17600 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17601 number of units as polynomial.
17603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17604 Alan Hayward <alan.hayward@arm.com>
17605 David Sherwood <david.sherwood@arm.com>
17607 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17608 * target.def (autovectorize_vector_sizes): Return the vector sizes
17609 by pointer, using vector_sizes rather than a bitmask.
17610 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17611 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17612 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17614 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17615 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17616 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17617 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17618 * omp-general.c (omp_max_vf): Likewise.
17619 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17620 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17621 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17622 * tree-vect-slp.c (vect_slp_bb): Likewise.
17623 * doc/tm.texi: Regenerate.
17624 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17626 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17627 the vector size as a poly_uint64 rather than an unsigned int.
17628 (current_vector_size): Change from an unsigned int to a poly_uint64.
17629 (get_vectype_for_scalar_type): Update accordingly.
17630 * tree.h (build_truth_vector_type): Take the size and number of
17631 units as a poly_uint64 rather than an unsigned int.
17632 (build_vector_type): Add a temporary overload that takes
17633 the number of units as a poly_uint64 rather than an unsigned int.
17634 * tree.c (make_vector_type): Likewise.
17635 (build_truth_vector_type): Take the number of units as a poly_uint64
17636 rather than an unsigned int.
17638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17639 Alan Hayward <alan.hayward@arm.com>
17640 David Sherwood <david.sherwood@arm.com>
17642 * target.def (get_mask_mode): Take the number of units and length
17643 as poly_uint64s rather than unsigned ints.
17644 * targhooks.h (default_get_mask_mode): Update accordingly.
17645 * targhooks.c (default_get_mask_mode): Likewise.
17646 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17647 * doc/tm.texi: Regenerate.
17649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17650 Alan Hayward <alan.hayward@arm.com>
17651 David Sherwood <david.sherwood@arm.com>
17653 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17654 * omp-general.c (omp_max_vf): Likewise.
17655 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17656 (expand_omp_simd): Handle polynomial safelen.
17657 * omp-low.c (omplow_simd_context): Add a default constructor.
17658 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17659 (lower_rec_simd_input_clauses): Update accordingly.
17660 (lower_rec_input_clauses): Likewise.
17662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17663 Alan Hayward <alan.hayward@arm.com>
17664 David Sherwood <david.sherwood@arm.com>
17666 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17667 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17668 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17669 (vect_analyze_slp_cost): Likewise.
17670 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17671 (vect_model_load_cost): Likewise.
17673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17674 Alan Hayward <alan.hayward@arm.com>
17675 David Sherwood <david.sherwood@arm.com>
17677 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17678 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17679 from an unsigned int * to a poly_uint64_pod *.
17680 (calculate_unrolling_factor): New function.
17681 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17684 Alan Hayward <alan.hayward@arm.com>
17685 David Sherwood <david.sherwood@arm.com>
17687 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17688 from an unsigned int to a poly_uint64.
17689 (_loop_vec_info::slp_unrolling_factor): Likewise.
17690 (_loop_vec_info::vectorization_factor): Change from an int
17692 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17693 (vect_get_num_vectors): New function.
17694 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17695 (vect_get_num_copies): Use vect_get_num_vectors.
17696 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17697 to an unsigned int *.
17698 (vect_analyze_data_refs): Change min_vf from an int * to a
17700 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17701 than an unsigned HOST_WIDE_INT.
17702 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17703 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17704 to an unsigned int *.
17705 (vect_analyze_data_ref_dependences): Likewise.
17706 (vect_compute_data_ref_alignment): Handle polynomial vf.
17707 (vect_enhance_data_refs_alignment): Likewise.
17708 (vect_prune_runtime_alias_test_list): Likewise.
17709 (vect_shift_permute_load_chain): Likewise.
17710 (vect_supportable_dr_alignment): Likewise.
17711 (dependence_distance_ge_vf): Take the vectorization factor as a
17712 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17713 (vect_analyze_data_refs): Change min_vf from an int * to a
17715 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17716 vfm1 as a poly_uint64 rather than an int. Make the same change
17717 for the returned bound_scalar.
17718 (vect_gen_vector_loop_niters): Handle polynomial vf.
17719 (vect_do_peeling): Likewise. Update call to
17720 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17721 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17723 * tree-vect-loop.c (vect_determine_vectorization_factor)
17724 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17725 (vect_get_known_peeling_cost): Likewise.
17726 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17727 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17728 (vect_transform_loop): Likewise. Use the lowest possible VF when
17729 updating the upper bounds of the loop.
17730 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17731 rather than an int.
17732 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17733 polynomial unroll factors.
17734 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17735 (vect_make_slp_decision): Likewise.
17736 (vect_supported_load_permutation_p): Likewise, and polynomial
17738 (vect_analyze_slp_cost): Handle polynomial vf.
17739 (vect_slp_analyze_node_operations): Likewise.
17740 (vect_slp_analyze_bb_1): Likewise.
17741 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17742 than an unsigned HOST_WIDE_INT.
17743 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17744 (vectorizable_load): Handle polynomial vf.
17745 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17747 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17750 Alan Hayward <alan.hayward@arm.com>
17751 David Sherwood <david.sherwood@arm.com>
17753 * match.pd: Handle bit operations involving three constants
17754 and try to fold one pair.
17756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17758 * tree-vect-loop-manip.c: Include gimple-fold.h.
17759 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17760 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17761 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17762 Add a path that uses a step of VF instead of 1, but disable it
17764 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17765 and niters_no_overflow parameters. Update calls to
17766 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17767 Create a new SSA name if the latter choses to use a ste other
17768 than zero, and return it via niters_vector_mult_vf_var.
17769 * tree-vect-loop.c (vect_transform_loop): Update calls to
17770 vect_do_peeling, vect_gen_vector_loop_niters and
17771 slpeel_make_loop_iterate_ntimes.
17772 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17773 (vect_gen_vector_loop_niters): Update declarations after above changes.
17775 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17777 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17778 128-bit round to integer instructions.
17779 (ceil<mode>2): Likewise.
17780 (btrunc<mode>2): Likewise.
17781 (round<mode>2): Likewise.
17783 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17785 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17786 unaligned VSX load/store on P8/P9.
17787 (expand_block_clear): Allow the use of unaligned VSX
17788 load/store on P8/P9.
17790 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17792 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17794 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17795 swap associated with both a load and a store.
17797 2018-01-02 Andrew Waterman <andrew@sifive.com>
17799 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17800 * config/riscv/riscv.md (clear_cache): Use it.
17802 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17804 * web.c: Remove out-of-date comment.
17806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17808 * expr.c (fixup_args_size_notes): Check that any existing
17809 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17810 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17811 (emit_single_push_insn): ...here.
17813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17815 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17816 (const_vector_encoded_nelts): New function.
17817 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17818 (const_vector_int_elt, const_vector_elt): Declare.
17819 * emit-rtl.c (const_vector_int_elt_1): New function.
17820 (const_vector_elt): Likewise.
17821 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17822 of CONST_VECTOR_ELT.
17824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17826 * expr.c: Include rtx-vector-builder.h.
17827 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17828 directly on the tree encoding.
17829 (const_vector_from_tree): Likewise.
17830 * optabs.c: Include rtx-vector-builder.h.
17831 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17832 sequence of "u" values.
17833 * vec-perm-indices.c: Include rtx-vector-builder.h.
17834 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17835 directly on the vec_perm_indices encoding.
17837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17839 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17840 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17841 * rtx-vector-builder.h: New file.
17842 * rtx-vector-builder.c: Likewise.
17843 * rtl.h (rtx_def::u2): Add a const_vector field.
17844 (CONST_VECTOR_NPATTERNS): New macro.
17845 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17846 (CONST_VECTOR_DUPLICATE_P): Likewise.
17847 (CONST_VECTOR_STEPPED_P): Likewise.
17848 (CONST_VECTOR_ENCODED_ELT): Likewise.
17849 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17850 (unwrap_const_vec_duplicate): Likewise.
17851 (const_vec_series_p): Check for a non-duplicated vector encoding.
17852 Say that the function only returns true for integer vectors.
17853 * emit-rtl.c: Include rtx-vector-builder.h.
17854 (gen_const_vec_duplicate_1): Delete.
17855 (gen_const_vector): Call gen_const_vec_duplicate instead of
17856 gen_const_vec_duplicate_1.
17857 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17858 (gen_const_vec_duplicate): Use rtx_vector_builder.
17859 (gen_const_vec_series): Likewise.
17860 (gen_rtx_CONST_VECTOR): Likewise.
17861 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17862 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17863 Build a new vector rather than modifying a CONST_VECTOR in-place.
17864 (handle_special_swappables): Update call accordingly.
17865 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17866 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17867 Build a new vector rather than modifying a CONST_VECTOR in-place.
17868 (handle_special_swappables): Update call accordingly.
17870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17872 * simplify-rtx.c (simplify_const_binary_operation): Use
17873 CONST_VECTOR_ELT instead of XVECEXP.
17875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17877 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17878 the selector elements to be different from the data elements
17879 if the selector is a VECTOR_CST.
17880 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17881 ssizetype for the selector.
17883 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17885 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17886 before testing each element individually.
17887 * tree-vect-generic.c (lower_vec_perm): Likewise.
17889 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17891 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17892 * selftest-run-tests.c (selftest::run_tests): Call it.
17893 * vector-builder.h (vector_builder::operator ==): New function.
17894 (vector_builder::operator !=): Likewise.
17895 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17896 (vec_perm_indices::all_from_input_p): New function.
17897 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17898 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17899 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17900 instead of reading the VECTOR_CST directly. Detect whether both
17901 vector inputs are the same before constructing the vec_perm_indices,
17902 and update the number of inputs argument accordingly. Use the
17903 utility functions added above. Only construct sel2 if we need to.
17905 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17907 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17908 the broadcast of the low byte.
17909 (expand_mult_highpart): Use an explicit encoding for the permutes.
17910 * optabs-query.c (can_mult_highpart_p): Likewise.
17911 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17912 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17913 (vectorizable_bswap): Likewise.
17914 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17915 explicit encoding for the power-of-2 permutes.
17916 (vect_permute_store_chain): Likewise.
17917 (vect_grouped_load_supported): Likewise.
17918 (vect_permute_load_chain): Likewise.
17920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17922 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17923 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17924 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17925 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17926 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17927 (vect_gen_perm_mask_any): Likewise.
17929 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17931 * int-vector-builder.h: New file.
17932 * vec-perm-indices.h: Include int-vector-builder.h.
17933 (vec_perm_indices): Redefine as an int_vector_builder.
17934 (auto_vec_perm_indices): Delete.
17935 (vec_perm_builder): Redefine as a stand-alone class.
17936 (vec_perm_indices::vec_perm_indices): New function.
17937 (vec_perm_indices::clamp): Likewise.
17938 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17939 (vec_perm_indices::new_vector): New function.
17940 (vec_perm_indices::new_expanded_vector): Update for new
17941 vec_perm_indices class.
17942 (vec_perm_indices::rotate_inputs): New function.
17943 (vec_perm_indices::all_in_range_p): Operate directly on the
17944 encoded form, without computing elided elements.
17945 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17946 encoding. Update for new vec_perm_indices class.
17947 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17948 the given vec_perm_builder.
17949 (expand_vec_perm_var): Update vec_perm_builder constructor.
17950 (expand_mult_highpart): Use vec_perm_builder instead of
17951 auto_vec_perm_indices.
17952 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17953 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17954 or double series encoding as appropriate.
17955 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17956 vec_perm_indices instead of auto_vec_perm_indices.
17957 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17958 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17959 (vect_permute_store_chain): Likewise.
17960 (vect_grouped_load_supported): Likewise.
17961 (vect_permute_load_chain): Likewise.
17962 (vect_shift_permute_load_chain): Likewise.
17963 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17964 (vect_transform_slp_perm_load): Likewise.
17965 (vect_schedule_slp_instance): Likewise.
17966 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17967 (vectorizable_mask_load_store): Likewise.
17968 (vectorizable_bswap): Likewise.
17969 (vectorizable_store): Likewise.
17970 (vectorizable_load): Likewise.
17971 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17972 vec_perm_indices instead of auto_vec_perm_indices. Use
17973 tree_to_vec_perm_builder to read the vector from a tree.
17974 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17975 vec_perm_builder instead of a vec_perm_indices.
17976 (have_whole_vector_shift): Use vec_perm_builder and
17977 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17978 truncation to calc_vec_perm_mask_for_shift.
17979 (vect_create_epilog_for_reduction): Likewise.
17980 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17981 from auto_vec_perm_indices to vec_perm_indices.
17982 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17983 instead of changing individual elements.
17984 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17985 the vector in d.perm.
17986 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17987 from auto_vec_perm_indices to vec_perm_indices.
17988 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17989 instead of changing individual elements.
17990 (arm_vectorize_vec_perm_const): Use new_vector to install
17991 the vector in d.perm.
17992 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17993 Update vec_perm_builder constructor.
17994 (rs6000_expand_interleave): Likewise.
17995 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17996 (rs6000_expand_interleave): Likewise.
17998 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18000 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18001 to qimode could truncate the indices.
18002 * optabs.c (expand_vec_perm_var): Likewise.
18004 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18006 * Makefile.in (OBJS): Add vec-perm-indices.o.
18007 * vec-perm-indices.h: New file.
18008 * vec-perm-indices.c: Likewise.
18009 * target.h (vec_perm_indices): Replace with a forward class
18011 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18012 * optabs.h: Include vec-perm-indices.h.
18013 (expand_vec_perm): Delete.
18014 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18015 (expand_vec_perm_const): Declare.
18016 * target.def (vec_perm_const_ok): Replace with...
18017 (vec_perm_const): ...this new hook.
18018 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18019 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18020 * doc/tm.texi: Regenerate.
18021 * optabs.def (vec_perm_const): Delete.
18022 * doc/md.texi (vec_perm_const): Likewise.
18023 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18024 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18025 expand_vec_perm for constant permutation vectors. Assert that
18026 the mode of variable permutation vectors is the integer equivalent
18027 of the mode that is being permuted.
18028 * optabs-query.h (selector_fits_mode_p): Declare.
18029 * optabs-query.c: Include vec-perm-indices.h.
18030 (selector_fits_mode_p): New function.
18031 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18032 is defined, instead of checking whether the vec_perm_const_optab
18033 exists. Use targetm.vectorize.vec_perm_const instead of
18034 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18035 fit in the vector mode before using a variable permute.
18036 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18037 vec_perm_indices instead of an rtx.
18038 (expand_vec_perm): Replace with...
18039 (expand_vec_perm_const): ...this new function. Take the selector
18040 as a vec_perm_indices rather than an rtx. Also take the mode of
18041 the selector. Update call to shift_amt_for_vec_perm_mask.
18042 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18043 Use vec_perm_indices::new_expanded_vector to expand the original
18044 selector into bytes. Check whether the indices fit in the vector
18045 mode before using a variable permute.
18046 (expand_vec_perm_var): Make global.
18047 (expand_mult_highpart): Use expand_vec_perm_const.
18048 * fold-const.c: Includes vec-perm-indices.h.
18049 * tree-ssa-forwprop.c: Likewise.
18050 * tree-vect-data-refs.c: Likewise.
18051 * tree-vect-generic.c: Likewise.
18052 * tree-vect-loop.c: Likewise.
18053 * tree-vect-slp.c: Likewise.
18054 * tree-vect-stmts.c: Likewise.
18055 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18057 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18058 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18059 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18060 (aarch64_vectorize_vec_perm_const): ...this new function.
18061 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18062 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18063 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18064 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18065 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18066 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18067 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18069 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18070 check for NEON modes.
18071 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18072 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18073 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18074 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18076 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18077 the old VEC_PERM_CONST conditions.
18078 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18079 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18080 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18081 (ia64_vectorize_vec_perm_const_ok): Merge into...
18082 (ia64_vectorize_vec_perm_const): ...this new function.
18083 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18084 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18085 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18086 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18087 * config/mips/mips.c (mips_expand_vec_perm_const)
18088 (mips_vectorize_vec_perm_const_ok): Merge into...
18089 (mips_vectorize_vec_perm_const): ...this new function.
18090 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18091 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18092 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18093 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18094 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18095 (rs6000_expand_vec_perm_const): Delete.
18096 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18098 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18099 (altivec_expand_vec_perm_const_le): Take each operand individually.
18100 Operate on constant selectors rather than rtxes.
18101 (altivec_expand_vec_perm_const): Likewise. Update call to
18102 altivec_expand_vec_perm_const_le.
18103 (rs6000_expand_vec_perm_const): Delete.
18104 (rs6000_vectorize_vec_perm_const_ok): Delete.
18105 (rs6000_vectorize_vec_perm_const): New function.
18106 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18107 an element count and rtx array.
18108 (rs6000_expand_extract_even): Update call accordingly.
18109 (rs6000_expand_interleave): Likewise.
18110 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18111 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18112 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18113 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18114 (rs6000_expand_vec_perm_const): Delete.
18115 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18116 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18117 (altivec_expand_vec_perm_const_le): Take each operand individually.
18118 Operate on constant selectors rather than rtxes.
18119 (altivec_expand_vec_perm_const): Likewise. Update call to
18120 altivec_expand_vec_perm_const_le.
18121 (rs6000_expand_vec_perm_const): Delete.
18122 (rs6000_vectorize_vec_perm_const_ok): Delete.
18123 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18124 reference to the SPE evmerge intructions.
18125 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18126 an element count and rtx array.
18127 (rs6000_expand_extract_even): Update call accordingly.
18128 (rs6000_expand_interleave): Likewise.
18129 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18130 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18132 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18136 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18137 vector mode and that that mode matches the mode of the data
18139 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18140 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18141 directly using expand_vec_perm_1 when forcing selectors into
18143 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18147 * optabs-query.h (can_vec_perm_p): Delete.
18148 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18149 * optabs-query.c (can_vec_perm_p): Split into...
18150 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18151 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18152 particular selector is valid.
18153 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18154 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18155 (vect_grouped_load_supported): Likewise.
18156 (vect_shift_permute_load_chain): Likewise.
18157 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18158 (vect_transform_slp_perm_load): Likewise.
18159 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18160 (vectorizable_bswap): Likewise.
18161 (vect_gen_perm_mask_checked): Likewise.
18162 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18163 implementations of variable permutation vectors into account
18164 when deciding which selector to use.
18165 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18166 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18167 with a false third argument.
18168 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18169 to test whether the constant selector is valid and can_vec_perm_var_p
18170 to test whether a variable selector is valid.
18172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18174 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18175 * optabs-query.c (can_vec_perm_p): Likewise.
18176 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18177 instead of vec_perm_indices.
18178 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18179 (vect_gen_perm_mask_checked): Likewise,
18180 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18181 (vect_gen_perm_mask_checked): Likewise,
18183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18185 * optabs-query.h (qimode_for_vec_perm): Declare.
18186 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18187 (qimode_for_vec_perm): ...this new function.
18188 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18190 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18192 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18193 does not have a conditional at the top.
18195 2018-01-02 Richard Biener <rguenther@suse.de>
18197 * ipa-inline.c (big_speedup_p): Fix expression.
18199 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18202 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18205 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18209 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18210 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18211 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18212 cond_taken_branch_cost 3->4.
18214 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18216 PR tree-optimization/83581
18217 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18218 TODO_cleanup_cfg if any changes have been made.
18220 PR middle-end/83608
18221 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18222 convert_modes if target mode has the right side, but different mode
18225 PR middle-end/83609
18226 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18227 last argument when extracting from CONCAT. If either from_real or
18228 from_imag is NULL, use expansion through memory. If result is not
18229 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18230 the parts directly to inner mode, if even that fails, use expansion
18233 PR middle-end/83623
18234 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18235 check for bswap in mode rather than HImode and use that in expand_unop
18238 Copyright (C) 2018 Free Software Foundation, Inc.
18240 Copying and distribution of this file, with or without modification,
18241 are permitted in any medium without royalty provided the copyright
18242 notice and this notice are preserved.